 Okay, so we're going to spend intense three hours to close the doors. And then we'll have another hour and a half tomorrow. And also, I guess maybe I'll actually start by apologizing in a way. So I just arrived yesterday, I spent today working on my talks. So I have not actually seen any of the talks yet. And that means that I don't know what you've been taught already. And I'm not exactly sure what would interest all of you. So hopefully these lectures will be roughly at the right level. What I would encourage you to do is interrupt me. The other thing I should say is that I haven't given this precise talk before. And I've never spoken about this for three hours. So I don't, and I'm very bad at estimating time. So we'll just see how this goes. Okay, so it'll be a discussion, please interrupt as much as you like. The general themes, this is not the plan. It's just a sort of general takeaway is I'm going to talk a bit about stem cell biology. And then I think that the general overarching theme is how I learned to love big data. And really why I think that for many, when I did my PhD in a physics department, there was this idea that bioinformatics was something best left to bioinformatician and biophysicists should really focus on models and theory. And I hope that you'll come away, those of you who have a more physics background, which seems to be more of the theme of these talks, hope that you'll come away feeling that there are some really, really interesting challenges for people with a physics training to contribute to big data analysis. Okay, now more sort of concrete, this is roughly the themes. And titles of the lectures are not quite reflecting this theme. So this is sort of what's come out after trying to write the lectures. So I'm going to just briefly think a bit about models of self-made choice. I actually called it differentiation here, but it's really about self-made choice. And then we're going to zoom out from a molecular level to a cellular level and think about how does the discussion that occurs at the molecular level and the discussion that occurred at the cellular level, they're slightly different. So it's almost like a sociological discussion of biology here with a bit of science in it. And then I'm going to hopefully use all of that to motivate why it really is interesting to look at single cells using unbiased profiling methods. And we're going to talk a bit about how single cell analysis can be used to ask questions about self-made choice and about population structure. And we're going to look a bit under the hood. So thinking about what it actually means to engage with data of this type. That's in the spirit of this being a school. So we should sort of just look and get a feel for the nitty-gritty of what working with single cell big data sets involves. And we're going to look at some pretty pictures. And that's maybe where we'll finish today. We'll see how the day goes. And then next, tomorrow, what I would like to do is actually probably mostly use the white board and talk a bit about some ways that we've been using ideas from spectral graph theory and statistical physics in order to make predictions on cell fate choice based on single cell data. And it's somewhat more formal. And there'll be a sort of theory component on the board. And then I will hopefully show you how we've actually put that to work in hematopoiesis to discover new receptors that control hematopoiesis and map out the hierarchy. So some real biology coming out at the end. So just to summarize, this is pretty much what we'll do today. It'll probably all be on the screen. And then tomorrow we'll spend an hour and a half on this last piece of work. So here's a starting point. And this is the first place where I'll ask for feedback. How much have you talked about bistable switches in this workshop already? Okay. Yeah. Okay. It's been touched a couple of them. Okay. Fine. So I'm not going to go through the maths of this. Let me ask you now a different question. How many of you have encountered a bistable switch in your work so far? Or read about it or being taught about it at? Okay. Yeah. A lot of you. Okay. Good. Really the reason I put this up here is that it's one of the most conceptually simple and satisfying pieces of analysis that you can do in order to feel like you're learning something about regulation. And I'm going to challenge this very soon, but it's a good starting point. So the general idea is if we're thinking about how cells make decisions, it's between making, becoming two different cell types. An appealing idea is that we might have master regulators, genes which are going to then orchestrate the expression of many other genes. Here they are represented symbolically as X1 and X2. And if these two genes repress each other, and there's some nonlinearity in the system, in this case it's represented by these autocatalytic loops, then we can generate a dynamical system which has two stable fixed points in which either one of the genes is on or the other gene is on. And this, it's pretty simple to generate pictures such as the one that we see here. So what we are looking at in this is a phase diagram where the fixed point, so if we write down a dynamical system in which we have say dx1 by dt being synthesized at a rate which depends on the concentration of dx1, and it's maybe being degraded by a rate which is dependent on X2, but of course that's the life, that's the rate of degradation, so we have to have a rate, a rate, it has to be proportional to the concentration here. Or alternatively, and there's many different formulations, we can have a synthesis rate which depends on the concentration of both components, and then we just have a uniform degradation rate, this is equivalent. We can write down, we can write down a dynamical system, and then do the same thing for, so if we use this as X1 dot, and here X2 dot, so this would be S1 gamma1 and S2 X1 X2, and we can just absorb one timescale into this to X2. What we can then do is we can for every point plot of vector showing the response, and if we have mutual repression we can end up with two stable points, and depending on the non-minerality of the system there might be a weakly stable point which is intermediate, so that if you were to take a cut through this diagram, you could imagine representing the dynamics by a potential landscape where you have these inflection points, which are the points of its stability, and then stable basins of attraction. So this is very appealing, and there are a number of examples that have been quoted for this, so here we just have some real examples, so if we're looking at say hematopoiesis, the common myelodendrocell can give rise to either megacarocytes or erythroid cells where the master regulator's got a one, or it can give rise to granulocytes and monocytes, and the master regulator's P1, and it's been suggested that there's a bistable loop there, sorry, a bistable system here with negative inhibition of GATA1 and P1, and likewise for embryonic stem cells the idea being that a totipotent cell can either become trifectoderm or in a cell mass, and there might be master regulators which inhibit each other there, and so on, this is in stromal cells or mesenchymal stem cells which can give rise to osteoblasts or adipocytes, again we may have two factors. Now this is, there was a sort of an explosion of excitement that we now have a framework to describe cell fate choice using a set of these switches, and it also seems to be very appealing, just go back one slide, that we can represent this using the idea of a landscape, because the landscape seems to be very, very visually, this idea that there's a potential which defines, which would stable attractor state seems to resonate very strongly with the way that we think about stable states in physics and so on. Yes, right, right, so this would be that maybe a minimal example, and then you could imagine a situation where you may have, this could be a gene X1 giving rise to X2, and then you have a second Y1 which then drives expression of Y2, and then you could imagine that this cascade goes on for a while, and these actually never inhibit each other until a certain point where you suddenly get mutual repression, right, and back again, so these are elaborations, and of course you could have interactions, so this would be a situation where you may have a developmental process where two different lineages, the gene expression program of the two different lineages can occur simultaneously and only at a certain point do they start to repress each other. Do you want to suggest another alternative, right, so you can now you can ask how do you make this combinatorially complex, how do you say make it four states, right, yeah, so if we're going to use the same type of ideas here, then you could either make it hierarchical, in other words, you make one subset of decisions and then another subset of decisions, right, a set of binary, or you could start to have multiple interactions where there's a set of binding sites where each factor is auto-catalytic but inhibits all of the other factors, so we can now essentially extend this out, and those are the least imaginative ways of extending this out, and I, just off hand, there's probably examples that people have published of more sophisticated schemes. I think that one place where this is going to clearly break down, so the olfactory system is an interesting one in which every cell there's several hundred olfactory receptors and every cell will choose to express only one of them. It's very hard to imagine that this type of scheme is acting when you have to choose between 