 Thank you for that very kind introduction and for the opportunity to talk to you today. So what I was going to do is represent the work that's happened in our group at NIST over the last couple of years by a number of our members of our groups that have contributed to advancing the application of NMR methods to assessing, as was mentioned by Andreas, monoclonal antibodies, but biologics or protein drugs in general. This was a problem that came to us probably more than a decade ago now. There was a real need, and I'll go through this in my talk, how we come about what kinds of problems we try to address at NIST. But there was a real need that was brought to us by our many stakeholders in industry and the regulatory agencies of having more precise, accurate and robust methods to assess structure of protein therapeutics and more complex therapeutics. And as I'll go through in my talk, the structure, particularly what's called these higher order structures, the foldedness of the protein, if it aggregates or not, these are really attributes of a protein therapeutic that are quite unique to these types of therapies and we're not really something that had to be dealt with in the development of more traditional small molecules. So again, thank you for giving me the opportunity today to represent our group's work. And so let me start off again by just introducing, I guess most of the folks on the phone now or on the call have already heard from folks from NIST so you may know about us, but NIST is a non-regulatory agency of the US government. We're in the Department of Commerce and within NIST, we support sort of commerce development and advancing particularly manufacturing and underneath the way we develop programs, we is by working with stakeholders and again, getting feedback as to where we could actually develop measurement science standards, technologies that will accelerate the development and manufacturing of advanced products in our case under the bio manufacturing issue as shown here, the manufacturing of innovative high quality biotherapeutics. And again, we do this as NIST as creating a community around us where we get feedback, we try to address because we're a relatively small organization compared to the large ecosystem of biopharma. Where are the large infrastructural problems, particularly measurement problems? So we try to get feedback to say, what problem could we address at NIST that would have the broadest impact across the entire ecosystem? And so we try to go after those. And then we also look at problems that we identify that we can actually solve at NIST that are truly measurement problems where we could bring our measurement expertise to bear and also our sort of strengths in developing standards and harmonization of methods. And we always strive to promote cross industry collaboration and open sharing. So much of our work is at the pre-competitive stage where we're trying to sort of again, raise all boats with the work we do. So the particular focus of this talk and the series is monoclonal antibodies and by again, quick introduction. The monoclonal antibody is really the king of biologics right now in terms of the platform molecule is being most used to develop new advanced protein therapeutics. As shown on the left here is the sort of cartoon structure of an IgG antibody with its fab fragments at the top, shown here with blue and purple are the variable regions which are where the amino acid sequence will change to hit various targets. And then there's the FC domain which binds to immune cell receptors. So as a platform drug, this type of molecule could be readily developed and repurposed for other targets and other diseases. It can hit conventional targets and it can also recruit the immune system. So it's an immunotype therapies and so you get that initial sort of natural benefit of using this type of platform molecule. They've been highly successful in treating many forms of disease. And to begin because of their rapid developability these were some of the first therapeutics that came in response to things like the recent COVID pandemic. And again, we like to point out because we're part of commerce that these are a large part or maybe the largest part of what we call the bio economy. So the sales of these advanced therapeutics and now complex vaccines like the COVID vaccines really represent billions of dollars of sales every year. And so it's really not only a large health, public health issue to develop these therapeutics but it's certainly a large economic driver here too. The other thing that to point out which came to us when we started to ask questions about where we could provide new measurements or address measurement gaps was that starting, you know, in the early 2000s it was recognized that a lot of the monoclonal antibody drugs that first came to the market as innovators would be falling off patents. And so the so-called patent cliff was seen coming particularly members of the regulatory agencies the FDA EMA had to develop ways to approve what would soon be coming which are these so-called biosimilars which were essentially analogous to the generic drugs that were developed for small molecules but because biologic proteins were considered a situation where they couldn't truly be copied identically they're considered biosimilar and not generic but the idea was they needed measurements for comparability to address what would be coming on sort of follow-on molecules. And again, there's a large driver here in the market because you bring these follow-ons you could potentially reduce costs to the patients by having competition. And the basic idea there and where our strengths at NIST come in because we're primarily developing measurements in physical and chemical analytical tools to look at proteins is when you develop a typical innovator drug it's really driven largely by clinical trials which are long and expensive. And the idea on a biosimilar development pathway is the more you could learn about that molecule and