 So thank you for the invitation to come and talk about sense. So the idea here is that I will tell you a bit about the basic principles of sense so that even if you don't know the details about instrumentation, if you've never done an experiment, you will at least know if it's useful for your experiment or for your project. So if you should think about using it or not, and of course then you can contact an instrument scientist at your favorite nutrient center and sort out the details and discuss your specific experiment. So today I will tell you about what a sense experiment is and some of it you already heard with Frank Cabell. So I can question you about it to see if you understood everything. Now with some of it you will already know, but we'll go over it real quick. I will also tell you about the sense scattering process of the different contributions and basically how you get your signal and why you would use it. So we'll look into the typical sense data analysis that you do for more for biopolymer solutions. And then we'll talk about data quality and I'll give you some recommendations with when you're planning a sense experiment. Okay, so in life sciences and biomaterials, sense is a very powerful technique because typical samples have a very large hierarchy of structures so you can go from small polymers or small protein, let's say, to a very large structure, you know, as big as a virus, for example, or even a cell. So sense is very powerful because you can cover a broad range of sizes. In terms of small angles scattering in general if you want to compare to compare with complimentary probes and depending on your background this, this may be useful for you to understand the where neutrons are placed in terms of the experimental range. So x-rays typically use slightly shorter wavelength ranges. Neutrons cover the range that you see here from one to about 20 angstroms and light scattering will cover larger sizes so you can imagine that you are looking at different sizes of things, even though there's a lot of overlap with the different types. Now of course also depending on the probe that you use, the exact information that you're extracting what you're measuring is not exactly the same so even though I tell you that it overlaps, the interactions are different so the x-rays are interacting with electrons, neutrons are interacting with nuclei and with light scattering you get a signal from differences in refractive index. So you're not quite measuring the exact same thing even though of course your samples is the same. The nature of the radiation will determine for example the sample environment that you can use. So neutrons for example are a non-destructive probe so you may be able to for example use a pressure cell or other environments because they will cross different materials. The lens scales that can be probed are different and even the nature of the information. So what can you measure with small angle scattering in general? So there are a lot of structural aspects that you can measure from size, shape, even fractal dimension. We're going to go over this a little bit. And also phenomena you can look at self-assembly, phase transitions, thermodynamics, kinetics, so a lot of different phenomena. And it may seem a bit overwhelming if you've never used it to know if you can use it or not for your research but that's why we're here to give you some background and of course you can always talk to instrument scientists about your project. Okay, so if you've never done a sun's experiment what is it? So the basics of it is shown here in this schematics. So you'll have incoming neutrons and this can be from a spallation or a reactive based source. I will not go over too much about how you produce neutrons. I'm going to discuss it if you want afterwards. There will be some way in which you can select the wavelength or the wavelength range that you want to use. Here I'm showing just the velocity selector that would select the wavelength of interest. There'll be a source aperture that defines the size of the source and then we have some collimation guides and apertures that we use to define the flux and the coherence that you're going to have. You have a sample aperture, you have your sample and there's scattering. So the scattering will happen at some angle. So the scattering angle 2-theta that I'm showing you there. And in very, very basic terms the scattering experiment is basically an experiment where you're measuring how the intensity of the scattering varies with this angle 2-theta here on the detector. Now it's called small angle scattering because we are measuring very small angles. So if you look at the correspondence between the scattering angle and the wavelength that you use and this D which is the real space distance or feature in your sample. If you think about the typical ranges that we're using, that we're measuring, the angles that we're looking at are typically from 0.3 to about 5 degrees. So they are very small angles what we're trying to do. So if you've done crystallography, these angles correspond to that range of data next to the beam stop that you very often actually reject from your data. So it is very small angle. And we Frank already showed you a version of this, this schematic so this is a very basic idea of the scattering process so you have an incident plane wave of neutrons. And the neutrons will see a nucleus or remember they interact with nucleus not with the like condensed cloud like an x-rays. And there will be an interaction between the neutron and the nucleus and the nucleus will then scatter as a spherical wave. So you can try the spherical wave with a wave function that will have an amplitude that is proportional to the scattering length and to the distance to the detector. Now the scattering length quantifies the strength of the interaction between the incoming neutron and the nucleus. So when the neutron comes in, there will be an interaction and there will be also a momentum transfer. So there will be a change in the wave number that comes in. Let's look at this with a typical triangle scattering. So in real space you have the incoming neutron with some velocities and some speed coming in and then coming out at some angle. So in this case your sample is like your single nucleus. In reciprocal space you have a wave vector coming in, and then there's a change in direction on the wave vector going out at a certain angle. And the difference between these two vectors is what we call the scattering vector or the momentum transfer vector, if you will. And it measures basically the difference between these two vectors. And if there's also a transfer in energy, we can measure that as well. For a smaller neutron scattering, what we are measuring is elastic scattering. So basically there's no change in the amplitude of the wave vector that comes in when it goes out. There's only a change in direction. And it is done very simple. It's basic trigonometry. If you look at this triangle, it's very simple to show that your Q, your scattering vector, is 2k sin theta. So this is half the scattering angle here. And if you know Bragg's law, it's straightforward to get to this relation that you will see many times in your publications between the scattering angle and the wavelength and the, sorry, the scattering vector, the wavelength and the half the scattering angle. In real space, the distances that you're measuring, and I believe Frank also mentioned this to you, are, have a reciprocal relationship to Q. So when you're measuring small Q, you are actually withdrawing information about, about largest spacings or largest distances in your sample or correlations. And when you're measuring large Q, you're, you're at a high resolution if you will, if you come from the crystallography background. So you're measuring smaller distances. Okay. So let's look at the geometry of the sans experiment. So you have your incident neutron hitting a sample. And they'll be scattering that can be defined in this solid angle the omega that you see here. And if your detector is far away enough. You can actually define this, this d omega angle and this to the angle, very precisely. And so we can also define a differential scattering cross section, which is basically the, the, the number of neutrons scattered into this solid angle d omega per second, normalized by the flux. And it can be shown as it's straightforward to show if you, if you think about the definition of being flux and it's solid angle that this differential scattering cross sections actually be square so remember the in the wave function the amplitude was B square. So if you square that, you get the amplitude. Now the total scattering cross section, which is actually what we measure when we do an experiment was also called a microscopic cross section is the integral of the differential scattering cross section. And so basically you're now integrating over four pie so so instead of just the one angle we're integrating into all directions. And you get to this to this equation that basically tells its four pi over the square of the scattering. Now the units of the, the, the microscopic cross section with which you measure our units of area. So you will also see them referred to as, as, as a barn, the unit is born which is 10 to the minus 24. So you'll see that a lot when you look at the tables of scattering cross section. Now the result that we want in the end of course is not the one for the one nucleus you want the result for your full sample so we normalize the microscopic cross section here to the entire volume of the sample. So that we then can compare results measured to different experiments or with different samples. And once your data is reduced you get to absolute units of centimeter minus one so in the tutorial, you'll use data that is already reduced so the intensities, the units of the intensities there are centimeters minus one. And this is what is the, the information that we can actually relate to our to the information on our sample that we want to from our experiment. And if we do, if we do the differential of the microscopic cross section, what do we get what information is in there. Again, we're integrating for our entire sample. And, and it's for small angles capturing we use relatively long wavelengths measuring very small angles. So the information that we're getting is not about inter atomic distances we're not looking at that level of resolution we're looking at larger distances and inter atomic. And in that case what the neutron see is it is a density of scattering links. So it's a density of a material not inter atomic distances. So when we integrate, and we do the Fourier transform what we get this actually scattering density distribution so you see the amplitude there. And the meaning of this, this amplitude square square is that we actually lose the phase information so we cannot just do an inverse Fourier and obtain the the microscopic cross section. Again, which is an important result. And also if you think about the density in here, what you have then when you're measuring small angle scattering is an average density of your sample, and there will be oscillations around that average so when when the scattering length density varies. So for example as you, as you're looking at your sample or your molecule let's say it's a protein molecule, as opposed to your solvent. There will be some, some, some difference there like you would have for example with light scattering and reflected index. And this is what we're actually measuring. It's where the signal actually comes from it's my little scattering it comes from in homogeneities in the scattering length density. So, just to step back for a second let's look at the dimensions of the things that we are measuring so if you look at the sizes of different things so for example the wavelengths of the neutron is comparable to the inter atomic distances that were that your sample has. The atomic nuclei on the other hand are smaller by a factor of four. So for neutrons you're the nuclei are actually a point scatter because they're so much smaller than the wavelength of the of the neutron itself. The scattering cross sections are on the order of similar order of magnitude. That's the velocity of the neutrons there just out of interest and the typical neutron flux that you'll have at the sample it will vary in the, but depending on the instrument that you're using but typically it's around 10th to the seven. Now this has the few implications and assumptions that we do when we look at small angle and scattering data. Before we move on I just wanted to list them there for you so. Just to remind you, some neutrons will pass through the sample and deviated right so the neutrons remember the interact with the nuclei not with the electron density cloud. So they can go through the sample and deviated. And we assume when we're interpreting the data, in most cases anyway, that the, the, the neutrons will undergo a single scattering event so that there's no multiple scattering from one nucleus to the other. We also assume that the incident beam is not significant is to significantly disturbed by the medium. We assume that the nuclei are fixed at their position in the solution that's why if you look at the tables for cross sections you'll see that they're, they're listed as bound cross sections so basically we're assuming that the nuclei cannot recoil, or that they're not moving so that they cannot impart energy on the, on the, on the neutron incoming, which is an assumption of course. So where does the signal come from and what are the different components so the total scattering that you're measuring comes from three major components so there's a coherent component and incoherent and absorption components if you look at the little schematic to the bottom here. If something's going to happen to the neutron it can get transmitted so it'll go through your sample. It can get scattered and that's what we want to measure, or it can get absorbed and your sample will have some, some thickness that you will define for your experiment. And this is one of the parameters that you actually decide when you're going to do a neutron experiment. So we can, we can quantify the, the absorption by measuring the, the intensity sort of flux if you want. They go through a sample compared to the incoming flux of nutrients. Now this absorption actually reduces the signal to noise of what you want to measure and it's, it's wavelength dependent. Now the other two components are our main concern typically so the signal that you want is your coherent signal. And what the signal is is, is a, it comes from correlations between the, between positions in your different sample, in your sample. So basically it's the, it's the structure of your sample and it's, and that is Q dependent and that's why we measure the intensity is a different Q's. Now your incoherent signal comes from which for all intents and purposes is the noise or the background for yours for your sands experiment. And that's from the, the, the fact that there are different nuclei in your sample that have different spin states. So, so for example, even for the same elements there are different spin states for certain nucleus if they're, if they're non zero spin. And so the interactions can, can vary in ways that are not related to your structure. So you'll have a type of scattering that is not Q dependent. You'll have a flat background for your, for your data. And this you can think of as your, as your noise really. So, how do you know what your sample will do and how much of these three components you have so you can look at can look up tables for the absorption contribution in particular if you know the isotopic composition of your sample. You can calculate it fairly accurately so you can look up the different contributions from the different elements and you can, you can actually have an estimate of what the absorption composition or is expected the expected absorption. Contribution is the incoherent contribution is slightly less straightforward. Because it depends as I told you it depends on them. It's not, it doesn't have information about the positions or correlations in your sample. And it is dependent on on the movement so at the positions of atoms at different points in time. So it's related to dynamics, and it's temperature dependent so it's more difficult to estimate it. And the coherent contribution as well it's because it's dependent on the structure and a priori you don't know it. You can get an estimate but again you don't know it to very very accurately so you, you have an idea by looking up tables as to what to expect. But it's not, not necessarily straightforward. Let's go back for a minute. So here we're comparing so on the, on the vertical you have the penetration depth of the, of the probe that we're using and we're comparing neutrons x-rays and electrons. And at the bottom you just have the atomic number. And you can see that the higher the atomic number for x-rays which are the squares here, the less penetration that you get because interaction gets stronger and stronger so they penetrate less and less similar for electrons. And then for neutrons you have these triangles here and they're all over the place, but you can see that the penetration depth is typically on the order of centimeters. And it is not dependent on atomic number of course because we're interacting with the nuclear and the nuclear have different different characteristics depending on not only on the element but on the isotope as well. This is why neutrons can do cool things like you see on image here so they can look into a high atomic number of materials like the metal that you see here. And they are sensitive to the presence of lighter elements so we can get a cool image like you see there. So, going back to the cross sections. So here you have your neutron scattering lengths on the left. And at the bottom you just have the typical atoms that you'll see or typical nuclei that you'll see in biological samples so these are very often present for proteins, nucleic acids, lipids, etc. Also in your buffer. And you can see if you interesting things. So you're in green you have your coherent signal and in red you have your incoherent and I told you that you can think of your incoherent signal as your noise. So if you remember that that hydrogen is arguably the most abundant. And most important element in biology you should start worrying because of course you can see that there's a big contribution from the coherent scattering so you'll have a lot of noise when you have a lot of incoherent scattering in your sample. And you'll see that the contributions for hydrogen and deuterium and just isotopes are very different. So for for deuterium not only the coherence scattering is positive as it is also approximately double that of hydrogen, but also it has this very interesting feature, if you will that the incoherent scattering contribution is much much smaller than that of hydrogen. And the main reason why we when you do a science experiment, you will almost certainly be using D2O buffers instead of H2O so you're trying to reduce this this contribution. It is also the reason why we can obtain different signals and almost tune the signal that we want by playing with a percentage of hydrogen and deuterium in our sample, as I think Frank has already showed you. Okay, so going back to the general idea of small angle neutron scattering so if you so on the right here you have real space and just a little cartoon description and on the left you have the reciprocal space. And if you focus on the right here for a second and if you think that your sample is a molecule here it's represented as a very anguished fish. But if you have your molecule in a