300 different receptors, so clearly there are going to be some cases where this is almost certainly going to break down. I think I'm going to suggest in a minute that we don't have any cases where we actually know that this even works, so that's going to, okay, I just set things up a bit more. This idea of a landscape which comes up, so one of the reasons it resonated a lot is because it's been, it's an idea that's been around in biology for a while now and it really is electrified biology in many ways. This is, so many of you may have already seen this picture being shown in talks. This is, or maybe you've looked at the original, so this is, this is Waddington's work from the 1940s originally, where he was thinking about the dynamical structure of developing embryo and this picture here is essentially a very evocative of the idea that development involves cells starting off in a wide basin and then as time progresses they're rolling down potential landscape and they encounter bifurcations and then the cells will commit to one fate or another as they roll down this landscape and the idea is that time is measured in this case sort of in on the depth axis or the vertical axis and that some sort of measure of phenotypic space, maybe it's gene expression profile is being reduced in this schematic just to a one-dimensional representation just to give us a sense of what's going on and there were a few important ideas that Waddington raised when he, when he discusses. One of them was the idea that, that was observation that cells, that an embryo starts off as a single cell which is undifferentiated and then specializes and that this, this, this may be, this, this, this is going to, this was already known to be a branching process, but more than that was the idea of canalization, the idea that initially the basin start off as fairly wide and that as differentiation proceeds the valleys become deeper and the cells become locked into their state more. Now all of this is very metaphorical and it's very inspiring to, to essentially, it continues to be inspiring to, to generations of people studying the volumetal biology. Just for fun, so here's actually a sort of a picture from the original with really what the, the very figure beforehand was showing and here this is the, the figure, you can see why the figure beforehand is less famous. You sort of, you stare at the landscape, you sort of instantly have an idea of what's going on. The figure beforehand is not so clear. So the figure beforehand is suggesting that now phenotypic space is being represented as a two dimensional plane and here we are in the undifferentiated egg. And as we proceed to an adult, the idea is that cells are coalescing into gradually narrower basins of attraction and that these attractors are these deeper and deeper valleys that that are emerging later on. So this is, this is sort of, this is an idea and you can imagine writing down a dynamical system with a series of attractors, which would satisfy some of these properties. Sorry. So the idea over here is that the attractors would somehow be a committed cell fate. Now the idea of potency is really a functional one is whether you can reverse your trajectory and whether or not this directly relates to the idea of potency is, maybe it's, it's not this is, this is, this is describing the dynamics rather than commenting on their reversibility. Yeah, but, but presumably from a developmental perspective, at this point, you're a totipotent, at this point, you're a multipotent, and at this point, you're a unipotent with respect to these decisions. Yeah, that's, yeah, that's to keep it simple. Great. Okay, so fine. So attractors, they're very appealing. And this is a picture from a review by Sui Huan, or an opinion piece by Sui Huan from 2009, which was really drawing, just again, reminding us of the relationship between these bistable states and these attractors. Now this is a sort of a stylized attractor basin. But the idea is that cells, remember, if you recall, we have an uncommitted metastable state where cells can now fall into committed states. And if this is now a hierarchical process, then once you drop down into the next state, you can make this decision again and again and again. And gradually, you know, become from a stem cell to a progenitor cell to a differentiated cell. So this is a picture. And it sort of raises a number of appealing ideas. The idea being that underlying this picture is a well defined dynamical network, which has noise on it. So they're fluctuations. And that provided that the fluctuations are small enough we're confined to a local minimum. And that with a signaling might remove one of these humps or push the cells over the boundary and cause them to differentiate. The idea being that the uncommitted state is somehow a ground state. And that cells which happen to be fluctuating more in one to one direction or another are biased in their fate. These are hypotheses that have been suggested, really inspired by this picture. And the further important point is that the fluctuations occurring in this basin are reflecting the natural dynamics of the system towards differentiation. So this raises another important concept, which was raised here, which is the idea of priming. The idea that already information on the dynamics that are going to follow are encoded in the fluctuations of the early state. So I'm not going to say a lot without writing down pretty much a single equation here. And actually, I don't want to write down equations because we could write down some formal equations which show all of these ideas. But they're just ideas. These are just hypotheses. And this is almost an idea you could have over a beer. I mean, maybe it's right. Maybe it's right. It's very, very appealing. Okay, so I've sort of already hinted to you that maybe it's, you know, this is just not all as sort of happy and potential land. And I just want to suggest why some people think this is controversial. So the first point, and maybe as a physicist and as physicists, you would feel this as well, is that the landscape, especially Waddington when he wrote this down, really meant that there's a metaphor. And it was very much inspired by the idea that potentials have played such an important role in physics from the 19th century onwards. But potential fields really in physics are never metaphors. So the idea, if it's going to inspire us to do some biology, that's very good. But we should not be necessarily trying to force our biology to look like physics. We should be thinking about how we can understand biology in its own, on its own ground. And really the use of a potential field, I mean, the reason the potential fields emerge in physics is that they're really, first of all, tools for solving problems. So potential fields emerge as it were spontaneously when we're trying to write down the properties of physical systems. For example, conservation of energy imposes the idea of a potential on us. So they impose strong constraints, they reflect constraints. Rather than imposing them, they really reflect them. They're analytical tools. They, even in places where we're emerging to the unknown in physics, we're essentially proposing a very specific hypothesis. And potentials are not consistent with certain properties, and they are consistent with others. So by saying that we think that a potential underlies a dynamical system, we are making very strong statements about the properties of that dynamical system. So I'm sort of being very technical here later on, and in fact, probably tomorrow, I will come back to the idea of a potential and think about it much more formally. So this is just sort of for fun. So this is from our recent review by Al Fonso Martinez Arias at Cambridge, who was really attacking this idea of Waddington's landscape. And he was saying, well, it's a good metaphor, but it's difficult to implement it formally. How do you actually use it? Then this sort of shows the problem with using ideas which are just metaphors, is that now Al Fonso is to some extent thinking about not about a biological potential, but a physical or chemical potential. And he's really claiming that well, cells are a system far from chemical equilibrium with no conservation of energy. Well, in physics, certainly potentials are associated with conservation of energy, but there's many places we could use potentials which don't require conservation of energy. Of course, without a definition, these types of arguments make sense. And then finally, this idea that as a picture, it's great, but how do you actually implement a potential in more than two dimensions to get some insight into it? So those are all those thoughts, and we'll come back to them tomorrow. This is the idea, so just to, hopefully, you've still been inspired by this picture, that there's a lot of interesting phenomena that could be thought of in this way. We'll come back to it. Okay, fine. So that was just a bit of a taste on the ways, this is a real, real touch on thinking about molecular regulation which could be used to create a switch. Okay, if we want, that was very, very, very light. If we want to really go and sort of work through this carefully, we can do that later if you like. So what I want to now do is actually zoom out and think about, well, where are these switches being invoked? They're being invoked to think about self-made choice. We could be thinking about, say, bacterial fate choices under stress. We could be thinking about yeast. I'm going to really focus more on mammalian systems. And in fact, you know, if you go to a workshop on stem cell biology, half of the workshop is spent