sort of inverting the so-called inverted pyramid scheme as shown here the more you learn about it analytically the less or you reduce the need potentially for the extensive and expensive clinical trial. So again, with the idea of bringing the biosimilars to market more quickly and at lower cost a robust analytical foundation of your pyramid could actually help could help do that. So with these protein therapeutics there's a lot of attributes of the therapies that have to be measured revolving around essentially four areas the potency of the molecule, how well does it work? Its identity, the purity and stability. And as I show on this slide there's 20 to 30 release tests for monoclonal antibodies there's lots of measurements that are done. And these are again because these are complex molecules there's lots of different techniques that come to bear to sort of get a robust characterization of the molecules to gain confidence that you understand and can confidently describe each one of these areas. And what I wanna transition now to talk about for the rest of the talk is the area that we've focused on and all of those which is the identity area of the molecule and focus particularly on structure. So again, for most in on this call perhaps this is nothing new but proteins fold, okay? They're not like small molecule drugs there's a primary sequence these fold into secondary structures and those secondary structures fold into higher order tertiary folds and then paternal structures. And in situations which are often unwanted these proteins can also aggregate because in an aggregation event is something that's considered quite serious because it can induce an immune response. But this is really one of the primary unique features of these protein based drugs. And in order to be able to understand the structure and measure it in sort of an analytical regulatory context where you can actually describe the structure with confidence with a method of high precision that's robust where you can actually detect changes that are meaningful is really critical. So there was a need as I said about 10 years ago it came to us looking for new methods that could actually do this. And this is where we stepped in and said perhaps NMR could play a significant role here. And again, the measurement need is that we have to show that these molecules are folded and that there is no changes to the fold or unattended aggregation that would make it unsafe or ineffective. So for those who are familiar with NMR NMR has a very straightforward way that had been established for many years with quote unquote finger printing proteins. Anybody who's ever worked with protein biomolecular NMR knows it. You can do two dimensional correlation spectroscopy. And in doing so here I show the example for one small protein biologic you can correlate the nitrogen and proton of every amide in a protein. And what you end up is what showed on the lower right is this two dimensional spectra where the cross peak patterns that you see there really represent the folded state of that structure. So there's a unique folded state pattern to those cross peaks and each one of those represents a particular amino acid in the protein. And so what NMR really gives you is these highly sensitive localized reporters across the entire protein in this case the backbone of the protein and the sum of all those reporters and the sensing of the chemical and structural environment those proteins gives you a unique pattern that tells you that that's the folded state of the protein at that time. And so if we wanted to consider using such a method as this for the application in pharmaceuticals we had to basically demonstrate that we could apply this method in a practical way that it was general, it was very precise but also robust and didn't drift and sensitive to changes. And the other thing that's important and I sort of alluded to in this slide is it has to be applied to as provided materials. So oftentimes or most oftentimes in basic research NMR is applied with isotope enrichment of proteins which allows correlations to putting the more NMR active nuclear like N15 or C13 into your protein through heterologous expression but this is not really the case when you're dealing with pharmaceuticals you basically have to measure what's in the vial. So that means for NMR can we do these kinds of experiments at high natural abundance. So the other thing that we were facing and so we could show early on that two dimensional sort of classic nitrogen proton correlation mapping would work very well for small proteins but the challenge the physical challenge or physics challenge at NMR is as you go to larger and larger proteins your lines broaden because your tumbling slows down. So these molecules tumble more slowly in solution the correlation time increases the lines broaden that's shown through a simulation here which shows you from the small 19 kilodolm protein up to the monoclonal antibody how the line widths in this simulation would be expected to broaden. So you brought in your lines you lower your sensitivity and you basically create a specter that are not as well resolved. And so really what was most interesting to the folks again because it was the most dominant platform molecule was whether we could apply NMR to monoclonal antibodies. And so initially we made an attempt to look at N15 correlations at natural abundance of things like monoclonal antibodies and I think you've already heard in an earlier talk from John Schiele about NISMAP. So a lot of what I'm gonna talk about today is focused on applying methods to NISMAP which is our publicly available standard which allows us to do a lot of these benchmarking of any of these analytical techniques. But as you can see here even if you're not an NMR spectroscopist the number of signals that you get from an antibody compared to a small protein goes up and the quality really goes down the lines get longer. And the timeframe it takes to take these experiments in natural abundance was