buffer, like you have here, what you're going to be wanting to do is reconstitute what you from the data you're going to be wanting to reconstitute what you what you have in yourself so basically the structure. So you're going to be wanting to know what's the contribution to my scattering that comes from the molecule itself which is the fish here that has a certain volume. So the, the, you're going to add that the, add the contribution from the solvent itself, and then you're going to want to subtract the contribution from the solvent that is not present because yours, your molecule is there right. So you have your, what we call the excluded particle volume so in terms of reciprocal space on the left. What does that correspond to so the intensities that you measure their proportional to the signal that comes from your molecules and you have your, your scattering intensities there. And you're going to integrate that for the volume of the sample, you're going to add the contribution from the solvent now the the solvent is for all intents and purposes of infinite extent compared to your, to your molecule and, and to the process so it's a delta function except for q cos zero. So we don't observe it in practice so this is basically zero. And then you want to subtract the contribution from the excluded excluded part of volume so it should be something similar to what you have here for your molecule except the scattering intensity now is for your solvent so you want to subtract that. And this is how we come to the concept of contrast that you have to to subtract the contribution of your solvents to to from the contribution of your, your mouth or your fish if you want there. And before we go on to contrast I'll just talk a little bit about buffer subtraction because it's actually important. So the image that you have there is just an example of data on that was collected on license I'm solution at different pressures. So you see from ambient to about five kilowatts so five kilowatts about 5000 atmospheres just for reference. So, and you have your, your cross sections here on the, on the vertical axis. So if you look at this inset here. You can see that the license I'm the cross section of the license I'm solution which includes the buffer and the protein has a profile as you see up there. And this profile is actually the sum of the contributions from the protein itself license and the buffer. And it changes with different pressures and this is a story for another day, but basically you get the point that you have to add these two contributions to get your license solution. And what you don't want is that the signal from the buffer is basically the major contributor to the signal in your, in your scattering. So here to your experimental data you see different curves collecting at different pressures and it's color coded. So you can see that for example this top curve here, which is the highest pressure. The data from your molecules on top and this flat line that you see is your buffer that was measured as well so we measure both you measure you measure the protein solution and you measure just the buffer. If you've prepared your samples right your buffers should match exactly the background of your sample at this high Q here. And you can then subtract the buffer contribution from your protein and you have your reduced data. Now, if you look at other curve at the other curves and the reasons why this happens under pressure are not are not important here but the important thing to see is that sometimes it doesn't match. If you look at this green curve here, you can see that the solvent is a bit lower than your protein. Same for the red curve and same for the slight blue and the black curves. Now when this happens in this case it's an effective pressure, but if you're, if you're just measuring at a standard condition, the standard temperature and you have no other reason for this to happen. You'd have to either apply a flat background subtraction or you'd have to scale the contribution from your buffer. And you can imagine that that you, you can introduce errors from this so what you really want is that your buffer matches the background of your sample here at high Q, where there are no more structural features main sense. And if that doesn't happen, then there are a few reasons that typically the reason for that is that you're you have what we call the buffer mismatch so when you prepared your sample. There's slightly different percentages of deuterium for example in your buffer or salt so you have to be very careful to make sure that your buffer matches. You could also have air bubbles in your buffer, or on your protein sample or you can have them on one and not the other. So air bubbles are a problem because not only they decrease the volume that sample that you're exposing but also if they are of the right size which is the size that we're measuring here in this Q range so if they have the size that the experiment has the resolution to see, if you will, they will contribute to the to the scattering. Then other reasons are, for example, in coherence or multiple scattering and also incorrect data scaling so if you if you have if you haven't scaled the data properly you can also have issues like that. Okay, so let's go back to contrast. I believe Frank has already shown showed you an example so I won't go too much into it, but still I want to mention it because it's important it's one of the main reasons why you use sense. So your contrast and sense as I told you when I showed you the fish. The difference between the scattering intensity were simple and that of the, the medium that it's in so you're solving the fuel. And when you hear, you hear is talking about contrast, it's a scattering like fancy contrast and it's the square of this difference. So now we come to this magic formula that you will see many many times. Hopefully you will remember if nothing else please remember this. This formula, your intensities will be proportional of course to the particle volume fraction. And the particle volume itself, they are proportional to the contrast on your sample to a form factor that describes the shape of your molecule so the shape of your fish. So if your fish start talking to each other and they start influencing each other and interacting then there's there will also be what we call a structure factor that we also need to model to fit our data. So, let's look at the contrast just a little bit more in detail for a second. So, the contrast as I told you it's the difference between the scattering intensity of your molecule and the medium that it's in. So your black line that you see here, sorry the vertical axis is the scattering intensity and then on the bottom you have the percentage of T2O. And this black line is basically the scattering density of water as it goes from H2O to pure T2O so it will increase. For example, if you look at this dark blue line here you have the signal for a protein. Again, as you increase the percentage of T2O in your buffer, the label hydrogens in your molecule start exchanging so you expect a slight increase in the signal as well. And at some point, something interesting happens is that these scattering density is crossed. So you can see here, roughly around 40% in this case the scattering density crosses. What this means is that you have the same signal from your protein and from your solvent so the contrast is then zero. So basically you don't see the protein in your sample. So you can imagine that if you had a system, for example, that had DNA present there. At this point when you were using 40% T2O you would not see the contribution from the protein but you would still see the contribution from your DNA and imagine you have a protein DNA complex for example. It's still there but you are now focusing on the contribution from the DNA alone. Why if you use for example a contrast or percentage of T2O that was equivalent to the point where the DNA now matches the solvent. You would still see a signal so it's a difference remember from your protein contribution from your protein, but not from the DNA so it's basically tuning different components of interest in your system. You can see in the image here on the left so here you have a representation of sort of a core shell structure or even a micelle if you will. And you can see that the structure and solution is still there and it's always the same we have not separated the different components or we are still looking at them as they are when they are in a full intact complex. But when you're in 100% T2O, this is just a schematics to give you a feeling of what contrast is. So basically looking at the core structure here so you'd be investigating the structure of the core. As you start increasing the percentage of hydrogen in your sample. You start seeing contributions from both of them because the scattering length entities of both components differ from that of the solvent. And then as you approach the 0% T2O, now you're matching the core of the molecule so you'd basically be seeing the contributions from the shell. So you can see how this is useful for us to work out the how different complexes and range and what's the structure when the whole intact particle is there. Okay, so let's talk a little bit about structure factors. So those little fish that I told you about if you use high enough concentration, or depending on the structure that they have, they may want to talk to each other so they will have some interactions. And when that happened remember that magical formula that I showed you you have a contribution from this interaction so this is the structure factor SFQ here. So I started introducing this factor that I didn't show you before this is basically the background so your signal will have the contribution from your from your molecule. And there will also be a background that we typically subtract but just to get you used to this factor that is always present. This is the income here in background. So your, your structure factor actually, if you have an isotropic system and interactions are all uniform, regardless of the shape of your of your molecule. The structure factor can be calculated and it's a function of the, this G of R which is a radial distribution function. And it's governed by the interaction potential so by the type of potential or the type of interaction fuel between the different molecules of the how the different fish interactive fuel. So this this potential, if we knew exactly the type of interaction we could calculate it very precisely and we can then model our structure factors but in biology, as I'm sure you already know the samples don't behave like we want to they will not do exactly what we want. One of the things that they, they don't do is they don't they're not necessarily uniform. So you'll have a slightly different shapes of molecules in there. Plus the interactions between them are not necessarily isotropic so they have a certain shape, and depending on the orientation of the different molecules or of the different fish the interaction between them will be different. And so, what we effectively measure is this S prime structure factor and apparent structure factor or effective structure factor if you will, that is affected by the fact that the interactions between the molecules are not isotropic. They will vary depending on the orientation. And so to account for that what we typically do is we use this, it's called a decoupling approximation so we introduced a little factor the speed of factor here. And it takes into account the effects of the orientation of your, of your, of your molecule so it basically introduces a little fudge factor into your parent's structure factor and it helps us module. And you will see this in the, in the tutorial that we're going to do after. So if your system is monitor disperse now this this beta factor will be one, and you're back to the structure factor of your molecule alone but very very often, more often than not really what you're measuring is this effective structure factor. Okay. So, how do we do this how do we model this structure factor so at the, at the forward scattering intensity which is intensity for the q angle zero. And your structure factor is actually proportional to the osmotic compressibility that you see there in red this is just the Boltzmann constant and temperature, and the osmotic compressibility is a very interesting parameter for many, for many reasons. It depends on the sector material coefficients, which is also tells you gives you information about interaction between the particles but also the molecular weight and the concentration of your sample. So if you want to know the concentration of your sample with accuracy as we'll go into that in a second. So for example if you're if you're interested in the different phases of your system if you're interested in the effects of the, of the, for example of pressure. And you can read it a little bit more about how to use pressure. I put the reference there with Sam's. This forward scattering intensity has a lot of information about the, about the types of interactions that you have in your, in your sample. And these interactions can have can be a different nature. So there I thought different types of structure factors and you have a diameter here on the on the x axis. And you can see that you can have, for example, cool on repulsion so if you if you have charged particles your molecules mostly charged, then you'll have a certain type of interaction if your sample is more like a hard sphere, which is not charged and you'll have a different type. And again if you have an attractive square well interaction, another, and you can see that at these shorted, sorry at these larger distance so basically at low q if you want. So these contributions are stronger. Okay, so it'll affect your data more at lower q. I'll show you plots in a second. But what we're doing by analyzing our structure factors is looking at the types of interactions that we have present. So not only can we see the structure of the molecules but we can also see the types of interactions. And this is a tremendous advantage of sounds as well. So let's go back to our magic formula. If you have an attraction or a repulsion type of interaction. Yes. Excuse me, I just quick question what is the difference between them structure factor like s and s prime. Okay, so let's go back to the question. You have your just your molecule and their interactions between your molecule. You have just your structure factor like you see on the top here. I don't know if you see my mouse. But the end and this structure factor is the structure factor that you would have if your molecule let's say was just a sphere. And these spheres were all the same across your samples all the same size, and regardless of the direction in which you measured so to be regardless of your q angle if you will. The interactions would always be the same but in reality, your protein can be let's say it's an intrinsically distorted protein and solution. And the direction, you know, on the relative orientation of these two molecules the types of interactions will be very different. So the, so our samples are not as simple as would like them to be basically. So it's so the direction in which they interact does matter. And in that case, what we're measuring is actually an apparent structure factor, or an effective structure factor if you will. We introduced this be factor that actually takes into account the fact that the orientation of the molecules matters we can't just apply an average form factor and then work out our interactions. And you will do that in the tutorials you will you will see you will introduce this factor and you will see the effect on the 15. So this is this s prime that I did I now refer to. It's basically the same thing but but more realistic if you will. Yes, thank you very much. Right so going back to our intensities. Looking at our structure factors again and here I just put a little reminder of the different things that you're looking at as you as you plot and send intensities versus cute so remember this reciprocal relationship between q and the real space dimension so low q you're measuring sizes and the intermediate and when you get to higher q you're now looking into more details of the internal structure. Okay, so if you have different types of interactions. Let's say you have your form factor here your FQ for your molecule. What you measure in the end is is really the, the, the combination of these two, the form factor and the structure factor. And if you have some attractive interaction. You will have a curve similar to this, and the sum of or the combination of the two notes and it will then show a little upturn here so if you see an upturn it's, it's typically because you have some sort of attraction potential on your, on your sample. And then you have to use what you know about your sample as well to try to find out what types of potential or interaction to expect. If it's a repulsive interaction, then you will have this downturn here. And when you combine the form and the structure factor you will see this downturn. And again, we'll look at this again but just to give you a more intuitive feeling of what to expect. Okay, let's go back to the form factors. So we've looked at this parents or effective structure factor, but about this form factor which is very, very often what you're really interested in from your sample. And this form factor is basically the shape of your single molecule right. So in space you'll have a molecule for example like a sphere with a certain radius. And in the real space as well you will have some some density that is zero when you're you move past the dimensions of this radius. And it will have a certain average density when you're when you're under that dimension in four years space. So if we look at the normalized form factor and this is just a form factor for the, for a sphere, you'll have some some signal that could look like this for example if you had this perfect sphere sphere, like I told you always the same say a shape always the same size of the uniform. Now in reality of course and this and I cannot overemphasize this. In reality, our samples will be polydispersed very very very likely to some to some extent at least. So polydispersities up to 20% are very very common so if you're doing by a problem if you're looking at biopolymer solutions. This is to be expected. Also to be minimized but but be aware that this will happen. So what happens to your data so your data is then smeared out a little so the more polydispersed your sample is the more smeared out your intensities will be and this smearing can not only come from the sample but it can also come from the instrument itself. So the instruments have a finite coherence in queue. So that to will cause some, some smearing of your data and when you're fitting your data. So what you have there is, you see the little points that those are your data, of course. And then in red, you have the model for a spherical shell, where smearing was not taken into account and you can see that the model doesn't fit the data, particularly in these troughs that you see here. But once you introduce the smearing that comes from the instrument and the fit is much much better to the data and that's the blue line that you see there. And you'll do that as well in your tutorial I will show you that when we do when we fit our data. We, we take into account the instrumental smearing which comes with your, with your small angles capturing data. So when you when you collect your data. So the information on the smearing is on the same file. And when you're fitting your data, you, you use that information to know what to expect. So basically you add the smearing to