thinking about what does it mean to be in a workshop on stem cell biology, because it really covers many, many fields. And there's no way I'm going to do justice to even a tiny bit of this. I just to mention some of the fields which are broadly covered by stem cell biology here. So this covers some of the earliest developmental choices and maybe some of the later choices in development. In adult tissues, and I'll spend a bit more time on this. That's the area I know a bit more about. Stem cells are defined as the cells which can regenerate tissue or renew tissue during adult life. Regeneration is really often considered its own separate field of study. For example, looking at axolotl where an entire limb can be removed and that limb can regrow. Then the question is which are the cells driving that regeneration? Stem cells are also using an entirely different context which is really to describe the properties of a culture system where cells taken from embryos or taken from adult and reprogrammed by induced pluripotency are then are then kept in an undifferentiated state through perturbation of their signaling environment and then only allowed to differentiate later. That's a really very quite a different state from any of the states that we see here. There's a very strong similarity between some of these cells and some of the very earliest developmental states but it's important to emphasize that the embryonic stem cells would only, the thing closest to an embryonic stem cell would only exist in an embryo for a period of hours or at most days whereas in a dish they can be held for a very long time and therefore the types of biology you could get in a dish could be quite different from what you see in an early embryo. So really if you're thinking about, so this is really maybe not a stem cell system at all but it is a system where fate choices and differentiation occurs. If we look at immune cells which are responding to an infection such as monocytes entering a tissue they will differentiate into macrophages and there's different choices for types of macrophages that they can become. They can also differentiate into dendritic cells so there's choices here and then of course is dysregulated fate choices so all of the above in disease and tumors really can define entirely new modes of decision-making. Yeah, yeah what's the other stuff? So actually I should actually say it's a bit unfair immune cells are really not considered stem cells they really are a place so cell decision-making is really quite ubiquitous and I just want to put that out there. What isn't considered a stem cell? Yeah that's a very good so actually you know what makes it even worse is that the very cells which are called stem cells can be studied by other people and just in a different field. So for example epithelia or self-renew and therefore there's a whole field of studying stem cell self-renewal and epithelia but of course epithelial biology is a very mature field of cell biology and there's a huge overlap between those two but there's the different flavor so it's really defining communities rather than necessarily a particular system and the most important maybe the most important point over here is that stem cell is really a functional definition here and they're defined functionally in each one of these cases in a different way an adult tissue stem cell has nothing to do with an embryonic stem cell they just happen to both be called stem cells. So often you know we were just talking now about this idea that there might be biophysical mechanisms for making switches using dynamical systems and an idea of mutually inhibitory molecules but really when you look at what what's really occupying the field of stem cell biology it's often not working out the molecular detail yet everybody would love to do that but real tissues are very complex environments and often the first challenge is just to define what cells we're looking at what cells what states of cells exist so here's a here's an example this is still very simplified very abstracted of an adult somatic stem cell the idea being that there's an undifferentiated cell it could self renew into so it could divide into two stem cells it could differentiate and then it could give rise to several multiple different cell types here it could also sit there and be quiescent for a long time now just just drawing this diagram and relating it to a tissue is not a trivial exercise there's a number of questions which need to be answered so what can we identify each of these cell populations can we find markers which we can use to pull out these cells and then functionally characterize them our is is for example this base of the hierarchy the stem cell is it one population or is it a collection of interacting states is it an ecosystem how many differentiated cell types are there have we accounted for all of the different cell types in a tissue and that may be important because they may be signaling back to the stem cells so even if we're not interested in their biology per se we may need to take them into account in order to fully define the environment in which cells are making decisions let's say that we know every single cell type do we know the structure of this hierarchy and that's very important because if we're going to study cell fate choices we need to know which choices are being made is it one cell which then can directly go from an undifferentiated state into five different cell types or is it a series of hierarchical decisions in which you gradually narrow your state and eventually commit to one lineage and so on so there's all of these questions are really and this is sort of important to say it's fun to sort of dive right back right into the detail but really in many cases the field is still debating these questions and you know here we are in the 21st century and the quantitative systems biologists are coming in and they think great we're just gonna model this we don't even often know what we're supposed to be modeling because we don't know where these decisions are being made and so on and then finally you know the point where ultimately the biology tries to focus in on is which are the pathways and genes that are implicated in these fate choices and and maybe not surprisingly much more progress can be made in knowing which molecular components are are implicated in a process then understanding how the whole system is put together and that's really ultimately where quantitative systems biology can contribute but we do need to have this big picture in order to put the pieces in place so okay so how do we go about doing this so and again this is sort of in the spirit of pet pedagogy here so apologies no it's a bit of a dry slide so you know if we were if I wouldn't show the slide and would brainstorm this I guess I could have asked you this is maybe what we would come up with well first of all we'd like to have a catalog of cell types historically this is being based on morphology or histochemistry gene expression and so on many of the names of cells reflect histological characteristics the difference between a neutrophil and an eosinophil is that an eosinophil staying strongly in ear with ear sin and a neutrophil doesn't because the difference in pH okay so that's a neutrophil so many of the names sort of reflect some of the very basic ways that we've defined cells well that might give us a catalog of cells but now we want to know their relationships so we might label cells and trace them in development these are some of the oldest experiments just injecting dye into an embryo and seeing where that cell that you labeled ends up and then making fate maps in adults the same thing can be done I'll show you a bit about how labels are labeling can be done later and of course this can be done in different contexts and each context might define a different functional stem cell so if we define the stem cells the cell giving rise to to differentiated cells and self renewing we may get a different answer in each of these cases okay what else can we say well we want to really not just observe these cells but we want to really see what they're capable of doing so another way that has been very powerful has been to take cells out isolate subsets of cells and then transplant them into another organism and see whether they're capable of engrafting and regenerating tissue okay so and maybe they're biased in the types of cells that they make so we can now find subsets of cells with particular fate biases we could you know if we if we're with with some effort and this is when I say we I don't mean me I mean the community can define culture conditions in which we can essentially grow the cells and culture and follow what they do and of course the main caveat with culture is that we might perturb the cells and doing so so we have to be careful there and then finally we could kill the cells and see what happens to the tissue so we could use for example diphtheria toxin linked to a gene to specifically remove one type of cell and then ask what how does the tissue care who compensates does the tissue fail and what way does it fail all of these are different ways that we can try to work out a structure such as the one that we see here in a very real context okay okay so that's that's hopefully I give the sense for the types of questions that stem cell biologists are asking so now let's just sort of way back and see how this looked like 50 years ago so this is really pioneering work this paper is from 1974 from Charles Philippe Leblond but Leblond was really pioneered this work in the 50s already and what he did and this sounds like something