multiple days which is really not practical. So if you can't get actionable information in a reasonable time, folks in industry lose interest very quickly in your method. So at the time when we were initially starting and seeing the performance of the N15 correlated experiments we decided to pivot to attempt to use carbon 13 correlations. And this is something else that had been explored broadly in the field of biomolecular NMR. Methyl groups had been used to look at large protein systems. There's some intrinsic advantages to using methyl correlation spectroscopy first of which is it's a slightly higher natural abundance but more critically the methyl groups have this intrinsically favored relaxation because of the rotation around the methyl bond. And so they can actually give sharper spectra than one might anticipate for other residues in such a large protein as an antibody. And the other thing as I show here it's in these sort of salmon spheres where all representing all the methyls is like the amide correlations the methyl correlations really do represent or sample the structural space throughout the entire protein. And so in 2015 which is now almost eight years ago we published the first paper to show that we could indeed collect these two dimensional carbon correlated spectra. This was work at the time that was done by Luke Arbogas who's since moved on to join Eli Lilly and Rob Brinson who stood with us and they basically showed that indeed we could collect these C13 correlated spectra similar to the way one would collect N15 spectra. We could actually because there were fewer methyls we could actually see almost all the methyls we would expect in a given spectra and we could collect these data in a reasonable timeframe in a course of a couple of hours rather than days. And we also did quite a bit of work to show that the N15 and C13 methyls data sets while orthogonal seem to be sensing similar aspects of the protein structure in terms of representing the structure in the folded state and also sensing the differences. So we had a much more sensitive fingerprint we could use in methyls and from then on this is where we've been focusing our attention on carbon correlation. And you can collect fingerprints of the full map. Anybody's can be fragmented into their FC and FAB fragments that are shown here. This is FC and FAB fragments and you can basically look at the cleaved mixture, the FC or isolate either one of the two fragments and look at them individually. This will become important at the very end of the talk when we talk about how we could put assignments on these spectra. So far what I'll talk about is how we use these spectra first in a non-assigned mode. So just in the pattern understanding the patterns but eventually again with NMR which is unique to the spectroscopy we could potentially put labels on each one of these peaks that assigns them to specific correlations within amino acids in the protein. So just to quickly highlight what we have to do when we develop these kinds of approaches is first of all, we want something that's broadly applicable. And Eva Bon who's a long-time collaborator of ours who's also been pioneering some of these fingerprinting methods with NMR was able to show that these carbon 13 as well as N15 correlations could be broadly applicable to any FDA-approved drug. He used about a half dozen or more pharmaceuticals they could get from the pharmacy. They have an easier time of it than we do because they're a regulatory agency in getting FDA-approved drugs to research. And they were able to show that as we might expect for as an NMR expert would expect that these methods are general. There was nothing special about the methods we developed for NISMAB but it could be applied to basically any antibody in the IGG class. And again, this is a nice feature for us because we're developing methods and because it's a class molecule any method we develop on NISMAB which is a representative of that class we would hope as we saw in this study from Eve that it would be broadly applicable. The other thing we need to show when we develop these methods is that they're actually robust and they're reproducible. And so Rob Brinson again led an intern lab study that NIST brought together groups from around the globe to basically show, it's a NIST does this quite a bit to ask the simple question if you give everybody the same sample and ask them to make the same measurement can they do it? And does everybody get the same answer? And so it seems like a fairly simple Monday in exercise it's interesting when you actually go through these exercises which you find and where the variance comes in but this really helped to and all of this work was published with about 48 co-authors and it basically showed that the method the NMR method was actually incredibly stable incredibly precise in terms of the spectra that people could collect across different NMRs at different field strengths I think as low as 500 up to 800 or 900 megahertz and that by doing this type of exercise we also could make suggestions about best practices for harmonization of the method and to gain confidence and broader more rapid adoption because everyone could see that and particularly when there's a question of how well will the method work from one lab to the other between different sites? Is it something unique? This will show that the community came to a consensus of how the method could be best applied and that's basically the way NIST works again because we are non-regulatory so we work together with our community stakeholders to get adoption. So we were able to show that the method is broad and applicable. So the next, what I wanna just point out too is that one of the strengths of NMR is that NMR can do selective spectroscopy so we can look at subsets in a complex mixture of signals and in particular all of these methods we'd like to be able to apply to as provided drug formulations. In order to do that we have to deal with these pesky things called excipients which are