your, to your model. And if you don't remember anything else about data fitting. Remember that you always add smearing to your model, and you don't be smear your data. Okay. And we'll talk about this a little bit more in your tutorial but this is an important point. Okay. Let's see how we analyze our sense data. So let's assume we have our sense data and we have as good a data as we could get. If we're in the dilute low Q regime so basically we're in the regime where we're looking at sizes that are larger than the size of your molecule so the size of your molecule again is this deceiving perfect fear that you have there. Remember I told you we have this effective structure factor that we can approximate we can introduce a fudge factor if you will, that takes into account the fact that there are different shapes present in there, and that the interactions depend on the orientation for dilute solutions typically this beta factor that you have there is zero and your, your apparent structure factors just goes back to being your regular structure factors so they're both the same and approximately one. In that case your forward scattering intensity, and will just be proportional to to your scattering intensity, and to the, the, the number density of your particles so basically, you, you get this molecular weight information back here. So if you can extrapolate your data so remember at q zero, we cannot measure the data q zero because that's the direction of the direct beam and the intensity is there and the direct beam completely dominate the signal that you can measure so we don't actually measure that, but we can extrapolate so we can measure to as low as low as we can and then extrapolate what the intensity would be. And then you can, if everything is well calibrated you can actually obtain information about the molecular weight of your sample. And of course there are other measures other ways to measure molecular weight but in solution. That can be important for example to determine if you have a dimer or or a trimer and your solution. Now, as Q, if q is not zero but it's approaching so at very small q, or as q tends to zero if you will. It can be shown so if you do an expansion of your, of your exponential so if you have your wave function, you do a table series expansion basically, you'll get exponentials that are a function of q and if, and as q is smaller and smaller it reduces to the contribution from this first term and your intensity will be proportional to your force scattering intensity. And to this exponential that you see there. And this is valid at least for it depends on the shape of the validity of this of this equation depends on the shape of your molecule but for globular proteins. It's typically valid to in the range where q times the radius of duration is one point is smaller than 1.3. If you look at this equation it should be straightforward to see that it's easy to linearize it so if you apply logs on both sides, you get this. This very interesting relationship between intensities and your q square, which you can plot directly from your, from your data. And so you expect a straight line. And the interesting thing about this straight line or this Guinea, as we call it, is that the slope will give you the radius of duration as your particle. So you're measuring information already. And the intercept will give you information about this force scattering intensity, which in turn as you see on the formula up there also tells you about them like the weight information so there's a lot of interesting information that you can withdraw. Now, I put some limits there for the q min and q max the typical ones at least that are very interesting. And that you should try to to look at so when you you're measuring your data you should try to cover a q min that goes at least down to pi over d max or the maximum size of the maximum dimension that you expect for your sample. So your q max should also obey this, this rule of thumb, if you will so that you have enough points to get your, your opinion. Sorry, just take a sip of my coffee. So what about this intermediate q range so this was a low q. Now we're looking at sizes that are smaller than the size of your of your molecule so here's your deceivingly perfect small molecule again your sphere. So we're looking at a resolution that corresponds to smaller sizes. He can also be shown that there will be a cute abundance of your intensity is that again, keep can be easily linearized and that brings very interesting information about the, about this and factor here, which gives us information about the shape of our sample. So depending on the type of objects that you have in your sample and there's an illustration there. So if for example your objects more like a one D type of object you'll see a Q dependence of about minus one, if it's 2d minus two 3d minus four. And then there are some intermediate values where you, you can characterize what we call the fractal dimension. So you can can investigate the types of structures that you have in there and they could be a mass fractal surface fractal. Directly from your data and without knowing the exact structure that you have in there. You can get information about the types of structures present, and this is at intermediate Q. If you're a Q that is much larger than the largest dimension of your sample something very interesting happens as well. So remember when we're in what we call this broad region. The intensities will be proportional to will have a q to the minus four, the dependence. And for example if you have a sphere like we keep showing you there. We get to call the pro law where the form factor is proportional to the surface to volume ratio. So again we can obtain information about the, about an important parameter for which is very important for example for industry, even for the food industry. Let's say you have a porous material in your sample. The surface to volume ratio may be very important information that you can measure with sounds. And just to show you in the, in the, in the profile where this is. So the Guinea region remember this is at lower key. And then as you go to intermediate keys and higher, you enter the pro region and you have this q to the minus four dependency and again I'm showing I'm, I'm deceiving you because I'm showing you very perfect. Very perfect line so this is all theoretical for this fear of say five nanometers, but you would expect a profile like this. Now, depending on the shape so of course your samples not necessarily sphere. You can have different shapes. And like I told you you'll have different q dependency so if you had more of a cylinder you'd see this q to the minus one dependency before you enter the q to the minus four. And if you had something like a disk, which has a surface, then you'd see for example this q to the minus two. So the profile of your just looking at the profile of your intensity versus cube. You can, you can already infer some characteristics of your sample. Okay, so what can we do with a whole key range. Now if we, if we assume that the Guinea law applies. So let's think again about the Guinea equation. If you multiply both sides by q times RG square, and you divide by I not so the reason why we're dividing by an artist just to normalize so that you can, for example, compare your protein solutions different concentrations. You get this interesting result. And if you plot it, if you plot this equation so you can plot this straight from your data. You'll get a profile that will immediately tell you something interesting which is the information about the folding of your protein. So for a globular protein can be shown that we expect a peak so this blue line here, you would we expect a peak at about the square root of three. And this you cannot be can easily calculate from the derivative of this equation. So if you have an intensity of about 1.1. So if, if the Guinea equations is not valid or if your protein is not quite globular they'll be deviations from this behavior. So if your protein is asymmetric, for example, they'll be a deviation to the right. If it's partially unfolded as well, or if even an intrinsically distorted protein. So you can imagine for example heating up your sample and watching it unfold. You're applying pressure and watching it unfold or change the pH and watching it unfold. And all this you can do straightly directly from your data, even if you don't have a structure for it. Okay. What else can you do with in with the, with the entire Q range if you will. So, going back to the formula for your, for your intensities. So your, your intensities depend on the form factor and you can integrate from zero to the maximum distance that you expect on your sample. And you can do an indirect Fourier transform. So you can invert this and obtain information about what we call the distance distribution function and what is this. Now, mind you, I say indirect Fourier transform for many reasons. One of them is the fact that we don't measure the entire Q range from zero to infinite. So there are some approximations that we make when we do this calculation and also our priority. You have an idea of the d max to expect to the maximum distance on your sample it's to first approximation you can say it's twice your G. But we don't really know this because we don't yet know the structure of our sample that's why we're studying it. So there are assumptions that you make when you do this calculation it's not a direct calculation from your data. It's an interesting calculation to do. And I'll show you examples. Because it gives us basically a histogram of inter atomic distances in your in your sample so a histogram of the all potential differences that you'll have in your sample so you'll get a peak for example, that will correspond to your RG. And where it crosses zero. So this is an estimate of your d max so the maximum dimension of your sample. And this is interesting information because you can relate it to to structure. Let me show you examples. So this is your data here on the left and again I'm deceiving you because I'm just showing you data for perfect shapes and remember, it'll all be smeared out in a real sample. You can see that in the tutorial, but if you calculate the P of R. You can see, let's pick an example so let's say for example, this dumbbell here in pink, you can see that the P of R reflects the two distances the two major distances present in your sample so again directly by calculating the the P of indirectly calculating the P of R. You can get information about about structure. Okay, so how do you use all this to assess your data quality so how do you have a quality semi qualitative assessment of your data quality so on your left you have good data. In the middle, you have some inter particle interference so it's not necessarily bad data which just means that there's some interactions between your particles, and it could be subtle it could be severe and then on the right you have aggregation. And in red you have very serious aggregation and in purple. Not to that. And again, just to, to so that you're not too discouraged when you measure science data very often we're between one of these two conditions. But let's assume you have the three cases. So how do you how do you know the quality of your of your data. Well, you can, for example, calculate again is so let's say you calculate a guinea for all these three cases. If you calculate your guinea and you have a dilute sample with no other interactions present your guinea will be straight line with no major deviations. It's obvious and you can calculate a very accurate RG from your data. If you have some inter particle interference you can see for example there's a little down turn down turn it to low q and I told you that corresponds to some repulsive repulsive interactions. You can get an underestimated RG if you use the data and this range so in this case you would cut the guinea range to lower, lower q so that you can get a better estimate so basically you wouldn't include the points that you get. And if you have something more serious like aggregation, you can see how it may not be very obvious. So there's still a fairly good fit but there was an upturn at low q. It's a very overestimated RG and it may not be obvious because remember at the beginning you don't know what the effect RG is. So what can you also use to know if you're in this third case and that if you're overestimating your RG. Well, you can also calculate the PFR so the distance distribution. If you have good data it should be this, this more or less model shape and you can see this asymptotic decay to zero as you would expect. Now the estimate of d max typically even for good data has errors on the order of 10 to 20%. So you have some interparticle interference. Again it may not be too obviously obvious because you still have a shape that looks kind of okay, but here you're artificially calculating a small d max and you know that because you're guinea indicated that there was an issue there. Now on the third case your guinea didn't indicate anything too drastically wrong, but your d max will show it because instead of having this asymptotic decay to zero, you have this sudden downturn so that your d max here is too short. So then you'd recalculate your PFR introducing a larger d max until you get a curve that decays more gently to zero. So there are ways in which you can, you can assess your data and of course you always want to collect complementary data as well to make sure that everything is consistent and that you know what you have in your sample. Okay, so let's talk in general about sans experiments. The sample concentration so all of these analyses that I showed you can be very powerful but remember that sans is a very underestimated. The sense data that you collect is is typically not two or three or four times more data than you have parameters to determine. It's a very underestimated problem if you will. So having parameters that you know well and that you know accurately is very important and one of them is your sample concentration so you don't want to have errors in your sample concentration. So you can be the most perfect pipeter in the world and the most you can have the best lab techniques and still have issues with your sample concentration so you want to verify it. Actually verify the sound the concentration of the sample that you're using. So, there are many sources of error when you're measuring concentrations so you want to know that parameter accurately. The concentrations in general are better for for for the signal to noise. But we, if you for example just want to look at your form factor and you don't want to have structure factors present. You should always do the concentration screening so measurement of different concentrations to see when your structure factor starts impacting impacting your data. If you've never done a sense experiment that and you're wondering how much something do you need so typical amounts of sample or about half to one mil. And the concentration can can be anything from one to 10 mix per meal that depends on the on the scattering the cross section of your sample. But remember that if you're going to use low contrast conditions. So, when when you're going to come closer to the in the buffer signal to the signal of your sample. The intensities will be lower so then you might want to use higher concentration so think about these things when you're planning an experiment or talk to an instrument scientist and they can, they can also give you a few recommendations. I told you about public dispersity. It is a problem because the no sense is a technique that averages the data of everything present. So if you have very public disperse samples you're basically losing resolution. So use complimentary techniques to ensure that you either have a minimal aggregation or that you know what the public dispersity of your sample is check for purity. Also remember the pH and PD effects so if you're working with the with D2O. There could be shifts in your in your pH that you do not expect so watch out for those and also the solubility of most biological molecules is reduced in D2O. So the fact that you're working in D2O buffers can also introduce some polydispersity. This is a sample environment so what temperature pressure pH and precision we need because this can determine not only if sense is an appropriate technique to use but also which instrument or which facility you're going to use. Think about how long is this the samples table for so if your sample is only one a disperse for half an hour after you've done a size exclusion chromatography. This probably means that you should be using sacks instead so you probably should be using a technique that can measure it faster. And for example if you're using pressure cells if you're looking at investigating phase diagrams for example, you may need the larger volumes. So the I think the overall messages that you have to think about all these things but also my strong recommendation is discuss these things with instrument scientists, they're happy to discuss it with you. And they can alert you for different aspects that you may not have thought about ahead. Now in the tutorial we're not going to cover, we're not going to do a whole contrast variation experiment unfortunately we don't have time for that. But I wanted to just leave you some software suggestions. You can estimate the contrasts of each component in your sample so that you know what to expect from your experiment that will also make your proposal stronger. If you include this information so if you do some calculations and you know what to expect then you are less likely to have surprises when you collect your data and you're more likely to have a successful experiment. Again, these are suggestions so there are other types of other software that you can use, but first if you have the two component complex you can use mulch, for example, and it will calculate the signal that you should expect from the two components. If you have more than two components, you can for example use the contrast calculator module of sassy, which is another software I leave you the references there so you can always look them up. So you can investigate it an unlimited number of components and it also gives you the forward scattering intensity as a function of the percentage of detail which may be also important information to help you know what concentrations you want to use. And it's also one of the parameters in your guinea so it'll help you know what to expect. Now if you want if you have a PDB file available, and you want to compare with the crystal structure or have an idea of what to expect. Again, you can calculate the theoretical sounds curve. Again, emphasis on the theoretical. But it'll help you for example know what Q range you're going to need to measure. And both the model the saskal module of sassy and chrysanthemum will do that for you. This is not necessarily straightforward so you have to add hydrogen to your sample or deuterium. But again just give us a shout if you're interested in doing something like this and we can help you out. So just to revise what we've covered. We've looked at the sound experiment and hopefully you have a better idea of when you would use it. There are two main reasons why you would use, for example, sounds instead of sex. Usually it's because you want to do an in situ measurement, or you want to use contrast or both. So in situ measurements are very powerful with sense because it's an undistracted probe right so that you have no radiation. So if you want to, for example, do a temperature screening, see how your sample reacts to different temperatures look at hysteresis. Reversibility of the effects then you're going to want