totally mundane today is he fed mice tridiated thymidine which is the equivalent today of using edu or brdu for those of you who use it and then by doing radiography he could trace which cells are dividing and where does the label end up and this was a way of looking at adult tissues and he immediately realized that there were three types of tissues and it was sort of it's amazing until the 50s and 60s there wasn't a clear appreciation that there's some tissues that undergo regular turnover there's some tissues that only undergo turnover during injury and regeneration such as the liver say and then there are certain tissues which don't seem to undergo any regeneration at all and in those tissues Leblond included the brain of course we know now that there are neural progenitor cells capable of limited regeneration but he didn't pick up on that at the time so among the this is now that's just seeing where we that's broadly classifying tissues now this is for example looking at the intestinal epithelium and looking at where does the tridiated thymidine first end up and then over time where does it end up localizing so the first thing that Leblond realizes that there is a the epithelium is this this is known already looking at intestinal epithelium it so this is drawn not to scale actually it's better so this is a villus actually maybe I'll start over here so if we take a cross section through the intestine the inner surface these undulations and these are villi so these are this is to increase the surface area of the intestine and the role of intestine is to absorb so we have absorptive cells on the surface and this forms a we want that a layer to be fairly thin so we've just got a mono layer and this mono layer is exposed constantly to to toxins and to to bacteria it's a very non-stellar environment so for you know I could tell you a just so story but the observation is that these cells only live for about three to five days and then after five days or so they die and they die at the tips of these villi so if I now here I have a villi of villus over here we have apoptosis or see the cell death over here okay so that and what the blonde found is that the cell for neural occurs the cell division is localized to these glands which sit at the base of the structure so these are some mono layer and there's cell division occurring all the way up here and if you now pulse these cells and trace them you can see that over time there's a steady flow of cells from the base of the crypt up so this now raised the idea that the base of these crypts is what we might call a stem cell zone and and that there's a transit amplifying zone here which our cells which are proliferating but they're going to get washed out so they're transient and that finally we have post mitotic cells which then die and this this picture establishes a hierarchy of cell states and if we look at the post mitotic cells the blonde identified that there are four types there's a panic cell is an enteroendocrine cell which is very rare a goblet cell which is secreting mucus I'll talk about the panacea in a minute and a columnar cell which is the absorptive cells so most cells are absorptive most cells are these columnar cells these goblet cells are still very prevalent they're about one to one-third to one-quarter of the cells and they're coding the intestine with it with a mucus to protect the intestinal lining and then we have these very rare cells which are secreting hormones your entire endocrine cells and finally these panacea cells and these panacea cells are interesting because they they emerge from the stem cells and then they migrate back down again and they sit at the base of the crypt and they're full of granules of lysozyme so they look like they play an immune role but we now know that they also play a very important role in supporting the stem cells so this is this is an example of how 40 years ago just by using a simple labeling experiment we could define the structure of the intestinal patulin okay so things have moved on you know feed forward and here we are 40 years later and in the 90s some a million biologists copied from Jusofla biologists and realized that they could make genetic constructs which would allow them to specifically label cells in a particular state with a fluorescent marker or or any other marker actually in this case is showing lack z so it would be it would be a histochemical marker and that the principle is is simple so we have apologized for the low resolution here so we have a gene locus of interest this would be a gene that marks a specific subpopulation it could be say keratin 14 or LGR5 some gene that I care about and now a downstream of that gene we have a cray recombinase which is fused to an estrogen receptor fragment which localizes the cray to the surface of the membrane of the cell extra cell of the membrane and now with the addition of tamoxifen which inhibits the estrogen receptor the cray recombinase is detached from the membrane and it can undergo nuclear localization at which point it finds two target sides which are these two lock speed sites and that allows it to permanently genetically edit the cell in which it's encoded and the genetic editing that is that we ask the crate to perform is to remove a short stop cassette which stops a reporter from being expressed so once the stop cassette has been removed we now have an active site which has a promoter driving expression of a reporter in this case it's beta galactosidase it could be a GFP or any or anything else and this promoter is put in a ubiquitous site so that once it's switched on it just goes on very loudly and we can then follow these cells okay so this is essentially the principle and now this if you think about the trial LeBlanc experiments LeBlanc was labeling the cells conditional on their dividing and asking a very important question conditional on the fact that you're dividing what happens to you next we can now ask that question generally conditional on you expressing a particular gene what happens to you next and this way formally test how the state of a cell will relate to its future fate okay yes that's a great question okay so the really good so if you put a reporter on a gene locus and the gene comes on you'll have the GFP on if the if that gene comes on transiently for example that gene might be a stem cell gene but the cell differentiates and it permanently shuts down that gene maybe that GFP will still be around for another day but it probably won't be around for another week and the time scale over which these questions are asked whether it's developmental biology which is slow or tissue adult tissue homeostasis which is even slower is far too long for us to really link the previous expression to future behavior using a transiently express gene so the beauty of this system is it allows you to say given the fact that at some point in the past a cell was in a particular state what is it now doing in the future and that that's really the that's why this is such a powerful system yeah so it's case in point you can find a lot of genes which are incredibly specific to this crit crit based cell which will then turn off as the cells differentiate so you could only label these cells at the bottom and then ask what happens to the next if you find a gene which is expressed everywhere you'll just label everything and you won't learn very much okay so that's good things are going to even more sophisticated this is a 21st century so in a few years ago the idea people thought well why use one color if you can use many and now there is a number of different versions of this it was initially called a brain bow because it was initially applied to new neural tissue by Jeff Lichman Joshua Sains in Harvard but it's there's now many many other examples of this zebra bow and zebrafish and there's probably other ones a fly bow and and the the the general principle is exactly the same but now we have multiple locks p sites and the locks p sites can undergo more than one recombination event which gives us random combinatoric outcome so this is a very simple one this is a four-color one it's known as a confetti system this was developed in Hans Cleaver's lab and the idea that a confetti mouse is that you end up with four colors and you can see these are actually the intestinal crypts so these stripes going up the side are really just clones which are what you're looking at in that picture is a clone going up the side of a crypt like that okay I've done that in red there's a red stripe there and yellow stripes and so on and this is this is I believe in in neuronal tissue and I'm not sure which tissue this one is but you can see this very striking combinatorial use of colors which now allows us to follow many clones very densely packed together so the disadvantage earlier was that either you just follow every cell labeled or you have to label at very very low densities in order to follow what a single cell does and this multicolor labeling allows you to increase the density of cells being tracked okay so that's one way that we can look at stem cells so let's let's go on to actually a sort of the the classical system for adult tissue stem cell biology and this is the hematopoietic system so the study of hematopoiesis really took off after the Second World War and I have to admit I'm probably not the expert for this but from what I've gathered and if there's someone here who knows more they can catch me this really came from an observation that radiation sickness was had had two components first of all it was causing people to have severe hematopoietic disorders and second of all that there was