put in there to stabilize the biologic. These can be a large fairly high concentrated things that are added to the drug formulation with signals that are orders of magnitude larger than the protein signals that could potentially obscure our protein signals which were particularly interested into fingerprint the protein. And so here's just some examples of some typical excipient sucrose, sodium acetate, thionine, polysorbate and you can kind of see the artifact signals that come from each one of these spectra particularly the sucrose and this is our fingerprint region and if those kind of things can really interfere with the ability to use the spectra and to actually particularly use them in statistical and PCA type analysis which I'll point out. So to address this Luke Arvigas took some work that he began as a graduate student when he was at Hopkins looking at selective excitation two-dimensional selective excitation. Again, I just show the pulse sequence but it's not really important to go into detail but the idea is one could actually use it in a two-dimensional spectra, selective pulses to basically either subtract away or excite certain regions of the spectra and what Luke was able to show was that by using these types of two-dimensional selective filters that he could actually essentially blank out parts of the spectra that they're showing us here an example that you could basically selectively excite and invert the signals that you're not interested in and then they could be removed together with some additional post acquisition processing. So you could go from a spectra with an artifact signal where you could effectively remove that signal in the two-dimensional spectra and then use the remaining signals robustly for fingerprinting. So since these are typically small molecule excipients there were other ways to potentially remove them. You can remove them based on the size difference between the large protein and the small molecule. These are typically done, we're done with filtering with diffusion. Small molecules move faster than big molecules and other types of filters. What's important with the Sierra filter is not only does it work very effectively but because it's really a selective, a selection-based removal filter and not based on diffusion or relaxation is that we actually retain almost 100% of the signal to noise in the spectra. So the filters often come at a price, as you can see at the bottom, the signal to noise goes down at a factor four in some of the pulse field gradient experiments. So again, as I mentioned early on we're doing these experiments at natural abundance. So we're constantly fighting signal to noise in the NMR to keep that signal to noise as high as possible. So it really is this filter really, again, other filters could potentially achieve the same goals but this filter actually can do it in a way with short timeframes for the filter such that you don't lose the signal to noise in removing your signal. You don't lose the signal to noise for the rest of the protein signal. And so the last thing to point out is that by the nature of the way this filter works you can remove multiple selective regions of the spectra that may be artifactual regions that's shown in gray here simultaneously with the filter all at once. So with the Sierra filter we really can apply these 2D methods to the as-provided pharmaceuticals. And we've shown many times over the years now that we can robustly collect these spectra at natural abundance and select the signals of interest and remove the signals that are interfering. So the question of course then comes the application of NMR methods can we detect and quantify assigned any structural variation from our spectra? This is always a question, how senses will the NMR be? And really somewhat philosophical questions of what does it mean to quantify structure? So, which is essentially how analytical chemists think but typically structural biologists do not. So in order to sort of test the NMR method we have again our test molecule mismab and one way to look at structural variation is to sort of do what's called glycan remodeling. So the FC portion of every monoclonal antibody has a single glycan site and when glycan and there are enzymes that will allow you to trim back that glycan position and then you can look at the response. There's a large interest and many studies that have been done looking at the glycans of FC because of their role in the function of FC. For our purposes, we just remodel them to look at their structural impact. The other example I will point out to show you how NMR could be sensitive is to look at oxidation. Oxidation is something that's of concern in the development of antibody and all biologic drugs as a potential in the manufacturing packaging delivery that the drugs can oxidize, particularly methionines and this could have potential effects on stability activity and aggregation. So again, for the glycan case study what we could do is we could take our NISMab using exoglycosidases, we could cleave the map off and show them by mass spec that removed it or trimmed it back. So we change again, this shows the mass spec pattern where we change the pattern of the glycans. These are complex glycans that are on these. So in the first example, we basically just trimmed back the terminal sugar and in the second case we kind of hit the protein or hit the antibody with a sledgehammer and removed the entire glycan and basically changed the charge at that position. So in one case we have a fairly subtle change to the glycans, the other fairly significant change at that position. So we can then look at our spectra and compare the two and what you can see hopefully is that in the case where there's a significant structural change visually on the left, you can see the difference between the blue and the red spectra. It's fairly obvious that these spectra are different but for the spectra or the samples where we just trimmed the glycans back, it's actually, it's not obvious to the