to use sense, or if you need contrast if you're sacks for example data does not give you enough contrast. You can use labeling to to to increase the contrast or to highlight a certain component let's say you have two proteins you want to see how they, how they complex together what are the structures. Then you can deuterate one of them and then the highlights the two structures in turn. So in this sense, scattering process a little bit and I, and I told you about this magical formula and I, I've tried to introduce to you the notion that what we measure typically is an effective or an apparent structure factor. But there are approximations we can use so there's this beta approximation that we can use to take into account the fact that your, your sample is not necessarily a sphere with the isotropic interactions in all directions regardless of its orientation. And I also told you about the interaction potentials that we plug into the calculation of the structure factor. So remember a sense can only measure structure but it can also measure the interactions between different types of molecules. And it gives you information about the types of interactions that stick. And in that during the tutorial we're going to work on this a little bit you're going to to fit your data with some structure factors and you will see the effects of this. We've looked a bit at about the different types of analysis that you can do so low Q you can have your kidney and have your PRR and your cracky and the powerful thing about these three is that even though for some of the kidneys look completely a model independent approach, but you don't need to know about your sample a priori to do these analysis so you can you can get a lot of information from your sense data even if you if you don't know too much about your the structure of your mouth. And finally looked at the data quality a little bit we spoke very broadly about how you plan your experiment. I don't know too much about labeling but this is something you will want to consider to increase the contrast of your, of your sample of your data, or also to just look at different components in a complex we spoke about a little bit, and just a final note so as Tommy mentioned I work at this so I would at this we have a suit of sands and pieces and you sense of different Q ranges of instruments. So if you if you want to do an experiment that missed just drop me a message or if you have any other questions. I again, a sort of a take home messages that you want to talk to instrument scientists. If you're thinking about doing a sense experiment or even if you're wondering if sounds is the technique that you want to do just to just reach out to instrument scientists because they're, they're, they're glad to talk to you and they, they can at least give you some pointers on on what to expect and what you would need. And with that, I think I am done I don't know if you have any questions. Any questions for Susanna. See no questions. So I'll give you the copy of these slides so that you can sort of digest the information. And your own time and if afterwards you also have questions as you look at the slides with more time. You can also drop me a note or discuss during the tutorials. So how do you want to go about with the exercises now. So, so we have some hopefully you saw that there's some data that you can use. There's you'll see in the in the link that as a as I showed you as a sense that there's, there are four data files that you will see there. Two sets of data in the diluted concentrated regime, and it's in the presence of salt and in the absence of salt so that's sort of a typical protein solution sample that you'd have. And the basic idea there was to give you data that is realistic so life is very well known and most people think it's very well behaved very very well characterized. And the idea of using the lifetime data was to actually show you how even a very well known supposedly well behaved sample can give you problems and how the data is not that perfect to very high resolution well defined profile that you might expect to give you a more realistic idea of what of what you can come across when you when you have your sense data and how we, how we deal with it how we analyze it and how we fit it. So, so that data is in that folder. Hopefully you downloaded it. If you haven't, you can maybe do it. Now which, and there are also two PDFs in there so one one PDF is called information, and it just tells you how the data was collected and what's your sample. So, think important things like the temperature and Q range and all that. So the second folder has a step by step instruction on how to fit the data so basically what we're going to do in the tutorial. Now in the tutorial I will first give you a little demo. If you've never used SAS for you, you when you open it to you might be a bit lost. There are some online videos that can help you but I'll give you a crash course of how to put data in there and how to start fitting. So I will show you that for example and then the idea is that you follow this step by step instruction. And then you try on the other samples on your on your own and I'll be there and can just tell me if you get if you get stuck. If that sounds like a plan. Okay, should we take a little bit here's from Helen. I cannot find a folder with the data. Okay, I think that as well I can send but hold on hold on. I'll fix that right away. Okay. I can write that in the. Here's a link in the chat now to the Indico page. And if you go to the Indico page on the timetable. And then today's exercise that there's the material that I received from Susanna. Plus that there was one document that was in the email that I sent out with with all the intro information zoom info for for the entire week. I'm sorry, I don't know if I'm. Is that okay Helen or yeah I mean I just because we said under exercise it would be or. Yeah, I don't remember if I put it under it's it's it's like there is like a paper clip. There's a paper clip. You can download them. I can also email it to you if it's easier. It's not a huge. I see paper clip on some of these but not on. There is one on the exercise on Susanna. Yeah, there's a paper clip and then there's four sub documents there with the ending sub and then there's yes. Sorry, I was in the long day. Yeah. It's okay. Okay, in the meantime, can I ask some questions. The first question. I don't understand why the playing wavelength after they meet this spot and then they become circles. Okay, let me share that again. Let me share. So this part right. So basically, don't get too worried about this this is basically just a representation of how we interpret the data but I don't I think Frank actually showed you an image of a beach right where you see a sort of a wave coming up to a beach and then you'd have some point interactions with the with some spacings and then you'd see the effects of the shame I don't have that image here but basically you can think of your incoming neutron so your neutrons are particle but it's also wave right so you can think of your incoming neutron as a plane wave and you can imagine it imagine what you would get imagine if this is water what you would get this as this plane wave hits a point, if you will, then the results is a wave or a change in momentum that we interpret as a scattered wave, because that's the that's where we can measure the results basically that's how we interpreted. That image here. It's just a way in which we interpreted, and it's also it comes from the fact that the wavelength of the neutron is so much larger. So it's a factor of four right larger than the size of the nucleus that it that it's going to interact with. And for the neutrons that it's as if the nucleus was a point scatter right. It's different for example from from x-rays where the interactions with the electron density cloud, which is much much larger than the size of the of the nucleus. Have a look at if you want to have a more intuitive or visual idea of you want to have a look at in Frank's. I believe it's in Frank's presentation otherwise drop me a message and I can send you that. The image of the beach where you see the incoming waves that you can think of as your plane waves. And there's a sort of a wall just before the beach with some openings and you can see the effect of the of the interaction on the beach so you can see these the tops of the spherical spherical is that'll give you a more intuitive feeling about the type of interaction, but basically it's just a way how we interpret it right it's just a mathematical treatment that we use. Yeah, thank you. And then second question. So you talked about a lot of parameters we can get from this experiment. The molecular weight on radius of duration surface volume factors structure and are there any other parameters we can get all these are all. So you can, you can infer a lot of a lot of information. So the other important parameters, the types of interactions that you have. And this is also very powerful thing about sense and so not only structure, but we're also measuring types of interactions between different molecules in your sample which is is very important in biology because it controls how the sample behaves right. So for example, the reason why you have aggregation or do not have aggregations because of the types of interactions that you have between molecules. So for a formulation, this is, this is critical or for the behavior in terms of phases of your of your sample. And again, because because we can look at, we're not limited to solutions I keep giving this example of solutions but we are not limited to solutions and neutrons are very penetrating matter. So we can look at dilute solutions but we can also look at concentration concentrated solutions and we can look at at a powder we can look at a solid. So there's a broad range of applications there. And yeah, I remember you also say that we can also get concentration. So, sorry, no concentration. You cannot directly get that so that we expected you know that about your, your sample. If you're, you can get molecular weight information if your data is well calibrated. And typically we do that by measuring a second sample data on a second sample, for which you know already the molecular weight with accuracy and for which you carefully measure the, the concentration so we can actually calibrate to get a semi accurate measurement of the molecular weight which is interesting for example if you have, if you want to know if you have a timer or a timer. So what type of sizes of molecule you have in the solution. But the sense is not the technique to, to, to measure concentration, not really, then you should use other techniques to to get more accurate information about that. So in fact you should come into a science experiment with us as much accurate information as possible from a range of techniques so you should know the purity of your sample you should know the percentage of deterioration we should know the concentration and average size, what to expect. So you should, you should come into your experiment with the with as much information as possible and again because there are a lot of parameters and to determine and, and comparatively not much data from sense if you want to have an accurate answer to your, to your you, you need to use complementary information so that you have a full story, because it is possible to fit a sense curve with the wrong model, and you can have a very nice fit but you and the model be wrong. So you need to, to, you need to know that that model makes sense for your, for your sample. Yes, and last question, what is this sense, the difference between the sense and neutral reflectometry. You can almost think of reflectometry as a specific case of sense where you have your, your, your, your molecules aligned or structured in a, in a surface. Okay, so in that case you're, you're looking at a surface or at your orientation of a molecule in a surface so that the type of sample and the, and the experimental setup is, is different. So you'd be looking for example at, at how a molecule binds to a membrane. And that's a very common application of reflectometry. So you're looking at reflections right so you're looking at the, at some surface of interest, which will be for example, a memory. So for example if you wanted to know the orientation of a protein that binds to a membrane. How does it orient when it comes into that to that surface then you would use reflectometry. For example, in solutions because it's all the molecules tumbling in solution it's all averaged, you would not get that sort of information. Okay, so the resolution is actually higher. Hopefully you can see my screen. Yes, we can see SAS view on it. Okay, so shall we get started so hopefully you've downloaded the data that should be for files. And they're not very large files. And you'll see one is called one lice, the other one is 100 lice, and then there's one lice and ACL and 100 lice and ACL. So they're in pairs, there's a dilute and a concentrated solution that were prepared in just T2O buffer. And then a dilute and concentration solution that were prepared in the presence of 150 millimolar salt. So if you've never used SAS view before. I don't know if you have it open. But if you do, you will see that on your left, there is what we call the data explorer so this is where you sort of manage your data where you load your data. And where you tell the program what you want to do with it there are different options at the bottom here. That we'll get to in a second. And then you'll have a fit panel on your on your right where you control where you're going to control your fits. So the first step the first thing you do is you load your data. I don't know if you can see the mouse but you do it here where it says load data. And if at some point you cannot so I'm just going to give you a little demo demo now, and then you will try it on your own. If at some point you cannot follow you cannot click and do exactly the same things. Don't worry just look at what I'm doing because afterwards you're going to do it on your own. And you have that file that gives you the step by step instructions and it has a little print screen so you know what to expect. So you will be able to do it on your own and I'll be here as well if you get stuck. But if at some point you're you can't follow don't worry just look at what I'm doing and then you're going to have time to do it on your own. Okay, so let's load the data there these dots sub sub is for subtracted because we've subtracted our background. So we're going to load that and you can load all of them in one go, because you can choose which one you're going to look at first. So they're all there. And one thing you have to be careful when you're using SAS view is that it's very often by default select data and you you want to decide which one which which data you're going to look at. So to make sure that you don't have some some some of these data files clicked by mistake. I always start by going here it says select and I go unselect all so that I then have to specifically go and click on the one that I want to look at right so that nothing is clicked by mistake. And you can see if you press the little plus here that you can open the information about the data. And then you'll have your data there and fit results once we have some some fits will be listed there as well. Okay. So we'll come back to this in a second. So let's start with this one license I'm so so I'll do it for I'll do the fits for this one and then you can work on on the other ones on your own. I'm going to click here. And I'm going to send the data to fitting. So there are different options depending on what you want to do so if you were going for example calculate the P of R you'd send it to inversion. I didn't tell you about this invariant option. But you can also do a calculation of your invariant you can do fitting you can do correlation functions are different options. We're going to focus on the fitting option here. So I'm going to send the data to fitting. Okay, and you can see that this this fit panel now becomes active. And it says data loaded from one lives dot sub so you know that this is the data that you think. And you can see that you have different tabs. So you have model fit options and resolution. Here's where you decide which model you're going to use. I'm just going to have my step by step guide here, so that I followed more or less the same steps as what I'm going to tell you in the guide. But we're going to so this is license I'm data. And I'm going to assume that you don't have the PDB structure available which is often the case when you look at a new structure. So let's say that we we kind of know that it's or we expected to be a sort of an ellipsoid shape so not a perfect circuit circular shape, but an ellipsoid so we choose ellipsoid. And now SAS view has a number of models already loaded onto it so you can, at least to a first approximation, you can load these to see how it fits your data and then if you need a more specific model or if you want to do your own. There are a couple of options to, to, to introduce your own model. And it's not too difficult. And, and again, if you, if you don't know how to do this, just give us a shout because the developers are very responsive and they can help you do that. And at least once a year we have a sort of a hackathon where we try to help people do some modeling so okay. In our category we said it's an ellipsoid and the model name is going to be ellipsoid again, and you see the populates a number of parameters and you can sort of change the thickness of the column so you can read it all. Okay, so remember that magic formula that I kept showing you that the intensity is proportional to the volume fraction to the form factor to the structure structure factor there's a background so this all comes back here in your fittings. So for scale here which is your volume fraction background that's the background of your data scattering like density of your model here scattering like density of your solvent. And because we have an ellipsoidal model you have a polar radius and you have an equatorial radius okay. So for scale. It's the volume fraction or solution is one meter mil. So that corresponds to a volume fraction. So it's point one percent right so point zero zero one. So, watch out for the units the units are here on this last column. The value of the parameter is on this one and then you have a minimum and a maximum so if you if you have a range that you expect the parameter to fall you can also introduce that here. Otherwise you just leave it as default if you don't know. Now, when you take on one of these options, it'll it'll fit that parameter if it's not ticked, then it will not fit the parameter. For example if you know the volume fraction and you're sure of it so you know that it's one make the meal you've measured that you've confirmed you do not want to fit this parameter ever. Then you can right click and constrain it to its current value so that you don't leave it tick by mistake. So that's just a little little trip that you can use. Now your background I told you that the data has background subtracted so to first approximation let's set it to zero now. Now, the intensity of your protein. Now, a priori you don't know it but I gave you a few links for software that you can use to get a rough estimation of what to expect. And if you had use it like for license and you get to a value that it's approximately 3.4 times 10 to the minus six. So it's already times 10 to the minus six here on the unit so you just put in 3.4. So it's 6.4 times 10 to the minus six. And again you can find that in a literature but you could also calculate it very easily. And