a serious damage to the bone marrow and this made people start to study the bone marrow as a source of blood okay I may be wrong about the precise order that but the seminal experiment came like this is a seminal experiment by Tillamacolic in the 60s is 1961 where they took they lethally irradiated a mouse and then they took the bone marrow from a donor and injected it into the mouse and they found that at a later point single colonies of cells were appearing both in the bone marrow but it was much easier to see them in the spleen of the mice these large distinct colonies which looked like they were emerging from single cells and when they had a look at these these colonies they realized that these colonies didn't have just one cell type in them but they had all of the different hematopoietic cell types they had erythroid cells and megakaryocytes and myeloid cells and this led to the idea that there is a stem cell which can give rise to many different cell types okay so this was this was really a big idea at the time that these multiples so that there's one cell and they then did something else which I don't have a separate slide for which is they took the bone marrow of this mouse which had been saved by bone marrow and graftment transplanted it into another mouse and discovered that the same thing happened again so that meant that these cells which were giving rise to colonies of differentiated cells could also give rise to themselves so that defined really two major features of a tissue stem cell it can self-renew and it can give rise to all of the different cell types in a tissue okay so we saw that already in the intestine that was actually later work this is really what kicked off the idea of stem cells and it's really also shaped the way that people think about tissue stem cells it's a very hematopoietic view this idea that there is this very complex hierarchy giving rise to many different cell types also and there's a stem cell that sits at the base of this hierarchy so now the the question is becomes pretty well-defined which is what is the structure we know that we know that there exists a single base to the tree and we can now try to map out what that tree is and from the 90s onwards there was a huge boost by the development of flow cytometry and monoclonal antibodies which meant that we can now take a complex mixture and try to fractionate it and look for transient states of differentiation so up until up until the 90s there was observations in colony formiasis you can take single cells and grow them in culture and that was a major really major innovation the idea that we didn't need to put them in a mouse to look at what was going on and to find which factors we have to grow the cells with in order to call them to allow them to grow that gave rise to many of the different factors which were later found to be that real regulators of these cells in vivo so that was a sort of a major major breakthrough but in the 90s onwards it really things really became transformed that here's a sort of cartoon so this is probably sort of too simple but just you know just to spell it out so we now have a complex mixture of cells and we throw in some specific antibodies and we can label ourselves right and then what we can then do is put them in into a flow sorter and we might be able to maybe combine up to these days without too much effort maybe up to eight colors and with a great deal of effort maybe up to 30 or 40 colors using mass spectrometry based readouts and we can now sort the populations according to their antibody profiles and then transplant these cells and ask whether they're biased towards one cell type or another or maybe grow them in vitro so here's an example of an vitro colony forming assay so you look down a microscope and if you've done this for a long time which I certainly haven't you would recognize that as a granulocyte macrophage colony and whereas this would be a an erythroid colony right here so we now can start to say okay well cells with a particular cell surface marker are no longer stem cells because they're already committed to one cell type or another okay so that picture that I showed you initially was a very abstract one here's now an example which is probably thought to be true a few years ago of the state of the art this was a state of the art maybe ten years ago for the structure of the hematopoietic lineage so the idea is that we have a stem cell itself in use and then it starts to undergo differentiation and gradually undergo differentiation until we get these unilineage committed cell types right and you know you can draw a potential landscape under this right if you wanted to right but anyway that's an aside it wouldn't teach you too much so one what are the implicit assumptions in this picture anybody want to say what are we assuming when we draw this picture sorry there's no back arrows okay that's good yeah so so that might be a limitation of the cartoons but you're absolutely right there's no background here so maybe somebody right okay very good so there's no background yeah sorry right so maybe with a forced expression of transcription factor or exposure to signaling conditions which are completely irregular we could cause an aberration of this picture but this might still be this may still reflect what's going on in vivo but you're right right so this is the what what is being hardwired mean in the era of reprogramming you know we don't know but this right so so that so there could be ways of breaking this right so that's yeah okay any other thoughts on yep right right we've given labels to these cells they're very very discreet right there's this notice that a lot of these labels are functional this is a colony forming unit they can give rights a granular site an erythroid cell a macrophage and a monocyte this is a granular site monocyte and so on so these are they're functionally defined but they're still considered to be entities which are very distinct yeah what else right so there could be some crosses okay so that's do you mean that there could be some crosses between these so yeah okay excellent so this is a tree it's a tree a tree is a very strong constraint right it doesn't necessarily have to be a tree and in fact you know this I said this was ten years ago the last five years there being quite a few dashed arrows going across this but why is it a tree it's probably a tree because people were looking for a tree okay because your way the assays are designed is that is sort of is looking for a tree so you get a tree at the end yeah okay good yeah that's that's really those are some of the those are some observations about this and again this is sort of important because a lot of systems biology starts from a picture like this and says now let's figure out what's going on it's really important to figure out to remember we don't really know what's going on okay these are the best guesses that we have what's going on there's a very very hot areas of research it's not a sort of a mature field waiting for for some you know wiring diagrams yeah now this is just a sort of a general I was going back to abstraction here so the experimental assay can also influence the conclusions on hierarchical structure I'm just showing this here for an example a really simple example if we say a label a cell we don't perturb the system and then we ask who are the stem cells you could imagine that in an unperturbed system this is a proliferative cell and it gives rise to cells in other states they may undergo some limited proliferation there might be some reversibility between these states but the net flux is definitely from the yellow cells to the green and blue okay now let's imagine that I irradiate the mouse well which which cells are going to be most susceptible to irradiation you know which cells die when you irradiate yeah the ones that divide okay irradiation really kills off cycling cells that's one of the reasons that radiotherapy is used for cancer right is to kill off the proliferating cells actually tumors are somehow very sensitive to irradiation irrespective so that's a different matter but but that's it so we might we might now put a very strong sort of fitness disadvantage onto the proliferating cells and it could be that as a result of that when we relax the system there's now a response from another cell population which hasn't been hit so hard so we now come to the conclusion that well you know actually you've got it all wrong you guys because really what we're into what we're defining a stem cell as a cell that gives rise to the tissue after injury and this is the cell that wakes up and this precise debate has just happened in the intestinal community trying to figure out who are the stem cells where homeostasis is one population but irradiation is another and that normally this the population that responds to irradiation is actually being swept out by the stem cells but it can it actually generates the stem cells no other way and then finally what are we selecting for for transplantation experiments what do you need to if I'm injecting cells into a mouse what do I need to what am I selecting for any thoughts so yeah survival so for example let's say I'm taking bone marrow and I'm sticking it into the blood of a mouse it has to find its way into the host bone marrow it has to buy it has to find as they engraft that may be strongly dependent on the concentration of integrins on its surface it doesn't mean it can't die in that process it may be strongly dependent on the phosphorylation of particular pro-survival pathways that could be selecting for yet a different cell population right so so these