eye which whether these spectra are the same or different and on the right, I show you some of the initial work that was done to analyze these types of spectra beyond visualization, which was linear correlation. And again, even using this type of an approach you can only really discern with confidence when there's a significant structural difference which you could already see visually at the top but if there's a small subtle change potentially this was not able to be detected using these linear correlation analysis. And I won't go into it but these methods or the correlation methods like this are much more sensitive to signal to noise. You need very high signal to noise to be confident that you're seeing a difference. So we basically moved away from those types of analysis, comparative analysis of the spectra to use PCA analysis of the spectra. So using principal component analysis, again, you can distill those two-dimensional and I'll show even one-dimensional spectra down into principal components. If we then look at the differences as a function of these so-called principal components one, two, and three that shows you sort of the difference what's differentiating the spectra that have been sort of distilled down to these single points on this in-dimensional space. And so what you can see here from applied of the principal components of PC1 and PC2 for our glycan example is that if you look at the native sample, you get a clustering here. The E-glycosylated which removed the entire glycan grouping up here and then the samples where you had just trimmed glycans, we see clustering here. So the PCA actually provides a very robust and easy way to readily separate even the spectra where visually we couldn't tell the difference visually and by linear correlation. The other nice feature of the PCA analysis is we could go back and using PCA loading pots which is basically sort of plotting the PCA component back as a two-dimensional spectral representation. We could see the aspects of the two-dimensional spectra that actually was contributing to the difference. So again, you can plot the PC, make the PC plots, look for clustering and then look for the principal components and overlay them back on your original spectra here shown in basically red and blue. And those sort of peaks show you what aspects of that two-dimensional spectra contributing to the differences that are seen in the principal component that are distinguishing the spectra. And so you can look in your spectra and try to understand where the differences are coming from. So it's something that could be used either in just sort of a default mode where you look for the differences, but you can also then look at that PCA loading plot and see where the differences are coming on your spectra. So we almost do all of our analysis now by PCA. It's an excellent way to look at many spectra. This is not the first application. There's many, many applications of principal component analysis for NMR and other sort of techniques. Some of the nice features is that certainly it can be applied even when the visual inspection is not possible. It also points out that in many, what we're also trying to achieve here is we wanna take the method such that a non-expert could use it and you don't need expertise to look directly at spectra because typically an expert, an NMR expert could look at a spectra and see things, but we wanna bring it more broadly to users who may not be experts. The PCA loading plots really can tell you where this spectra are coming from. We use the total points of the spectra. So there's importance in alignment and scaling. So there's certain things you have to do. You have to match many things. It's very sensitive to differences, but it's basically one of the good features is that it really is robust and I'll point that out in another example to random noise and the resolution of the input. So you need good measurement technique, but you can tolerate pretty high, pretty low signal to noise, which again in the NMR world where the signal to noise is always an issue, particularly when you're collecting spectra natural abundance or you wanna do spectra measurements quickly, that's really important. And I'll point out an example that Frank Delaglia who's really leading all the spectral processing in our group has developed where we can actually even use PCA in a quantitative mode to sort of start to address this question, can we quantify a structural difference between two different proteins? So the second example I wanted to just point out quickly is looking at oxidation. And this is where the Sega Solomon who was a postdoc in the group who recently left us for GSK. She led a study where we looked at tracking together with Robinson who was our mentor, the oxidation of the NISMAP over a time course. And so what's shown here on this first side is it's very easy, for example, to see when you oxidize methionine as shown in the box, you get the methyl sulfoxide and chemistry is very easy to pick up. Oftentimes in NMR, you see the shit and the peaks as the chemistry or the modifications of the methionines happen, they shift to a new position in the spectra. But really what was most interesting to us is to look at the impact of that oxidation over time on other parts of the NMR spectrum, the structural consequences besides the chemical change to the particular amino acid. And so again, using PCA analysis, different time points were able to be multiple acquisitions were taken at each time point, these could be clustered and then plotted to look at the sort of variance in the NISMAP as a function of time, how the structure was changing and plotted according to the PCA, the PCA scored particularly the PCA2 score. And again, at each time point, one could use loading pots to look where to see where the variance is coming from. And again, I didn't point this out on the glycan spectra, but things like removal of the glycans or oxidation of the methionines, these signals are moving, these are sort of easily