then you have this ellipsoidal polar and equatorial radius. And for license I mean I give you that in the instructions for a first approximation you're going to leave that at 20, and you're going to put the equatorial one 10, just as a first starting parameter. And I'm going to take the polar radius I'm going to say that I want to have a look at that one. So it'll be the only parameter that it will fit so you just click on show blocks just so that you can see what the data looks like so you have your data here in blue. And in your orange you have this what the fit for the model would look like so it hasn't calculated a fit yet this is just the calculation that it has done from the parameters that have input. Now another interesting thing as well and important is what what is the key range that you're going to fit. So if you look at your data you can see that you have a very, very strong upturn here and this is one micrometer license so you don't expect aggregation. Although it's still possible but you don't expect it necessarily. And then you have it's very noisy data because it's not a very it's a very dilute sample. The protein is not very large so it doesn't have a very strong cross section. I think it's noisy again as you as it flattens up towards the background and again like I promised you, I did not give you very beautiful perfect data I gave you realistic data where you would, you'd have to deal with these things. So it's, it's kind of noisy data as you would expect for a very dilute small protein solution. I mean you're going to do your fit you have to decide which key range you're going to use. Now, I don't know what happened there and my guess is that there's probably air bubbles in the solution so it's it's contributing to this low key regime because we're only at one micrometer so I didn't expect aggregation. And then it gets noisy here again as it flattens out to the to the background. So we're going to fit the data, including this part here at low Q and this noisy part in the, in the high Q so you go to the second tab that's called fit options. And we're going to limit the, oh, sorry, the Q range to point one, and then up to point four. And you can, you can see that this, this fit line is now much shorter. So we're going to try and fit only in this range defined by the two lines. So we quick, oh, sorry, and at the bottom here you have the residuals so the residuals are simply the difference between the model the orange curve that you see there and your, and your data. You can also see here on the, on the fit panel is a little fitting error this chi square here this is a reduced chi square so you want it to be as low as possible so it's it's basically your regular chi square. You can think of it as a goodness of fit if you will. It's basically your regular chi square and if you hover it gives you the definition. You can actually wait by the number of points in the number of parameters you're trying to fit. So if you take on more options, it'll, it'll, it'll be fitting more parameters and this will affect the calculation of the chi square but you want this to be, well ideally zero but as low as possible. Okay, so I'm going to click on fit. And you can see that it adjusts the curve so it's trying to fit the data and it's not yet quite perfect. It's kind of okay but not perfect. So things that you can do you can say well, I don't think the background is exactly zero so let's try and fit the background. Now one thing to be careful when you're doing these fits is that these parameters are often correlated. And you can see that SAS view actually separates the scale in the background from your very specific model parameters. So, so remember that magic formula, your intensities you have your, your volume fraction. So your scale in this case that is being multiplied by, by your form factor and your structure factor. So if you, if you change the scale, and you change one of these parameters at the same time they're correlated. So you can bias your, your results so you don't want to fit too many parameters at the same time so you can for example untick the polar radius here and fit your background and let it find more appropriate values that looks a little bit better. And now we can go back to our polar radius for example and fit again. Okay, and then we can say okay this is a sort of, you could say this is okay it's a reasonable fit for this quality of data. But remember I told you that you need to know what to expect and if the fit is reasonable or not. And as it so happens I looked at PDV structure for Lysosign and I saw that the, we don't expect an ellipsoid ratio or in other words the polar to equatorial ratio to be too large because I know that the structure should not have a polar radius that is so much larger than the equatorial radius. But the step by step guide also tells you is that you should try to increase the equatorial radius which I haven't fitted yet to a slightly higher value and try to fit again just to decrease this ratio a little bit because it's not realistic. Given what we know about the structure. So let's fit that again and see what happens. Now the chi-square is dropped significantly so I trust this much better and I can also visually see that it fits the data much better. We don't need to click the box. Sorry? We don't need to click the box if we want to fit this one radius of the last one. The equatorial? No, so what I did is I changed it manually. And then I left the other one tick to see if by increasing this one the ratio would look more like what I expected or what I know from the Lysosign structure. Okay, so imagine you had a very large molecular complex you had done for example EM so you know the aspect ratio more or less of your molecule. If the result that you're getting here doesn't make sense then you can change this by hand to see if you get something that is more consistent with the information you do have for it. And it does change so it's now more like the ratio of two, which is more consistent with what we know and also the fit to the data is better and the chi-square is also better. But I didn't manually change this one but kept it fixed. I only fitted the second one. Sorry, do you mind saying the numbers again because your screen is so small that I don't see it. Oh, sorry. So it'll be in your instructions but I'll say it again. So in the polar radius it now fits to 22.1 and the equatorial radius I changed it to 12. It was 10 before. So I increased it slightly to see how it would affect the other radius I changed it to 12 and then it fit to 22.1 instead of 20. And the chi-square now dropped to 0.65. Yeah, I don't get a chi-square. Do you get it here at the but I don't know if you see the mouse on the phone. Yeah, I don't get the number there it's just like the line. But I can look through the tutorial maybe. Do you have something? Do you have an option one of these parameters is sticks? So if no parameter is sticks then you won't get a chi-square. Yeah, I'll click the second to last one under ellipsoid. And you see your data there? If I click the shell plot I guess I mean I don't get the other one but it looks a bit strange. I don't have like the black lines I don't have those and it's entirely orange it's no blue. Okay, so what that means is that your data is not loaded. So check in this data explorer window on the left. So what you can do is basically start over go to the data explorer level on the left and click unselect all. So that you're sure nothing sticks by mistake. And then click on the one-lice.sub. So it's the one-lice.sub. So click on it and only that one and then click on send data to fitting at the bottom here on there. So it should open a tab on your fit panel. It'll open a second one probably called fit page two right and then you again you select your models. See so on your tab on the fit panel it should say data loaded from and it should then say one-lice.sub on the header of that tab. That means your data is loaded. And then if you click show you got it. No, there's something. It requires some kind of command tool. So I'm going to install that first. I don't know if something went wrong when I installed it. Let's do like this. I'm going to go on with the demo and then when I stop and while the others try to do their own. You can share your screen with me and we can try to do it. Okay, all right. So we've got that. So you're going to have to trust me for now, even if you don't see it yet on your computer. So we have an okay fit for a form factor. Now the residuals that you see here at the bottom like I told you they're just a difference between the fit and your data. And you expect it to be so they're they're plotted in terms of a number of standard deviations away from the mean. And you don't want them to be much larger than plus or minus three standard deviations. Okay, so this looks kind of normal. That's what you would you would expect. Okay. So now just to show you how to also model the structure factor. Now for this solution we're at one micromil. So we don't expect much contribution from a structure factor. I will just show you something. So let me just show you the one and the 100. So if it's too small you and you cannot see I'm clicking on the one lies dot seven on the 100 lies. Not so here on the explorer panel. And where it says plot I'm going to create a new plot. So I'm just going to look at the two data sets for the no salt concentration. Hopefully you see this this curve that just popped up. So the one lies is the blue data and then 100 makes for me we can see that the data is much much better quality much stronger. And you get this interaction peak that you see here. So now definitely there's, you know, we don't come up to a plateau at low Q there's definitely some interaction present so that you so you definitely want to want to fit a structure factor to this data set. For the one micromil, apart from this, this strange area here that I think comes from their levels, you wouldn't expect so much of a structure factor but we're going to see we're going to verify that. And I'm going to show you that for the for the one micromil, just so you see how you do it. Okay. So let's go back to our fitting panel. We chose a category model of ellipsoid we chose the model name of ellipsoid now we're going to choose a structure factor remember that magical formula we also multiply the form factor by the structure factor. And in this case you can see that there are different options of different types of potentials were already loaded into SAS view. They're not all here but again, if you if you want one that is not here we can get on that to the program. We're going to choose this one called hatred and I say. So here we're assuming that this ellipsoid model is a charge particle. And that there will be some cool interactions between the molecules. Okay, so then a lot of other parameters, so you see here the structure factor parameters at the bottom. Hopefully. Now, if you don't know what the model do does or what is the parameters that's using or what the parameters mean. My strong recommendation before you dive into the fitting. As if it were black boxes, you click this out option here, and that'll open a web page that you probably cannot see because I'm always sharing the SAS view window but when I click that open the web browser for me. In the browser you have very complete documentation on program, and it will also give you information about the, the, let me just share my screen, let's see if I can show you that you share. Okay, let me share my entire screen that's better. Can you see my browser. Yes. Okay, so when I clicked on help, this is what popped up. And then it gives you information about all the different, you know, structure factors, fitting data, and it's quite complete so you can you can really rely on it. And if you, for example, let's say you do a search on hater in the say let's say I don't know what the structure factor is to do a little search, you click on it, and it actually explains what each parameter is what are the units. What's the default value, and it will assume, and it then explains what the model is what's what's it designed for and very importantly, it gives you references at the bottom, and please do read about the model before you dive into the fitting so that you know, one if it's appropriate and to what's the way to apply the model so for example, let's go back to my SAS view. You can kind of see, for example in your model you have a radius effective you have a volume fraction and a charge temperature concentration, etc. And if you read the reference it tells you that when you combine it with a form factor, you should set the scale parameter to one because remember this scale is a volume fraction but it's a multiplying factor in that magical equation of ours. So we don't want to apply this scale twice, and you see that your structure factor already has a volume fraction parameter of its own, right, so you should set this to one. Again the reference tells you to do that. This is why it's important to read about it. In the background we're going to leave it as what it was, but you should set it to one and the reason why you should do that that's why it looks so so wrong at the moment is because otherwise you'd be applying to volume fractions right to be multiplying twice. So if I set that to one, you can see that it brings it down to the more reasonable value. Now the background we're going to leave it at what we already hit. Now it asks you now what's the structure factor mode and what is this question. So it's asking do we just do a pfq times sfq, you know remember the magical intensity formula or do we apply that beta approximation. And we're going to say yes we want to apply that beta approximation because we know, and that's the model that we're using that we're not going to have perfectly isotropic interactions between the particles because because it's not a perfect sphere. So we're going to use that. And the radius effective so basically the radius effective is the radius that the structure factor sees. So it doesn't. So it's kind of an average of this polar and equatorial ratios if you want. And we're going to set that to unconstrained so that we can vary them freely. Otherwise it sets a specific relationship between polar and equatorial radius and we may not be able to change this too much. So we set that radius effective. It's now set to 20. Now if you have a polar radius of 22 and an equatorial 12 and again that's in your instructions as well. If you calculate the volume of this ellipsoid and you work out an average radius you'd get something around 15 so I'm going to put that as my starting number and again you would know this by reading the reference it would tell you how to work with these parameters. The volume fraction it's one mic per mil. So we expect that to be point one percent. Now charge now this model again you know if you read the reference and if you looked at the help in the documentation. This structure factor model is made for it's not made for a very dilute solution so it's made for more concentrated solutions where the charge has a significant value. But we can kind of simulate interactions of this dilute regime if we put a very low charge so we're going to put it to 0.05. Again don't worry if it's on your instruction sheet so even if you cannot read it it will tell you what to put in here. And you can see that just without even trying to fit just by computing it it's already closer to your data. Now in your information sheet I will tell you that I collected the data 25 degrees so we set the temperature to 298K. Sol concentration is zero for this sample for the NACL samples we have to put 0.15 here. And this is the dilute constant of the medium we're going to approximate that with the 1.5D2O let's put 73. Okay. All right and now let's click just on the effective radius. Oh sorry and let's check that the Q range is still what we selected before it is. Just before we move on you can see that there's a third tab in here called resolution. So this resolution is basically where you can tell the program if it's going to use the instrumental smearing or the smearing on the data caused by the instrument and that information is on your data file or not. And the default is yes because you should use, you should add this smearing to your model I showed you an example of how to the impact of this into your model because otherwise your model is a very perfect uniform isotropic shape that is not realistic. Okay. So let's go back we take down radius effective. Click on fit. I won't see much of a difference because of course I was telling you approximate parameters already but it's it's done it's fit. The chi square is still pretty low if you cannot read it it now reads 0.65. It's pretty good. I could do things like trying to the background, try and refit the radio the equatorial and fuller radius so you could, you could adjust this further and further and look at the and look at the effects. For now, I'm just going to click here on the left panel on the data explorer if you so if you open the option for your for your data. I'm just going to click and select also that nothing is quick by mistake. You can see that now you have fit results in here. We have this M1 M1 is the the basically the curve for your fit. And you have the PFQ so your form factor and your SFQ so your structure factor that we've just fitted. And if you want to look at what that looks like you can click on this SFQ option. And you come to the bottom here and you say create new plots. And this is the plot of your structure factor and it may look very shocking because it's very dilute and you think but I don't expect much of a contribution of a of a structure factor and this is where you have to be a little careful. Because it's not showing us scale here right. So if you right click. And there's an option to look at your structure factor and click on data info. The information on your structure factor. And you'll see on the X column, the Q values and on the Y column, the structure factor values and you'll see that they're all very close to one point 98 99. So they're actually all pretty close to one it's just a scale of the plot that's making it look very exaggerated. All right. So then what you do before you go on is you right click again. And there's an option that says save points as a file so we click on that. And I'm going to save it as one. Nice. Let's say sq. I had already done this. If I like to exist already. Okay, so now I have it there. Right. Now, if I was now going to move on to the next sample and fit the next sample. I'm going to put it in the same windows and they don't disappear the information is still there. But if for some reason you close the program by accident or a crash for some reason. You want to be safe and you want to save your project before you move on to a different sample so you can save it. And a location that makes sense to you. Here. So that if the program goes by mistake for any reason you could open it again and you not lose your fit parameters. Okay. So then you move on to your to your next one at this point I'm going to let you do it on your own. So again you have the step by step instructions. Let me know if you get stuck at some point to try it for the for the other samples. Okay, so now I have a few suggestions of questions and things for you to think about. And if you don't have time to think about them. Now you can always email me and we can discuss. But I think now maybe I can look at. The long time. I can hear Tommy. I want to talk about Susanna. So, let me see if I can mute Tommy. See. Maybe so okay so the rest of you can maybe go on and try it. Sorry, let me see if I can mute them. They have comfort. I think I muted them. Sorry about that. I had a