and again these types of debates they go on there's sort of it gets tedious after a while for example the whole debate on cancer stem cells has been infected as it were by this where if you label a cell in a cancer and see which cells grow it's very very different from the cells that survive chemotherapy and it's very very different from the cells that if you transplant them into a host will give rise to a tumor okay those are different cell populations and then each one of them is called the steps okay so sort of a negative talk about all the things that are complicated wrong but it's a sort of making you hopefully all a little less naive about the complexities and the questions to ask if you're ever studying these systems okay so yeah yeah so so long as of course so at one level and I think this is your point this is purely a semantic discussion and another level it's important to just realize that it's a semantic discussion because it's not always some of the you know this gives it's a good way of generating high profile papers is to be confused about your terms and then argue about what system so okay yeah you're right this is semantics at the end of the day all of these things are correct okay but it means that if you're if you're for example switching assays midway through an experimental project in order to test different aspects of your system you may end up with a with very surprising outcomes okay and this is already been mentioned this is the final point which is that we until now this picture has been balls and sticks and these balls and sticks really may be simplified as maybe there's a continuum of states to worry about okay and this is itself is not entirely obvious that's true for example chromatin is presumably not in a continuous state chromatin is either open or closed so so not everything is is continuous there are discrete aspects to cell biology but but by and large we can't assume that it's all discrete okay so so you know so and so now this is sort of which it's half sort of half education and half seminar here I'm sort of going to set the questions for the next stage so do we really know this landscape do we know where all the branch points are is the structure a tree so these are some of the questions that were coming up and this is just planting the seed again this is somewhat for tomorrow where I'm going to be going deeper I'm going to developing some theory and then applying it to him at a poise is it's just a very specific example that we really don't know what's going on so we talked about him at a poise is a bit if we think about erythropoises in particular so we start off with a multi potent progenitor which is in an undifferentiated state and then we know that beyond a certain point onwards there's some very very very unambiguous states we can isolate these cells based on their morphology we can actually isolate we can isolate these cells based on flow cytometry so a cell that expresses CD 71 but no turn 119 has just started your terminal differentiation a cell expressing high turn 119 has is later on in terminal differentiation the the profile of these cells because of their size changes the forward scatter on the fax machine so we can really pull out these cells and study them but if we look earlier on there's this region which you could call a sort of dark matter which is between a multi-point progenitor which we have good markers for and this point there's a sequence of events which we can we can I retrospectively define these cells based on their colony forming potential there's this is more mature this one looks like it might be less mature but we really don't have a handle on how to isolate these cells so again you know there's there these points and I'll come back so I think one of the conclusions I'm going to tell you about tomorrow is that we have now defined all of these states and we can now pull out cells at each one of these states along the way so okay so all of this was at the cell level and now let's go back to the beginning of the picture so what about molecular switches and landscapes and you know these ideas are definitely have been very very appealing to to thinking about cell biology and to stem cell biology and this is just from a review specifically actually taric enver is a hematopoiesis biologist and he's he's one of the people who's really driven this idea of thinking about landscapes and so just a review of his from 2009 and the review is full of pictures of landscapes and bistable switches and so on and so forth so so it looks like it's very appealing but when you look at even the textbook cases so I showed you earlier got a 1p1 being a textbook case you actually this is actually a good again sort of a pedagogy is to ask how do you actually go about testing whether one of these bistable switches exist and this is one so this is a this is a very very nice piece of work he's just been published by Tim Schroeder where Tim isolated cells and put in labels for got a one so he has a got a one gfp and a pu1 yfp and he could resolve these two colors and now he he he can now follow them over multiple generations and ask what are the levels of got a one and pu1 at the point where fake commitment occurs and now maybe it's sort of a good idea to sort of draw what what we might expect from a bistable switch so let's see so let's say this is the point of fake commitment and we're looking and this is the expression of our two components so what is we've got them both at this so you know again we're in our we're in our metastable state here the cell can flow one way or the other so in this metastable state we should expect to see some fluctuations and here's our maybe some fluctuations but really too low to cause an effect and then at some point one of these genes wins out and the other one is quenched that's essentially what we'd expect so there should be a very simple statement which is a decision point should occur at the point where one of these genes will win out Tim made a very simple observation which is if you now follow these cells over time and this is you may have to squint here this dark red region over here are cells that are starting to express gotta one at high levels and you will notice something very clear that large regions of this tree are either red or dark so the decision is being made simultaneously seemingly simultaneously by all of the cells in a lineage well before gotta one levels have become high okay that means in other words of the decision to commit to erythropoiesis happened long before this point here so really what he's suggesting is there's some mysterious point back here where the decision was made and we cannot see it so this doesn't rule out the idea of a by stable switch but it certainly rules out the specific players p1 and gotta one as being driving the switch okay so yeah yes so presumably that there is a switch and but it's probably not occurring with gotta one not not by it's not it's not about gotta one going up to very high levels and repressing p1 as a result and it could be that there's a factor like tall one or a bit BHLH which is it which actually pre-presupposes gotta one or it could be that the switch is not about the concentration of gotta one but about the activity of the very low levels of p1 and gotta one already having an effect and that gotta one rising is a much I mean essentially what is that covering all of the alternatives right yeah so as many ways so in fact you know when when we saw this we were having a discussion about this in the lab and one of my students was saying you know this is sort of depressing this is one hypothesis out of the infinitely many hypotheses for how a by stable switch could be implemented it's specifically that p1 and gotta one interacted and yet it took a huge piece of work to just test and falsify this one hypothesis as many many other ways that this by stable switch can be implemented but you know this clearly rules out the sort of the textbook model for what was going on okay so let's see it's four five past four this is the summary so far this is sort of being a series of vignettes I guess we've discussed molecular and cellular perspectives of heterogeneity and self-voice self-made choice we've given a very simple molecular model for fake choice based on a by stable switch we've suggested that these ideas of switches and the idea of stable attractors in general of dynamical systems formalize what in terms metaphorical landscape but that we've just seen now that these models might not apply even in some textbook cases we've discussed how stencil systems have been defined by tracing by transplantation by colony forming cultures a bit about the sort of early history of that field and and we've also discussed the fact that we're sort of limited right now by experiments to thinking about ball by labeling cells using in discreet monoclonal antibodies or reporters for particular genes in order to define our hierarchy of states so that's essentially the the state right now so yeah an activity process process right right right okay so in these in vitro so if we were say this is a good point so if this was purely still can actually the paper was pretty precise about saying it's not a raster castic in the pen a sort of memoryless process which would correspond to say if it was a thermodynamic system it would be an exponential waiting time for a cell to cross a potential barrier right and then you'd expect all of the sister cells to have exponentially distributed waiting times but but the fact that they all go together rules out the idea that it's the mere stochastic activation of one of these switches now if it was in