coming to pop up and seeing contributing to the differences, but there are also these other positions or signals in the NISMAP they're changing which are indirectly more related to structure and not directly to that chemical modification. And so the interesting thing that Sega did is we all forget this question of correlating a chemical change or a primary amino acid change with a structural change in that implication that structural change for stability of the molecule or for function of the molecule. So at every time point, Sega also collected binding data looking at binding of the NISMAP to protein A and at protein, these are small proteins that bind to all monoclonal antibodies and then looked at the intrinsic triptopan fluorescent melting profile and looked at how the stability of the protein was changing. These are all fairly well understood that as you oxidize the MAV, you would basically decrease its stability and also change its ability to function to bind. Normally it's protein binding partners. What I found really quite interesting is when you look at the NMR, so here's what we've shown here is if you look at the things like affinity for binding affinity for these proteins or peptides, the protein A, and you plot them as a function of change in the PCA2 score, there's a fairly good correlation between the structural changes that are detected by the NMR as read out in a PCA score and the change in basically ability to bind at high affinity, these binding partners. And the same thing is true with stability. If we look at the changes in the stability of the NISMAP as a function of changes in the higher order structure as read out by the NMR spectra process through the PCA, again, and plot the PCA difference as a function as a function of stability change, again, you see a fairly good linear correlation. So as one might expect the, and this sort of data has sort of borne out when you have a chemical change like oxidation that manifests in structural changes to the protein, the structural changes should correlate often with diminished, could correlate with diminished stability and also diminished binding affinity because the structures are changing its oxidation, in this case, the stabilizing structure and also changing its binding interaction. So it really put the NMR in between the, to show where, how these sort of structural measurements could be directly correlated. And that question comes up quite often. How do these structural changes we see by NMR correlate with function, efficacy and stability? And so that this study, which just was published in the last few weeks, I think, or a month or so, basically is starting, and others are showing the same, other groups are starting to show this, how you can correlate the structural data with function and stability. So again, as I already alluded to, we wanna ask the question, can we look at two spectra and say, are they the same? And go beyond just saying, same different. Can we say something about how similar something is structurally? And so to do this, Frank Delagli and the group, basically has shown that in this particular case where you're comparing two sample classes. So if you use the PCA to look at two different molecules, the replicate measurements within the two classes can be used to show a distribution, a numerical distribution around the PCA score. So the PCA can be used to cluster and then by using the PCA, any variance in a repeatable replicate set between two sets of samples, anything beyond the second principle component represents sort of a variation in that class. So what that means is you can take the PCA, which we've already seen many times, it can be used to cluster two different types of protein, to say NISMAB with oxidation or not oxidation. These can be converted from a PCA score into a numerical score. And then in terms of the distance and standard deviations around those scores can be used to create Gaussian surfaces. And so you essentially get a Gaussian surface around each of the PCA clusters a numerical Gaussian distribution. And then by having that sort of conversion from PCA to Z scores and then establishing this Gaussian surface which represents the cluster, one could then look at the distances between these two clusters and begin to get a quantitative measure by the distances between the two clusters of how close or similar they are. So again, you use the structural variation within a class through replicate measurements and then convert that numerically to allow a distance measurement between the two classes. And so one could imagine a scenario where in a pharma company, you have a reference molecule which is well established with its and is used for comparison to new lots or biosimilar. So it's again this two class comparison and you can sort of say how well does each new lot compared to the reference material or how well does the biosimilar compared to the new material quantitatively. Okay, so that's basically showing the same thing I did. You can get the minimal interclass distance to get you sort of begin to understand or to describe the differences in a quantitative way between the two structures. Okay, so getting towards the end of the talk but I wanted to point out that NMR is a multifaceted multimodal tool. There's a lot of ways it can be applied. We've been sort of focusing a lot on two dimensional correlation. There's a lot of information there that can be acquired in a relatively rapid timeframe and it's shown to be very robust for fingerprinting but there's also a role for one dimensional NMR which is not as highly resolved but has the potential to be applied much more rapidly at higher sensitivity particularly if you're doing proton correlated proton spectra. So at the same time that we were really sort of working to sort of advance the two dimensional methods group at Amgen was also pursuing and developing the application of one dimensional proton spectra applied to monoclonal antibodies in a method they call profile. Just to briefly summarize what