little bit of a question. So, from what I know about, like, high square values and goodness of fit and their sort of distributions, an ideal high square value should be equal to one. And if we're going to very low high square values like 0.5, 0.6 were essentially overfitting. So I wondered if you have some thoughts in that realm of things. So in this case, it's the reduced chi-square, so not just the standard chi-square, and it will be affected by not only the number of data points that you have, but also the number of parameters that you're fitting. So the number of options that you have ticked here when you're fitting your daggers. If you sort of hover the mouse over the chi-square, it gives you the formula, but also if you go to help it or tell you. So in this case, you actually do want it to be the ideal value would actually be zero. But you are absolutely right in saying that you cannot refit. It's extremely easy to refit SAMS data. And that's why it's important to not go into the fit blindfolded. So for example, this is Lysosim. I should know that the form factor for Lysosim should correspond approximately to a certain size. When I, for example, for the first fit, I was obtaining a polar radius of 30 something, which was three times bigger than the equatorial radius. But I know that from the crystal structure, I know that I don't expect that elongated of a particle. It would not make sense for a globular protein, right? So there are things that you should know so that you can help the fit. Most of the times we are overfitting, to be honest. But you have to keep an eye on your chi-square on the light of how many parameters you're fitting. The other thing is not to fit parameters that are correlated. So when I'm fitting different radii, for example, I should not be fitting the scale at the same time because it's a multiplicative factor in that formula, right, in the intensity equation. So if I'm fitting the scale, which is a multiplicative factor in that equation that affects everything, in that formula, I should not at the same time be affecting the radius because you can see how you could bias the fit by increasing more than lowering the other. So the possibilities are tremendous, but you do have to be careful about how you do the fits. There is a different option that I did not tell you about yet. So the fitting algorithm behind this is the Levenberg market, the algorithm that you're probably familiar from other applications. It's a gradient descent algorithm, if this makes sense to you. There are other algorithms that we can choose. So we can choose the fit algorithms if you go to the fitting menu. And if you use, for example, one called dream, the robustness of the fits is much stronger and it also gives you correlations between the different parameters so it can guide you as to what you can and cannot fit at the same time and if you're biasing with it or not. The reason why I didn't start with that one is because the fit is much, much slower. So it's a lot more robust, but it's much, much slower. So typically I would start with something like the default algorithm, which is Levenberg Markard, which is faster, it gives you a rougher fit. You can get stuck in a local minimum, but it's at least a good starting point and once you have some starting values, you could then move on to a more robust algorithm, knowing that the same fit will then take maybe 10 minutes instead of seconds, right? So it would not be your first approximation, but it does give you a lot of extra information about correlations between parameters, about how your fit is converging as well. Yeah, so you're right in saying that you can't overfit and that's something you have to think of. That's why I cannot overemphasize use complementary techniques and come to a sense experiment with some information about your samples and not completely unaware of what to expect. Of course, we can have surprises, maybe your samples, not what you expected, but then you have to justify that well. You need some supporting information to know what to expect. Does that answer your question? Yes, thank you very much. Okay, hopefully you guys are now going through the fits and trying them out. Can I ask you a question? Yes. You said something about constrain, how we can define this between the different parameter that we know they are somehow related and how we should use that in SAS view. So SAS view, are you still seeing my screen, hopefully? No. Oh, sorry. Sorry. Can I stop sharing for some reason? I can see it now. Sorry. Can you see it now? Yes. I'm trying to make it bigger. So SAS view tries. So remember that formula, let me open the page for that. Remember the magic formula that I keep telling you about, right? So what we're doing here is we have our data, which is our intensities, and then we're trying to fit the rest, right? So we're introducing, as a fixed parameter, the volume fraction, because we think we know what that is, we measure the concentration. So this is kind of your scale factor, if you will. The contrast, we input, again, as fixed parameters, but you could refine that. We put in the scattering length density of your protein, the scattering length density of your solvent. So we're inputting this contrast parameter as well. We're fitting the form factor as the ellipsoid, and then we're fitting the structure factor as that hater at MSA approximation. Now, as you can see, these are all multiplying factors, right? So if your scale that you have there in your SAS view is fitted at the same time as, for example, the different radii in your form factor, of course they're correlated, right? So if you lower your radius, but you increase the scale, you can fit the same intensity, right? So they'll be correlated for sure, but also other parameters like the background will be also affecting, because it's a sum, so it'll affect it less, but it will also be affecting how you do your scale. And what SAS view tries to do is it tries to put on top these multiplicative factors, so the ones that will affect your data more strongly and more directly on top, so that you know not to fit these at the same time as you fit the rest. But basically what it comes down to is knowing that when you're doing your fit, you are applying different parameters on these formulas. So any multiplicative parameter will have a strong impact in what you're doing. Your scattering length density as well, again, and it's squared. So if you start fitting your scattering length densities at the same time as you're fitting a scale, your model is going to be all over the place, right? Because you're fitting two parameters where you can increase one and lower the other and get to the same intensity, right? So just to give you a sort of intuitive feeling about what you're doing, you have to remember this formula and what you're fitting, right? So while the background, which is just a plus B factor that you add here, it'll be less sensitive. So basically that's what it comes down to. But also, again, I recommend click on the help, go on the documentation and look up the reference for the models that you're using because they will give you a lot of comments and recommendations. They will tell you this model will not produce a stable fit. For example, it'll tell you it's not designed for very dilute solutions. It'll tell you it's not designed for particles that are not charged. So if I had put a zero here, it would not get a decent fit at all. So the reference helps you and the documentation helps you know what you cannot do. But beyond that, you should just remember the magic formula and that if you're using, if you're fitting some multiplicative factor, you should be careful and you should not fit multiplicative factors at the same time. Okay. Just one more question. Can we constrain some, if in two different data sets, for example, I have done something hydrated, something hydrogenated and I know at least the radii should be the same for both. Yeah. So you can. So there are options. If you look at the data for a window here on the left, you can see that there's an option to click for batch mode there that I didn't click here. So let's say for example, we were just fitting a form factor. Let's say for a second now, just for a second example, we were not going to fit a structure factor. We just want to fit a form factor. And we know or we presume to know that the form factor will be the same for the one liter mill and for the 100 liter mill data. So we want to fit them both with the same structure factor. So then you would click there, you would load the two data sets and that would constrain the fit to have the same values for both. So you can do, you can do constrained bits. Also you can add again, we didn't get into it too much, but if you go to the fitting menu, there's an option for constrained or simultaneous width where you can define specific constraints that you want for your model. So I don't know, you may have some information from other techniques or you may have a hypothesis that you want to test. So you can for example, say I want the polar radius to be the same for all the samples, but the equatorial radius can vary between two values. And you could set that up. So you can actually introduce that as conditions that you want. Beyond these minimum and maximum values in your step-by-step instructions, I give you an example. So for example, this radius effective, I could tell it that I never wanted to be smaller than two and I never wanted to be bigger than 100. So I could set that here and do the fit and then it will constrain it between these two values as well. So yes, the short answer is yes, you can constrain it not only within the one model, but also from between different datasets you can do that. So it is quite, I mean here I'm just showing you a very general example to give you more of an intuitive feeling of how it works, but it does, it's a powerful software. Yeah, you can have many options. Thank you so much, Helen here again. So another practical question. So I'm looking at, I re-installed the first few, I don't know what happened. Anyways, so now I'm in the program and I'm under the fit option and you. Do you want to share the screen? Sure. Okay, let me see. Hopefully it lets you share. Yeah, there we go. There we go. Okay, so it's in the, in the tutorial it says to adjust the Q range to 0.01. But when I, when I'm in the fit options, I get the minimum range and maximum range. So which one of these should I? So your minimum range is basically your minimum Q. So that one you want to set to 0.01. And then your max to 0.4. So your max range is your maximum Q. Are you using a Mac? Yeah. Okay. So then if you go back to your model, then I'll look at your plot. Hopefully it should have limited. Yeah, there we go. It's between those two black lines at the window behind it. You can see there. Oh, you want to see the plot? Yeah. It's the window behind it. Behind that. Yeah. I can try to send you after this. I can try to send you a version for you to test. Because some versions of the program were best with some, depending on what, what Mac. And are you using Catalina? Oh God, I don't know. It's a very old computer. Okay. I'm going to invest in a new one. But I just haven't had the time yet. Yeah. So basically now you see on the plot on the right, that's your data and your fit. And you see these two black lines. So that's the limits. You can actually drag these lines. To this, you know, instead of typing, you can drive them. To find. Yeah. So you could increase that if you, if you think, well, actually, I could use a little bit more data. So you could do that. Okay. We're going to be careful not to fit noisy data, but you can, you know, just to test because you're playing with it now. You want to see what. Right. You could stretch that out. So you will have more data parameters, but, but you can see that they're in noisy or so it's, it's something you do have to play with when we have your, your data. But that's one of the reasons why I wanted to give you realistic dating or not a data that has a perfect plateau. You never wonder where to cut your data because that's not what's going to happen. When you, with your sample. Right. Very probably. Bio samples are temperamentals. Okay. Okay. But then I think I'm just going to continue down the tutorial. Yeah. Fantastic. Yeah. But yeah, let me know if you get very, very stuck. I can try to send you a previous version of. That might. Yeah. On your computer. Okay. Thank you. Susanna. I have a. Problem here. The. The. There is some information jumping out the CC command requires the common. And the line developer tours. Would you like to install the tools now? So I should install. Yeah. Yeah. Yeah. Yeah. Yeah. Yeah. Yeah. Yeah. Yeah. So I should install that one. Can you share your screen? Just so I have a quick look. Yeah. Right. Can you see? Oh yeah. It's very accessible. When did you get this message when you open? I tried to change the background to zero. The value at that point. That's not related to the fit. That's a program that I would recommend restarting the program. Okay. Thank you. Did you. I don't know how far you had gone, but if you saved your project, you can reopen it afterwards. Okay. Now this, this looks like a bug. So the developers of the program. They're constantly introducing new developments. And if you go on the website or drop me a message, if you don't know how to do it. You can actually tell them I'm trying to do this and it's not working yet. This error message and they will fix it for you. So they can have their, you know, they actually welcome. Comments and they update it all the time. Okay. Thank you. They are, they, they are very happy to hear from you. At any point, or if you want to do something that the program cannot do, you can send them a request and they can try to help you. Like I said, at least once a year, often twice a year, they, they organize a sort of intensive. Code camp where you don't necessarily need to know how to program Python very well. You can come in and say, I would like to introduce this model where I would like this option that the program doesn't quite do. And they can try to help you to introduce that. And then the idea is that you then, they then, you know, include that in the program and it becomes available for everyone. So it's sort of a community effort. So all the models that you see in there were developed by someone else and they became incorporated and they're available. And of course they always refer back to your contribution. Okay. Now, yeah, the print screens that, that I did were on my computer. So I have a Windows laptop here. So they may look slightly different on the, on the Mac, but the general options should be there. They should be the same. It should, it should look very, very similar on your computer. It does depend on the, on the version of the, on your, for example, especially if you're using a Mac on the. Which version will you recommend for the Mac? Because my graph also look like a Helene. Helene showed you her in her graph, but it also looked like mine. Yes. So which version you recommend for this? So if, if you're using a Mac and, and you haven't updated it for, for a while, it may work better with the previous version. Throw me a message after this with the version that with your Mac version. And maybe I can send you a, an older version of the SAS view and you can try and see if it works a little bit better. Okay. Thank you. I mean, ideally if you can obviously update your system, that's always the preferred option because the older versions are not being maintained anymore, right? But for, you know, for a temporary solution, then we can, we can try to do that. Hey, Suzana, I just have a comment about this of version as well, because like I have Windows 10 and I was trying to install the version number five that is, I think it's the one that you were using. Yeah. It just didn't open at all. It's just up, it started the first pop up and it just didn't start. So I start, I installed it and I had now the version number four. It's a bit different, but I guess it has the same functionality, right? It does. So they are constantly updating it. So to give you an idea, so version four is not that old. It's just that they're updating it so often that the version numbers move fast. And late February, March, I was still using version four. So it's not that old. It's just that they fixed little bugs and they improved little things. So it might look slightly different. But these basic options that I'm telling you about here should be there. Okay. Thanks. If not, if there's something you need that you cannot find it again, just give us a shout. We'll try and help. Okay. Thank you. I have a question also about this in this tutorial. It says that we're supposed to change the structure factor mode to these P times beta. How do I do that in there? Because I also have the version four. Okay. Do you want to share your screen just so I can see what you're seeing? Sure. We can find it. It's possible that it's not in there. Which is interesting. Because it looks like in the tutorial, it looks like you just kind of scroll down menu here. Yeah. Which person? This is four to two. It may not have the beta approximation. Okay. So if it doesn't have in this case, because we're just approximating with a little ellipsoid, it won't make a huge difference. Okay. In other systems, it would, you know, if it was a big micro molecular complex, it might. In this case, it would not make a huge difference. So you can still carry on just using the, so knowing that what you're fitting now is a structure factor that assumes uniform isotropic interactions. So your chi squares might be a little bit worse. A little. Okay. Yes. I don't see it there. And it should show up and next to the. For the person for. Yeah. Okay. So I know. So it was inserted more recently because people ask for it specifically. And now I just use it all the time because I, I see that the chi squares. For this case, it's slightly better. It's not a huge difference. But for larger systems, it really does improve. So if you're, if your molecule is much larger, you're going to, you're going to want to use it. I'm just going to share why you guys are busy. I'm just going to share some parts. And to plot the data so that you can see what the data looks like. So that's the no salt. I'm just trying to create questions in your mind. And that's the salt data. Same molecule. So I'm just going to put some thoughts on the right. The solution has salt. Again, one of the things you may want to consider when you're doing a science experiment, how much salt do I put in my bottle? It's the same molecule, right? You will see in the tutorials that I, I put some questions there to teach you to. Make you wonder about a few things. So in case, in case you were not. You don't have any questions and you just followed all those steps and everything works. So I'm just going to put some teaser questions in there to make you think about this a little bit. And there may be some, something about the slow queue upterm on the one minute per meal data that goes beyond air bubbles. We don't really know. But again, that's something I'll leave for you to think about and we can discuss at a later point, if you like. So one of the things that SAS view is in my view, at least at the moment, not greater is that producing very good results. It's a powerful fitting program, but it doesn't produce a nice pretty publication like. On the other hand, it does let you save everything, you know, the form factor, the structure factor. So you, and you can export that into your favorite, I don't know, origin, Excel, whatever you use, but yeah, pretty pictures is not the strong point of SAS view. You can do reverse fitting, but for publication, I