vivo you could argue well actually all of these cells are in a single environment and that environment is in contact with the bone marrow and the bone marrow just lowered the potential barrier for everyone by by providing a particular signal so then that wouldn't really invalidate any of these results because of course as soon as you add an extrinsic source of noise you can couple all of the individual cells in your system this was actually done in vitro and the cells are fairly you know they migrate apart from each other so these cells are not experiencing the same micro environment they're certainly not experiencing a more similar micro environment than the other cells in that picture so there's no argument to be made this is really this really indicates that it's a decision that was made by a earlier generation of the cell which then undergone underwent around the divisions yeah does that yeah sure you know absolutely okay so that sorry so that is consistent with the idea that it happened earlier and in fact that's that's pretty much what the suggestion is at this point in that paper is that all the cells are switching on God of one at the same time because there is there was a commitment point and then a delay and in that delay there were six cell cycles and but during those six cell cycles there's a clock running and at the end of that clock all of the cell switch on God of one that's essentially the suggestion and the commitment they don't really have I mean they know that it's happening earlier they don't know whether it's a random process or not they just know it's not gotta one coming up so it's not this specific by stable switch which involves the absolute protein concentration of God of one repressing p1 activity what's the reason behind thermodynamic well this is so this is a I mean this is not a closed system by any means there's energy I mean formerly there's energy being consumed and you know the energy being consumed is I can write this down you know in a vector form right is in several places right so this is synthesis right that consumes energy right and then degradation is an active process of proteolysis that consumes energy everything consumes energy so the whole and this this system is certainly not a term in thermodynamic equilibrium it can it can have a stationary state but that stationary state is not a thermodynamic equilibrium okay it's it's it's okay that's different it it doesn't have to obey detailed flux in any way yeah yeah so so the idea that it's a very precise counter seems amazing and I don't know whether that's really true how precise it is what what what you could imagine I mean there are there are some really good examples mid blaster transition in early vertebrates where there's a cleavage and the cleavage you know it's a very precise number of divisions and you can perturb it in a very predictable way by say increasing the amount of DNA or reducing the amount of DNA which will reduce the number of divisions by one either depending on how you how you go and I think that in the last about five years ago a very particular chroma DNA binding protein was found which is undergoing dilution and that when it dilutes enough you get zygotic transcription switching on so it's really counting the it's really the dilution the exponential dilution of a factor is really counted okay yeah and yeah it's a possibility and actually exponential dilution seems like a very noisy thing to do right it's very hard to count more than three right I mean so I yeah so I agree with you but it's certainly a good example of where there is a counter so counters do exist in this case the other thing that I think we know about this particular stage of erythropoiesis so we know two things first of all I'll show you later gene expression and we can see that there's a very very specific stage associated with pre-gata one you know this amplification module and the cell cycles are incredibly fast there are like four six hours or something so there might be some dilution going on it's you just just dividing like crazy yeah which which structure oh no so it's sorry I didn't mean it's an assumption what I mean is it's a model look any any time that you draw any piece of science you're drawing a model for what you think is going on so the question is I guess go back that's a very philosophical statement not a very useful one what we're doing over here is we're formalizing the results of transplantation in vitro colony forming experiments okay so there exists a set of certain cell surface markers and if you isolate cells based on those markers you will find cells which can give rise to all of these cell types is this a pure population we don't know all right there is another cell population which has a restricted set of fade choices is this a pure population we you know we don't know actually we do now so some of these are completely impure they're just a mixture of different cell types okay so that's an exact this is what I mean you know this this was the best that you could do with antibodies because you had a very very unique phenotype associated with a small number of antibodies when you look at the transcriptional profiles you find that you're mixing together different cell types and some of the when you get to sort of sub subdivisions of these cells you're really just mixing different proportions of cells and you're getting slightly different outcomes yeah so okay so this might be a good time to take a break and then we can we can then let's see the next thing is to talk about single cell analysis okay so it's 515 watch or sorry 415 because I have a long time sort of taking the liberty of changing the breaks a bit what actually so you tell me I can start the next bit now and talk for 15 more minutes and then we take a break is that I mean again I'm sort of I don't want to ruin the back to back okay I have absolutely no idea how long it takes because we have tomorrow why don't we okay hold on wait I'm confused it's 415 you're saying I should continue until five oh I see okay I'll get my so that's thank you that's what I wanted just to tell me what to do okay okay good so alright so now and now for something completely different okay so now we're going to talk a bit about how to profile cells and we're then going to do a bit of technical sort of looking under the hood and then at the end of today we're going to look at examples of differentiation processes and let's see how long it takes to I don't really know yeah yes right so I think I think this idea of tunneling between potentials is a very nice one and it raises these ideas of let's say directed when you take a fibroblast and we reprogram it you can think of it this way the problem I have with potentials is that we're always using them as figure seven and not as figure one okay we're telling a whole story and at the end we say oh and we can think of all of this as a potential that's for me I mean that's fine it's cute but if you actually want to create a theory which are going to use in order to infer some non-trivial biology then you're not getting there because you know you're just telling you're just just sort of it's just a nice way of summarizing some some ideas I mean what is tunneling means something very very specific in quantum mechanics and in thermodynamics as well you know thermal thumbling right this exponential crossing of a barrier in this case you know we don't have a barrier we have never defined a barrier it doesn't mean anything so so that's really I mean that's a very nice idea but that's not the problem I have is not that people haven't included the idea of tunneling is you can include whatever you want and it'll sound great that's the problem I have right right right right so that's when it becomes interesting yeah so that's when it becomes interesting if you can come up with a potential which is derived from say looking at the density of states and you can use it to say something about dynamics it becomes interesting because then you're using it as an analytical tool and then you know if we use that enough and it starts to be real we might pay attention to it and say hey maybe there's really a real potential landscape and then we can ask about its properties what's a topology and and so on you know localization and there's all kind of questions we might ask yeah okay so yeah no actually any other questions before okay okay fine so so I guess all of this is really being about stem cells and and and mapping out some so continue actually that's really what I was thinking about when I did my PhD and also when I started getting into single cell analysis and the thing I think I didn't really realize but it's really sort of hit in a big way recently is that you don't need to worry about stems I mean I should have this is sort of obvious when I say you don't need to worry about some cell diversity only when you're thinking about proliferation and turnover and really cell diversity is a confounding factor for any time that you're looking at a complex tissue this is just a cartoon from a paper that we've just published looking at the pancreas and if you're looking at the pancreatic islets which is where beta cells are localized you can appreciate the fact that there's many many cell types in the pancreas and now just imagine you're trying to you're looking at a patient who has a disorder which you can associate with a pancreas maybe diabetes for example and you're now trying to understand where what is changing so besides the fact that you need to contend with in a human with differences in age and lifestyle and ethnicity and genetic background you also have to contend with