profile does typically the challenge with a one dimensional NMR spectra as you can see all the way on the left here is there's a lot of signals and antibody again is somewhat broad and it really is difficult to pull out any specific signals from this and with all the overlap it really doesn't have the resolving power of the two dimensional spectra. What the group at Amgen did was use a processing method to basically remove the broad underlying spectra from the one dimensional spectra using a manipulation of subtracting a Gaussian broad and contour spectra that's shown in the middle and then the subtraction of these two spectra generated the so-called fingerprint spectra which represents the fine structure of the one dimensional spectra and the patterns of those were shown to be again correlated with the NMR with the structure. So a postdoc came to our group so we actually decided or worked, contacted the group at Amgen and said we would like collaborate to look at the application of one and two dimensional NMR look for fit for purpose. Again, that's something we always were think about us where should the method be applied and at what level of resolution do you need? Again, using things in time frames that are best suited for the application. And so we wanted to explore the one and two dimensional methods and see where they worked best. And in the course of this collaboration and a joint, this was with Matt Wilström in a postdoc, Wade Elliott who's above Luke in this picture, Wade who came to us to work with us with Amgen and again with Luke Arvigast. It was sort of appreciated that one could actually use again the PCA approach to get an analogous type of way to sort of pull out the fine differences in the spectra without doing it in a sort of an alternative way to the post-processing contour subtraction to generate the fingerprint as was done with profile. And this so-called PCA approach to one dimensional map spectra was dubbed profound. So protein fingerprinting using the policy composition but essentially if you look at it analogously the first principle component again the average of the spectra looks very much like the monochrome antibody broadened spectra here in B and then the second principle component actually represents the differences, the finer structure which is again, it's a different analysis but somewhat analogous to the fingerprint. And basically the value of using the profound method is that there's no subtraction. So it's an easier approach for analysis and it's much more as I'll show you robust in terms of signal to noise that can be handled. Whoops, excuse me. So by using again the one dimensional proton spectra and applying the PCA component they can Wade and Luke were able to show that you could apply this sort of analysis to one dimensional spectra down to very low signals to noise which again is the value of the one dimensional spectra. The profile spectra oftentimes would take many, many, many scans which could take hours to collect. So to get the signal to noise one would need to do the post acquisition processing here the using the PCA approach to the direct one dimensional pulse field grade field gradient stimulated one d spectra we could directly pull the data into PCA low signal to noise was tolerated again. And again, using the profound results we have reference data on the left which shows the thousand scans which is something that would be typically taken for profile but we could get the same robust clustering of the different antibodies using micro probes smaller number of scans and even lower field NMR. So one could imagine taking this pushing this all the way down to bench top NMRs with very low field low resolution spectrum that should potentially still have the discerning power to distinguish between these different antibodies. So the one d again in this context using the PCA gives you all of the advantages of the one dimensional spectra it's highly sensitive it's rapid it can give you these yes no answers is it the same as it clustering? Certainly one could then move to two dimensional spectra if one is interested in the more details of the structure of the molecule. And there were also it turned out due to the nature of the methods there was some degree of complementarity between the different aspects of structural variants that one could pick up with one d versus two d chemical changes versus things like the diffusion changes. So everything I've spoke to you so far about is using the NMR finger printing or NMR is a finger printing tool and really being agnostic to the what the where the signals actually come from we're just looking for patterns and matching patterns and if patterns match or don't match you can say to what degree the structures are the same or different or highly similar or not and that serves a very useful function but as NMR spectroscopists we know that where we always want to be able to put labels on those peaks and so if you could put labels on those peaks not only could you do the patterning sort of recognition and comparability but you could also start to think about well if there is a structural change and this is another question that comes up often is well what's the action if there's a change in your NMR spectra? What's the functional of course in stability outcomes how does it correlate? And then could you basically if you could assign that structural change to this particular amino acid or structural element in the protein that would be highly valuable. So coming back to what I pointed out in an earlier slide in the talk the MAB molecules essentially three domains three domains the FC and two FABs connected by a flexible linker. So the two fabs are in blue, the flexible linker in green and the FC is in orange here and if you look maybe hard to pick this up over the Zoom call but essentially the FC FAB spectra are fairly closely resembling the summation of those two spectra in the intact antibody and that's something that one might expect if you have flexible domains that are