would export the data and just use a different software. Thanks for the quick question. And so I follow through the, the tutorial kind of doing all of the steps, exactly in order. And I have a slightly different values for the polar radius than in the screenshots. Is that bad? So that is perfectly normal. So it's, they probably, you know, because in this case, we're not fitting a lot of parameters and we fixed a lot of them. So they probably will not be too different, but it is a fit. And the algorithm does randomize the parameters. So it's sort of, it's like doing an annealing. It randomizes the parameters, and then it tries to make them converge. So you do expect to not have exactly the same results every time. So depending on the order in which you fit the parameters, what you fit first, what you fixed first, it is reasonable to get slightly different. Because it is a convergence, right? It's not a straight calculation. So it's because of the way the algorithm works. And if you use even more robust algorithms like green, so green is a population algorithm. So you would see even larger differences between one fit to the next, but they should converge to similar values. You're doing the same chi-square. So not exactly the same as normal. Okay, cool. Good question. Okay, so we have, I think, 15 minutes left. So I think I'd like to take a little bit of time, even if you have not finished, I'd like to take a little bit of time just to discuss the data and if it's a little bit. So I realized I didn't show you how to obtain CRAT keys and PVRs and Guineas with SAS view. Another reason why I haven't done that is because if you look on the SAS view website, they actually have video tutorials and PDFs, detailed PDFs on how to use these different options. So you can use them later if you'd like. But I can quickly show you. So here on this plot, I just plotted 100 micromil data for the salt and no salt on the same plot. And just to quickly show you how to do, for example, a CRAT key, if you right click on a plot, there's an option called change scale. And if you click on that, then you can customize what you're plotting in X and Y and there are some preset plots. And one of them is Guineas, the other one's pro. And for example, the CRAT key, I'm going to select CRAT keys just because it's easy to read. Then it immediately shows you the CRAT key plot. So we know that 100 micromil for both samples, the protein is nicely folded. And this is not the normalized CRAT key. So here you don't have the Q times RG squared. You just have the Q. So the intensities are not divided by I naught. So it's not normalized. And that's why you see differences in the intensities here. So what this is telling us is two things. One is that one sample is slightly more compact than the other. And also the concentrations are different. So the intensity of the peak are different. And of course they're well folded because you have this nice shape that you expect typical of the global effort. But yeah, so just to give you a quick idea of how you could then go about and change it to do a pro, to do a guinead, and things like that. Another option, especially in the last versions of SAS view, that is not displayed automatically, but you can right click. And there's an option at the bottom that says toggle navigation menu. So if you click on that, so it's a very last option when you right click, then these options pop at the bottom. And one of them that I like and I find useful is the zoom. So let's say, let's put this back to our data. Let's do a log log. So say I wanted to sort of zoom in here to see what intensities are there in the background, then I could zoom in here, right? Then I could change it back. So you can zoom in and look at different areas of your plot that I like. And you can also change how you display in which areas. You could save the data as well. You have different options there that is not shown by default. And on the older versions, this little option for the navigation menu pops up automatically, but on the more recent version is not necessary. Okay, so just before we wrap up, I put some suggested discussion points in your instructions. And also here, I've plotted. So on the left, you have the one and 100 micromil solutions with no salt present. And on the right, you have the one and 100 micromil solutions of life assignments present. I don't know if you already thought a little bit about it. If you look at the two of them as to why they look different. So let's assume for a second that they're exactly the same concentrations. Why do they look different? Once you add salt to it, have you thought about it? Does anyone want to comment a little bit? On the bottom, what I'm plotting here is just the structure factors. Well, salt can interact with the charges on the protein and do a little bit of shielding of the charges from each other. So without salt, I guess there should be more electrostatic interactions between the different protein molecules to each other. Exactly. So it's a very simple thing. So it's just that when you have salt present, the charges at the surface of the protein are more shielded by the counter ions in the buffer. So they interact a bit less and the structure factor is less strong. So there's less interaction. So there's less contribution of the structure factor. So if for some reason you wanted to study your solution, let's say 200 mix per mil. And that can happen. For example, people looking at, for example, formulations of antibodies that are sometimes used in pharma at very high concentrations. Then you may be interested in knowing what's the behavior of the sample in those conditions. And you don't want to cause too much extra aggregation or you don't want the structure factor to completely dominate the scattering of your profile. Then you could use a little bit more salt, for example. So don't forget about the buffer, not just the pH, but also the salt that you have in there. And if you're comparing solutions that are in completely different ionic strengths, you may not be comparing apples without those. So I kind of did this on purpose, not just so that you saw differences in structure factors, but also so that you thought about the salts that you are using, because it does make a difference. And this model was specifically designed for monovalent ions. That too makes a difference. So if you had livalent ions, that can be adjusted on the model. And again, if you read the reference, it will tell you how to do this. So you can adjust the parameters to take this into account, but be careful with that too. The type of ions also matters, of course. So there was a suggested point of discussion was if you fitted the volume fractions, so the scale in your parameters, why was it, for example, less than 0.1 for the 100 mix per mil? I don't know if that's the result that you obtained when you tried your fits. When I did my fits, I ended up having a scale that was less than 0.1, which is what I would expect for the 100 mix per mil. And that will probably happen more for the solutions that are less than 0.1 for the 100 mix per mil. So why do you think that for the 100 mix per mil solution when I fit my volume fraction, I ended up having less than 1. What do you think happened? Could it be that there's also solvent inside the protein and that's not considered the volume fraction of yours? That's a very good point. So your protein will have solvent filled cavities. It will have also hydrophobic cavities. So what neutrons are seeing could look different, but also especially in the case that we see here for the no salt condition, if there's some aggregation and if you're not rotating or agitating if you want your sample while you're measuring, you could have some of the protein precipitating at the bottom while you're measuring. And then the actual concentration that the neutrons see is not the same. So that's not unusual to see. And it's one of the reasons that when I'm measuring scattering data, especially at higher concentrations, I measure the concentration before and after my SANS experiment. So remember neutrons are non-destructive so you can recover your sample and you can measure the concentration after your experiment just to make sure that there was no settling that your concentration hasn't changed drastically so that if you're doing a fit and you see something like this and you know if it's an error of the fit or if it's a genuine, so we always say that neutrons do not lie. So neutrons are telling you exactly what they saw and if what they saw was less of the total amount of sample that you put in, then that's exactly what they would say. So that's probably what it meant that there was some aggregation because we can see it in the structure factor as well. But again, in the information that I give you, I tell you that I measure the concentration after preparing the sample and I don't think I tell you if I measure it before or after the SANS data collection and because data collection on neutrons scattering so it's not like SCATs, the data like SCACs, the data collection can take hours easily. So it's plenty of time for the protein to precipitate at the bottom or for aggregates or larger aggregates to start settling toward the bottom and move out of the beam. Now another point, do you have any questions? If not, another point of discussion that I suggest in there is what would be a better structure factor for the higher concentration data, so the 100 mixed per mil in 150 millimolar. How do you thought about it? So we chose this Hater MSA structure factor and if you had clicked on the help, it'll tell you so this is a structure factor that assumes the charged particle and assumes a coulomb types of the interactions between your particles, but we were just saying that when we have a decent amount of salt in the solution then these charges might be shielded somehow by counter ions. So what would be a better structure factor? So what would be your instinct? If you don't know a lot about structure factors, and you're just approaching this, you've never looked at structure factors, you don't know what the software has to offer, then you can click on the options and it shows you the ones that are already preloaded and it will show you there's a hard sphere, there's a Hater MSA which is the one we're using now, there's a square well and there's a sticky hard sphere. So these are the ones that it has available. At the moment there are more sophisticated structure factors out there and we could introduce them here, but these are the four that are available at the moment and if you don't know what each one of them does, what I strongly recommend is that you first help and you go and you click on the structure, I hope you see my screen, you click on the structure factor and you look for information about each one of them. So let's try this structure. Thank you. Why is it on a dude mind putting the screen that you're working in and full screen? Oh sorry. Oh hang on. Does that help? Yes, thank you so much. Thank you. If you click, oh it's small again, but let me, I can make it bigger as if you go and structure factors, sorry on mine it's maximized, it just doesn't show it very good. So you have the different structure factors that it's offering and if you click for example on the hard sphere, it'll tell you what it does and it tells you for example that it calculates the structure factor of monodisperse spherical particles interacting through hard sphere interactions. So it assumes uncharged particles which may be for the case where you have the salt shielding the charges around your particle, your molecule may be acting like an uncharged or like a very low charged particle. So I would probably try a hard sphere structure factor for the cases where you have salt and you can try this on your own time, you can maybe try it. Instead of using the H-R-M-S-A, you could try the hard sphere and see if that improves your fit. And hopefully it will for the salt case. But again, you have to do a little bit of reading about what the models do. I mean you could just try and see what fits but then you're kind of blind and you can be unlucky that you fit your data and it's not a physical sense, right? So your model has some physical sense with what you know about your sample. And again, this was sort of a teaser so I kind of made you fit something that is not ideal. In the suggested discussion I asked you why didn't we choose a hard sphere factor for the zero molar sodium chloride case since the structure factor is not made for very dilute solutions. That's why because we expect some charges so we know that the protein, we know the PI of the protein so that's something you will know from before your experiment, you will know the PI of your protein. You will have decided what pH you want to use so you know if you expect your molecule to be charged or not. And you decide how much salt you put in there so there are a number of things that you already control to some extent. Okay. I don't know if you have other questions. You can always drop me a message. I guess one general question we're dealing with protein data in this case. If we're lucky enough to be working with a structured protein like lysosine there is likely a structure available from crystallography or cryoEM or NMR. So in what situation would you be setting the sort of structure factor models with SAS view and in what situation would you rather just feed the PDB into software like cryosome? So you can so let's say you had a model from EM for example or NMR you had a model from a different technique so if it's NMR you'd have an ensemble of structures so you probably have to use an average of your structure, get a PDB then what would you do you would add hydrogens to your model if it's not there, if it's NMR it will be there but if it's just an extra crystallographic structure it may not be. So remember you have to add your hydrogens to the model because as we've seen hydrogen or DTRI make a strong impact into your spectrum profile then if you just want to do a very rough simulation you can try to calculate the profile without any further calculations if you want to do a proper simulation you have to let the structure relax because you're using a structure factor from a different technique let's say if it was EM for example it's probably prior cooled it was blocked at a surface so you want to do a slight simulation to relax the structure to what you would expect for a solution and that you can do a cryosome as well so you do need to do a little bit of simulation work it's not too if you've never done a simulation you can probably still do it just following the instructions and again you can drop us a message or the SASC developers also welcome people contacting them and they help you with making it work if you need to especially if it's a very complicated but you would do that and then you would calculate a curve and compare it with this one so that's the ideal in some ways it's a blessing and a curse it's a curse because it means you're going to have to do some simulations to relax the structure and to have a good starting model for your data on the other hand it's a blessing because you have a starting structure so you don't have to do this more crude approximation let's say it's an ellipsoid which is a very crude approximation obviously we know there's a lot more details in the structure than that but you don't always want to see that so maybe you kind of already know what the structure is you just want to look at interactions and Amara's not telling you that and Crayom's not telling you that you want to know what types of interactions are going on because you want to stabilize the formulation for example then you don't want to use SASC you could still use that starting structure but then you don't want to use a fit of your structure factors and try to fit your data and try to find out is this an attractive interaction is it a repulsive interaction do I need to add more salt is it a question of varying the pH how high in concentration can I go before these interactions contribute too much so these are questions that you can answer with this and you can also use more complicated structure factors which have both attractive and repulsive interactions at different hue ranges so you can get to higher levels of complexity depending on what you're doing but that's the power of SASC is that you can look at the structure but you can also look at interactions between them all I keep repeating myself but it all depends on what you want what are you trying what question are you trying to answer but you can absolutely you can use models from other data it will, I'm not going to lie it will get more and more complicated structures become larger and if it gets too complicated or if you don't have enough data then you can use these kinds of simplifications where you just use a more generic model because you may not be interested in those structures per se just in interactions for example like we're doing now so yeah it depends what you want to see as always yes as always thank you any other questions I have a question yeah go ahead so in the tutorial it says to save a couple of times do you want us to send you these files or it depends for our own sake that's more for you so that if you close the program by accident or if you have a bug or something and you need to close it and restart so that you don't lose the fits you already have if you do have a question and you want me to look at something then it's useful to send me the project because then I can more or less see what parameters you have but it's more for you especially when you're doing a lot of fits I mean in this case we're fitting for samples but very often you're going to have 20 samples of different concentrations lots of different you're going to have a lot of data loaded onto your SAS view and it's very frustrating because you've already done 10 fits you need to close the program and then you lose all that so it's more of a I'm trying to get you into that habit of saving your project just to save your work so that it's not a you never know what's after it's just a good habit to have but also if you want to exchange information with your collaborators or ask a question to someone then you can always see if they have the data and they have the project saved true, thank you it's already 5 past so I do not want to rush you but I think we're going to have to end soon again just drop me a message because you have to digest all this information and if you haven't had time to finish which is perfectly normal just do it in your own time and then if you have other questions that you remember in the meantime just let me know I'll try to help, I'll do my best okay I don't know I think they may not be back yet let me just check the schedule I think you're not supposed to have anything else after are you? I think after this you are free so you can spend the rest of the day playing with SASV or not or you can just take your time and digest information a little bit because I know the program is intensive I'll check the schedule real quick yeah we still have some time am I right? do we still have some time? I'm happy to sit here and spend more time a bit too soon I think the schedule officially says 4.30 we still have some time I'm happy to sit here and sit my coffee and wait for questions as they arrive if you want to play with it get used to it a little bit more share my screen again just in case it's useful yeah sorry I lost track here it's only 10 in the morning so if this was live I would be walking around and peeking over your shoulders can you hear someone talking at a distance? I don't know if it's a question or if it's another question sorry the sound is not very strong a question maybe someone