the fact that there's many cell types and your phenotype could be localized to any one of these and now starting to specifically investigate each one of these could be a lot of work and the worst thing that you could probably do which is of course what has been done very commonly is to just mash it all up and get a single sort of omics profile of everything and now you're sort of averaging and if you're trying to figure out now which compartment is being affected you're you're in real trouble so the reality is that this has become a very very very popular field of research now where almost any problem which involves a complex tissue is now being mapped and there's in fact now these huge initiatives to map every cell in the human body and so on so those are the kind of things you might want to sort of you might have a knee jerk reaction against but but so but this is just a general point which is that that there is a lot of a lot of diversity beyond tissue homeostasis so so really this is a general framework now for thinking about how to use cell single cell analysis the idea is that you start off with a mixed population which you don't have any pre prior knowledge or maybe you do and now you use the phenotypic information that you have to do one of two things well the very general thing that you do is you try to discover the structure of the phenotypic space that you're looking at so every gene is a coordinate you have a very high dimensional space and you're trying to describe that space so the simplest way you could do that is by clustering and that of course works very well if you're looking at very very distinct cell types like in the example of the pancreas I gave you endothelial cells are very different from stellate cells and they're very different from beta cells which are very different from gamma delta and epsilon cells and alpha cells so that works very nicely for particular problems and of course for more complex problems we may have a complex manifold this is showing a tree but again it doesn't have to be a tree so we can now try to essentially extract these types of structures and then interrogate them and this data can be very noisy but once we've got these structures we can take the average of these clusters or we can take a moving average along these manifolds and then get a very good view of what's changing as we walk along these structures so that's the general picture this is just from a review from Oliver Stegel, Sarah Teichmann and John Mariani who are a bioinformatician so okay so that is a very general view I guess that that's particularly leaning towards thinking about having a very very large phenotypic space to explore but let's actually think about the different single cell profiling methods that we could use we've already discussed live tracking briefly and you are probably going to hear more of that if or Stefano you've already presented your work, oh you didn't talk about this? okay all right so you've already seen how powerful that can be in order to infer spatio-temporal information this can be done by genetically encoded reporters or by just using vital dyes or morphology we can now fix tissue and then use antibodies in C2 hybridization there are now methods of multiplex readouts so that we can look at many different genes at a time we can look at facts which we've discussed briefly in the context of hematopoiesis and then there is a sort of whole transcriptome profile so these are just some of the paradigm experimental paradigms for profiling and just sort of I think I've already shown you some live imaging this is just another example this is from Valentino Greco's lab who looks at the epidermis and by so this is in a mouse and using a two photon microscope they can over several days they can follow cells and track every division that they make and then you can start to reconstruct full lineage trees and now if you can associate these lineage trees with what the local environment is doing or with the activity of a pathway you can start to ask about what is regulating these the the fake choices that are being made at each point along this tree okay so that's just that's an example this is very very powerful I clearly very very few reporters can be looked at at a time and very few cells for that matter this is just a shout out shameless sorry I had so Stefan is just written a very nice perspective about live imaging go and read it and he's got some very nice examples from work that he's been involved in looking at zebrafish renewal and he's this is now so bring together some of the things we discussed this is using a multicolor labeling system so now that this makes live imaging really easy because it's easy to follow the cells over time because they have I'm sure it's not that easy but it's it's possible because now these cells have unique colors and you can really follow them so okay so there are trade offs and I think that the other methods probably don't need too much introduction I mean I there's no need for me to show you an antibody stain section so the the some of the trade offs in choosing which technology to use are you is the effect that you're interested in occurring on a short time scale if it is probably live imaging is a very appealing way of looking at your process because you'll be able to capture what's going on very rapidly is it a very rare population for that you maybe need a method which can look at very large numbers of cells in order to catch your rare cells are you looking at a very complex population in that case maybe a noisy method is okay but you just need to sort of get a lot of a lot of different readouts so maybe single cell sequencing is great because you're looking at something that's very very diverse or are you looking for very very subtle differences in which case maybe microscopy in single molecule methods might be better where you can really pick out very very sort of very precise differences between cells restricted to a few genes so there's the different sort of different ways of looking at this sorry this is sort of very management consultancy like so so here are where different profiling methods and you can think about how many different measurements you can look at and how many cells and their various trade-offs right so single cell gives you lots of dimensions but medium numbers of cells and some of these give you about the same number of cells but fewer dimensions but then they're much lower in noise and then there's some that give you many cells and so you have different different trade-offs here and and they're different methods so I'm going to talk mostly about single cell RNA sequencing which poses really unique challenges just because of the sheer dimensionality of the data yes so that it's a moving target I can try so facts and also it depends on your experiment so for example facts in principle you can look at millions of cells but if the cells that you're interested in are only a fraction or if you're looking at a very rare population so your starting material is very dilute you could spend hours on a fax machine to just get a few hundred cells so you know it depends but in principle up to millions of cells here easily microscopy actually this is a bit on well fixed microscopy in principle you can also do millions of cells actually this should be a smear okay with a sufficient you know frameworks and computational platforms you can automatically scan huge huge numbers of cells if you're doing multiplexed microscopy you're sort of more limited in the number of cells you can look at just I mean that maybe with great effort you could increase the numbers so these are methods in which you add in one set of labels you wash them out you add another set and you then gradually cover many different many different genes live microscopy I guess you might have a better sense for there's a limit to how many stage positions you can look at and there's a limit to the magnification so you really can't look at that many cells at a time so so I would say that with life microscopy we're probably talking about tens to hundreds of cells depending on the resolution you need per session and you can you know build that up up until million so there's a log scale here okay single cell RNA sequencing talking between depending on how you're doing it but a typical number would be tens of thousands or thousands of cells okay and okay dimensionality you can pretty much look at every gene but it's very noisy here so maybe 10,000 genes say you would live microscopy or down to a handful like two three four okay so and with these methods maybe between tens and hundreds depending on if you work with someone who really knows what they're doing it that is a real art this middle section and then in terms of measurement noise I guess one useful way of thinking about this is what fraction of molecules do you detect okay and all everything follows from that if I have a very abundant gene then I don't need very high sensitivity to say what it's doing and if it's low I need high sensitivity with RNA sequencing it might be of the order one to five percent of molecules are detected and with single molecule fish and so on it might be around 80% okay so that's that's that's sort of the range and with live microscopy you're typically looking at reporters and then depending on you know for very very good setups you could even detect a single GFP molecule but that really that's in bacteria usually so typically we're talking about 20 molecules and above okay so okay let's take a break and then we'll get back and talk about single molecule RNA single RNA single cell RNA sequencing in particular after the break