really interacting with each other. So the value of this observation is one could potentially express these fragment molecules assign them and then transpose those assignments onto the full intact MAB and the reason that's important and I'm not gonna go into any of that detail but we've spent years in the group trying to establish platforms for the expression isotope and Richmond of monoclonal antibodies and their fragments and just to say monoclonal antibodies are produced and typically in show cells or mammalian systems and for those who are not familiar with mammalian cell culture growing things particularly on DTO is really not possible and deuteration is really critical to get the assignments at this size of a protein. So without going through all of the attempts to do things with E. coli I would just jump right into where things have worked recently and this is again the work of Kinlan in our group and Brad who are on the left who are these protein expression experts and then work by Sega who I already mentioned who work on this oxidation work Ryan Robinson, Kinlan and Brad were able to take a expression system that was initially developed in the group in Health Canada and optimized which was used early on in that inter lab study I mentioned as a suitability sample and they were able to optimize the protocols to allow triple labeling of deuterium, carbon and nitrogen with the ability to triply label the protein dealing with some other tricks about reprotonating the amides which had to be done because it turns out the back exchange didn't happen completely in such a large stable fragment as the fab they were able as I can show you here is to assign these labels which again represent the backbone amide assignments so for those who are not familiar with NMR assignment with the ability to label the protein one can actually jump from Adam to Adam from proton to carbon to nitrogen and by creating these patterns of relays of correlations one can line them up and get essentially sequential assignment through the whole protein backbone and by doing this they've reached a confident assignment of almost 95% of the molecule and now this opens up the possibility to overlay these assignments on the for example in this case we focus first on the fab on to the intact antibody and look for spectral responses of the antibody and assign them to specific structural regions and then the idea here from a pharmaceutical perspective is again as a platform molecule we would hope that going through the assignment exercise once should be sufficient to transpose these assignments onto the the platform molecule and represent since these platform maps have very few amino acids that are changed as they are retargeted to different indications a lot of the initial work in development of an antibody to do these assignments first of all for a particular antibody once the assignments are done they do not have to be typically repeated and so you basically have this information throughout the life cycle of the drug and potentially transported to different drug development campaigns again within a company using highly sequence similar amino acid antibodies so with that I'll just summarize what I said today so hopefully I've given you a pretty broad overview of how one and two dimensional NMR methods can be used to fingerprint and fingerprint maps we just point out that when we started this work there wasn't a lot of confidence there's still some skepticism particularly in the industry about how well NMR would work for such large proteins as monoclonal antibodies and how it could be woven into their development quality control manufacturing of these types of pharmaceuticals we've come a long way to show how these methods are robust in general the two one and two D methods as I said through our work with Amgen are quite complimentary they are somewhat sensitive to different aspects of the molecule and again the power of NMR as you can look at mixtures of samples and using selective pulsing we can actually filter suppressed signals we want or pull out signals we don't want and again this allows us all of these approaches are driven towards the idea of applying these methods to the pharmaceutical that's coming out of the syringe or the vial without making any changes to the formulation of that drug or requiring any tagging or labeling and I sort of gave some overview of the spectral processing that's involved here in the idea of basically taking the method out of the hands of the experts to create sort of abilities for non-experts to use it to answer simple yes-no questions does it inspect or out of spec how to actually start to think about quantifying structural differences which as I said is not typically how we think and again the last part which we've been very excited about in the last year or so to really put assignments on the spectra to start to ask even more detailed questions to try to help again guide understanding structural changes which can help improve and develop new antibodies and I've sort of alluded to folks as I've gone through the talk but this again is a summary to thank everybody who's been involved in the group Rob and Frank in particular I think they're on the call today Luke who was with us for many years who recently moved to Eli Lilly a number of very talented postdocs all of whom have left us for industry in the last year or so and again our long-standing collaborations with Eva Vaughn and his group in Health Canada and I didn't talk about it but we really have started in the last year or so to get more involved with Bruce Yu who is a University of Maryland faculty member here at IVBR and in the application of Benchtop NMR and as you can see from this picture here at IVBR and this we want to go from 900 megahertz all the way down to a Benchtop system to show NMR as sort of applications in bit per purpose in different aspects to be practical. Alright, so that's what I have for you today I would be happy to take questions kind of went right up to the last minute