already asks this but can we copy feeds from previous sample to a new sample or we have to set up each time all these parameters you do have to set up each time so if you've done it before and you saved your project it will populate the fits that you've done before but when you're moving from one sample to the other you do have to type them in I believe I don't think you can often it does carry over if you're doing them in a sequence sometimes it does carry over but because you have to each time select the category and the model name it does tend to bring it back to default because it assumes that if you're fitting a different sample you want it to input that information so unfortunately typically you have to it does but if you save your project you only have two at once I don't think there is that option good question I'll ask the developers to have an option to just import those parameters you're going to tell me it's a dangerous option to have because many people they're input values but you're right if you have I don't know if you have 50 samples then it's nice not have to and again here things like the scattering intensity of the protein and the solvent I gave you the values to try but these were values that you can estimate and that I calculated using other tools I used SASE and I calculated what was the SOD that I expected and here's the PDV and I got this value of 3.4 in reality for your sample you don't know precisely the scattering length density of your protein because you don't know the structure you don't know how compact it is or not so this can vary a lot you have an idea but this can vary so if you're doing a contrast experiment you would want to measure experimentally what's the match point of your protein for example you would calculate it and put it in your proposal so you would say I expected to be let's say 30% but then you'd also measure it I ask you a question this batch mode you showed us and I tried it and I have a version 4 on windows 10 because I installed version 5 couple of weeks ago and I couldn't get it to work so I had to go back to version 4 but it didn't work in version 4 at least the things you showed for the batch beating tell me what you're trying to do so you have different samples I just tried the same I tried to sample and then keep the structure factor for one and add it to the next sample or at least use the first sample as an initial parameter for the next one as a reference but apparently it didn't work I'm just wondering why it could be a version issue or it could be the way you want to share your screen maybe I can take a quick look let me stop sharing when I want to add them close this one I feed the first one which is here oh ok I think I don't remember for version 4 but I think that you have to select so let's say you wanted to fit the one license I'm and the one hundred license I'm sorry I believe you would have to have them selected both so let's yeah those two for example before you click to send to fitting on batch mode so that it knows it's on batch mode and that those two are associated I believe is that what you did yeah I tried this also and when I see yeah there you go but they are not good but if I want to have this on a plot I have to again replace it yeah there is an option to you want to have them put it on the same what because I cannot see the fit now for example now I'm trying to batch fit and I use the first one as a as a reference and then I try this okay I say fit for example this parameter I'm not in changes but that's that's a different issue that could be because the fit is not converging because you haven't put approximate parameters in there so the equatorial radius is way off the scale is off the SLDs are off so it may be because it's not converging at all if you yeah it should be showing you're right when you say it should be showing you the both of them is it behind that window oh yeah there you go it's showing you the two fits so M1 is for the for the other license sign so it's on yeah that will be graph 3 I think so it's showing you it's not showing you the other data because the license sign is the one you're using as a reference it's just showing you the square for that one but it's showing you the line for the other for the other data yeah so it takes a little bit of getting used to managing all the window did it crash I just oh you stop sharing okay thank you but yeah so you need to have probably what I would do is choose the one you're going to use as a reference and do a rough fit at least with this 11 bar markers algorithm so that you have approximate values that would work so you know it's difficult and then use that one to propagate into the it is a more computationally demanding calculation because it's trying to fit more things at the same time so if you don't have a good reason to fit it that way then don't because you can be imposing things that you shouldn't on the other hand if you do have a good reason you expect I don't know the form factor to be the same for all of them then yeah then it's a way to increase the number of data to parameter ratio right so there are instances where you have good reasons to do that but again if you're trying to look at some data and it's not working for you for some reason because of a parameter or something do reach out because the developers do want to help and it might be just some silly thing that we're overlooking or that you didn't see and they will you know instead of wasting too much time trying to do a fit if you think that it's there's a bug or there's a format issue contact them sooner rather than later so that they start you into the right direction the other algorithm you mentioned is it this dream at least in version 4 call so what sorry can you say that the other algorithm you said dreams yeah it has a bunch of them so dream is a it's population it's a it's a different principle you're more likely to not get stuck so so with 11 remark and for something simple like this it's perfectly fine and for I would say pretty much any system as a starting point that's what it is right it's faster you can check quick quicker if you have a roughly okay fit if you have especially if you have a more complicated system you would want to then change the algorithm to dream it's more robust it'll take a while so it's going to it's not going to be a minute it'll take a while but then it has a lot of advantages it's more robust so you're more likely to find the real minimum and and also it gives you a lot of extra information on the correlations between parameters so it shows if they're really randomized or if you're biasing them so if you're starting if you if you've been fitting your scale and your radius at the same time that you shouldn't do it'll show you that it'll show you a strong correlation and some journals when you when you use especially if you use this kind of algorithms will ask you for that they will they will ask to see that to see so what is what's the correlation between the parameters can you show us this and so you have that data and if you want to publish the data you can show that you can see look this is convergent I am not biased to fit and that's what it is and you may and you will especially want to show that if your result is something that's a little bit more unusual or unexpected if you have other complementary data and it's all consistent then no one no one's going to question it it all makes sense right but occasionally your molecule will act differently in solution or but it will surprise you and then you want to show that you're not biasing that it's you trust your physical so yeah then you want to use but yeah but you are going to set it to fit and walk away from the computer and leave it there for a while it's more robust but it's going to be slower sure yes I'm fitting something that it's vesicle plus one like my cell plus disk and when I added I tried I I couldn't get it work on version 5 so I get back to fashion 4 yeah you can you have some when you change the algorithm you have some options to change the number of steps that it does so you can actually reduce the robustness at least to begin with so you can make it a little bit less demanding so it'll still output all those options but you can reduce the number of steps you can there are options that you can manipulate a little bit to make it less exhausted so that it doesn't randomize the parameters in such a large space you can you can constrain that a little bit and so when you have a really large systems you can simplify a little bit and that will work for most cases but yeah but it's still going to be slower for sure so yeah start with the default Levenberg markers don't go to dream straight away because it'll be frustrating you'll use that when you're more more confident of your convergence and of your model but not to the end otherwise you need a very good computer you may not have access to it I noticed sometimes it's it's just crash yeah now stick to Levenberg marker even Levenberg marker you can reduce the number of steps it's going to do but yeah it's a better option to begin with and of course use all the information you have from the data if you know the volume fraction or in other words fixed parameters that you already trust you don't want to fit too many parameters at the same time use as much information as you have and if you're very curious I have a whole data set at different concentrations of license if you're starting up your project and you don't have data to play with I can send you more and you can do a little simulation if you want and again it's real data it's not perfect data that will perfectly match your sample you will be asking yourself questions about the queue range and why does it fit better or not I didn't want to give you a perfect solution that would make you not ask any questions because your sample is probably not going to be like that if your sample is more well behaved then you're probably not going to complain anyway license time is also not a big protein so it doesn't have a very large class section that's the other question you can and should ask instrument scientists what's the lowest concentration that they recommend you to use because not all instruments will be the same so not all sources are the same so ask them what's the lowest concentration that they recommend you can see that the one bit per mil data was okay and you can simulate again you can use the software to simulate your forward scattering intensity your I naught and you can let the instrument scientists know listen I calculated and I can see it's 0.05 is this acceptable or should this give me enough signal to noise data in your instrument or what do you recommend and they will give you some advice again don't hesitate because they want you to collect good data they want your experiment to work so they're happy when you contact them with a 0.05 mid per mil example that you'll never get anything out of did you manage to go through all the data did everyone sort of finish good I should have given you more than I'm a bit low I guess that's okay if you start playing with the parameters you can see so obviously I gave you good starting values but if you start playing with the parameters you can see that the fits can escape your control very quickly so that should give you a feeling of how important it is to use what you know about the sound with these tabulated values for cross sections for absorption coherent and incoherent and also with the software that you can use to simulate the scattering profile all these these available software packages they simulate coherent scattering right so it's not going to be the same as your as your measured profile because your measured profile will have incoherent scattering contributions it'll have a background there'll be absorption so remember that too when you do a simulation you're simulating the coherent scattering so there'll be other factors contributing that will reduce the signal to noise would be nice if we could simulate then we wouldn't need to do an experiment if we could simulate the full thing Susan I had a slightly off tutorial question that's asked me in some of my samples that I want to fit in this ask you I need to add together models and I was kind of as you were saying with the at least with the structure factor including the structure factor you already have an issue of so you have scale in one model and then you have volume fraction in another if you're adding together models and especially if they are custom models how do you avoid that kind of problem again well it depends so if you're adding models it depends what you're doing so let's say you're if you're adding it depends on what your equation it goes on it boils down to your equation and the developers will help you implement your equation but they will not tell you what equation to use so you have to come in with that so if what you're summing is let's say you have two form factors and you're summing two models and there is no interaction between them so you can have a sum and you can refine the different volume fractions so the different contributions of each in reality very often there is a cross factor so they influence each other so it's rarely as straightforward as that and very often or almost always you have to refine that you have to refine the scale factor for the both of them and also very often you have a fit so you have the model for one the model for another and then you have a third factor that accounts for the interactions between them so they're not typically but again you do your best to try to come up with an equation and you read the literature and you try to come up with a model reach out to someone who is done models reach out to one of us if you come to like a SAS view code camp and you don't have your equation they probably will not be able to help you unless you talk to someone like you know unless you talk to one of them we will also do models for example talk to one of us basically and then try to help you with the equation what they do help you with is when if you have the equation you just don't know how to implement it on the software that's straightforward to them so they will give you advice on how to name your parameters how to code for it so the basic things also common pitfalls is that you name your parameters the same things as for the variables that are already defined in the program you probably came across that already so that sort of thing they will help you with they will not tell you how to do the equation that you have to reach out to one of us usually you have to refine usually you have to refine just a little bit of it when you say their interaction between the two models what do you mean I don't quite understand what you mean by that okay so probably as I just told you so like the if I'm looking at with the ones in SASTU at least combining a core shell model with just like a broad peak lorenzian to have like a core shell particle with internal structure what would be kind of the interactions there is it just kind of scattering from so there are two ways of thinking of it so the intuitive idea is that they're both present in there they influence each other this is the intuitive idea I'll get to the maths in a second yeah that they're both present in there they are in the same structure so if one is larger or if one is more charged it will influence the structure of the other so they're not independent if you think about the magical equation of the intensities what you're doing is instead of a single p of q you have two p of q's in there so you have p1q plus p2q and you saw I showed you in the equations that that's a square we square the amplitude oh okay so mathematically you have the square of a sum so you're going to have the square of the first term the square of the second term and then you're going to have the two times the multiplication of the two so mathematically as well you can see that there's a third factor in there where you multiply the two of them so even if intuitively you can kind of imagine that they would affect each other but also mathematically because of how we calculate this there's a we call it the cross factor yeah I think I got a bit too stuck on interaction rather than but yeah it's a square so it's a square of a sum so you're going to have a third factor that multiplies sometimes to a sum approximation you can get away it depends on the weight of that of that third factor if one of them has a much weaker contribution maybe but yeah okay cool the good thing is that if you come up with a good model and you implement it then it becomes it's a community effort right so if you've done it then you can just have it there and it's almost as good as you know you can send it to them even if you don't know how to do it send it there and it will become available to other people and again if you're very very lost and you don't know what to start a good way would be to find a reference with a similar structure what you think is a similar structure or a similar model and start there even if you don't think it's exactly the same okay implement that one because that will show you what you have to do and it will show you the pitfalls so you can work out through you can do some progress so you can work out through the debugging of implementing the model and then look at the fit it's not going to be perfect because you know yours is slightly different and then you can just change that so start with something that's published if you feel a bit too lost that you think is kind of similar like a starting point it's easier to do that way you know to build on a complexity than to start from scratch if you're a monkey and it's very different it's a different story but but usually at this point there's some model out there that's going to be kind of similar yeah it's a the hard part rather than the models definitely are there somewhere have you been to one of the code camps? no I haven't Karolina told me about when she did a code camp which sounds really cool yeah because I've talked to Andrew Jackson a bit about it as well I actually took a full up with that yeah try to get yourself into one of those it helps if you know a little bit of Python but you don't necessarily need to and you have them there for a week just for you so you can sort of take their brains most of the times there will be people there like Andrew who can also help you with the model part but if not just take a model that you think is similar and find out how to implement it because then you have an example to work from thank you complicated sample exactly these days even my eyes can be complicated it's not a beam time unless something becomes weird for no reason absolutely there's a reason why it hasn't been done that's my previous sketch the advisor used to say that there's a reason it hasn't been done yes my sister is also doing some research and her supervisor said it would be really great to do this and she went to a confidence level and said it's really great you're doing this good luck the thing to do is not let yourself get too stuck reach out to people sometimes it's a silly thing sometimes it's just a bug in the software and you're worried about your model other times it is something but people who have seen lots of different models may be able to do this yeah reach out don't isolate in the corner and get too stuck especially in these days don't have those many interactions I was just thinking we're a bit over time so is there anything else that we should do oh yes sorry no if you finished if you've gone through all the steps that's perfect again if you want more if you don't have data to practice from I can send you more or if you have questions later I'm happy to answer otherwise you're free to go and enjoy the rest of your day