13:02:24 So, today at 1pm we have Jonathan Friedman, giving a research talk, direct from Israel where is it. It is 11pm I believe so thank you. 13:02:35 And at three Andrew Murray will speak so sorry if there was any confusion about the schedule it got updated yesterday. 13:02:43 Okay, Jonathan go ahead and fix it. 13:02:47 And so I just want to say that them episode that I should prepare for an hour, but be ready that it might take 90 minutes and. 13:02:57 And I really hope it does. 13:02:59 Because questions are the best part of of any talk, and because they forced me to things to think about things that they haven't thought about yet. So please, if you have any questions, do go ahead and ask. 13:03:13 And, but then I'll start by by just setting the tone and telling you things that you already know, and. But still, I think it's important to remember right the microbes. 13:03:27 They're everywhere, and, and, and they're not just there they do important things right that they're important for human health, they do things like digest our food suppress pathogens and food and agriculture, they promote the growth of plants and provide 13:03:41 them with nutrients and also industrial application many valuable compounds are being produced by by microbes and the relation of pollutants bioremediation and things like that. 13:03:53 And, and therefore this this great interest in micro biome engineering or generating an official mobile microbial communities, which is the overall mission of Milan. 13:04:06 And to give you a sense of, or sorry, and so this, you all again know that this is difficult because then microbes and don't work alone but rather they function in communities composed of multiple interact and species so trying to engineer or control 13:04:22 the systems is quite difficult. And so there are many approaches to trying to generate some desired microbial community to perform some sort of function, and. 13:04:35 And just to give you a short overview and the most common approaches is reverse engineering so you might sequence and 16 s or shotgun sequencing of of natural microbial community, some of them might be beneficial like these ones, promoting the growth 13:04:52 of this plant and some might not be. And like these ones that maybe don't protect it from a pathogen and try and find the differences between them. And by finding differences you might find the key player or players, and that contribute to the function. 13:05:08 That's the most common approach that people in the micro biome world are taking, so far, not quite successfully, probably because of the complexity of the systems. 13:05:19 And then there's what probably most of us know and like, which is forward engineering or rational design, so maybe you write down some model parameters it, and with some measurements and then you eventually design, okay to get the best function I should 13:05:34 I should put this set of species together. Why shouldn't engineer this species to produce some compound. 13:05:41 And so that has met with partial successful for a mostly because we don't understand the system so well. So there are other approaches which are independent of our understanding of the system and one of them is direct discovery through high throughput 13:05:56 screening. Right, so this is typically done in other complex things like drug discovery. And we might just try many many many different things until you find that few things that work. 13:06:10 And, and this is being done for my groups as well. And last but not least is artificial selection right so you let evolution and work for you. And as long as you're able to impose some selection pressure for the community to do the desired function, the 13:06:25 great deployment then protect the planet and so forth. And then you can you can do selection and let the evolution, find a solution for you. So, these are think are the four main approaches that people take him today. 13:06:40 And I'm focused mostly on forward engineering or rational design. And then for the last bit of talk a bit about the high throughput screening. And of course, these two are not independent right by doing, and throughput screening, we can understand a bit 13:06:57 better how the system work and inform our models, and our models can also inform our screening, how to best direct screening efforts which strings to use and so forth. 13:07:11 And so I'll just jump right in and say that the for this rational design approach, and it's the classical bottom up approach to trying to understand how community functions and, and in the ideal world what we should be able to do is take some information 13:07:27 about the environment say the nutrients that are available, and the properties of the species maybe the genome, maybe some other parameters about them. 13:07:36 And from that infer how the species interact and what kind of structure, and the community is going to have which species are going to be there, how much of each one and what's going to be their function, are they going to be beneficial or the Truman, 13:07:51 that's the overall goal or hope or dream. 13:07:55 And so this is a difficult problem, and much more than a single research project so we broke it down into several smaller problems, to try and start making headway. 13:08:06 And so the first thing that the will try is to limit the scope and say if I know what how species interact, can I predict, and what the community's going to look like when I put them all together. 13:08:21 And in particular, we wanted to ask our pairwise interactions interactions between pairs of species sufficient to predict community structure. Right. So, when we draw one of these nice network diagrams, and we're. 13:08:34 These are species in the edges between them represents some sort of interaction inhibitory or civilization. 13:08:40 Then we implicitly assume that all interactions are pairwise right that we can draw a particular edge between a set of species. That's independent of the presence of additional species, what are the green species is here or an order there's a purple species, 13:09:01 it doesn't matter. There's always this kind of inhibitory interaction between the orange species in the blue spaces and, but this might be wrong right we know that the higher order interactions might exist. 13:09:07 And so, so that's the question, could we use just pairwise interactions this pairwise approximation and get decent results. Predicting community structure. 13:09:16 And so how does one go about it, and ask what are people's intuition so sufficient. 13:09:31 of of each species, I abundance is equal to its exponential growth rate when it's growing alone, and some leaner pairwise interactions with all other species, and in this alpha is the strength of of interaction with the company extreme when the. 13:09:40 And the naval approach, or I don't know if they've but the first approach that one of my take is to try and them, and write down and model, and with pairwise interactions, like this, classical generalize local terremoto will say that the rate of change 13:09:55 So we broke down this nice airways model. Now we can get some data, try and parameters this data, and using the parameters that we've been furred maybe these are his growth rates and of ij the coupling coefficients. 13:10:11 And then we can try and predict what happens when we put multiple species to get her. 13:10:15 And then if predictions fail, we would say okay directions are not pairwise, and if predictions are accurate then we can say in directions might be paradise or is pairwise approximation is sufficiently accurate. 13:10:28 And the problem with this approach is that if it fails if we're unable to, to make accurate predictions, it doesn't mean that interactions are not paralyzed. 13:10:37 Because in the real world, and the simple model is typically not the right model, right, we know that species don't just grow exponentially. They grow with some complex function which might be different for each species and interactions are also am not 13:10:53 they could still be pairwise interactions or at least when I say pairwise I mean, p always between species not between individuals, so they can be some complex nonlinear, and interaction or function, which is still just a function of these two species, 13:11:09 independent of the presence of other species. 13:11:12 And in this case, if, if our predictions from the simple generalized multiple Tara linear model fail. It's not because interactions are not pairwise it's because they're nonlinear. 13:11:23 And so, we, we won't be able to tell whether higher order interactions and requiring are affected by the presence of additional pieces are there are no. 13:11:33 And so instead we decided to take another approach and, which is slightly more in qualitative and to just say, If all I know is the outcomes of pairwise competition so in this pointer to an example of three species A, B and C. 13:11:52 And I put each pair of species together and I look at the outcome qualitatively do both species coexist and survive or this one of them go extinct. 13:12:02 And this cartoon example, A and B B and C, both coexist and survive but when I put AMC together, and I'm left with just a and CCC goes extinct. So if I have this information. 13:12:17 Am I able to make predictions about what happens when I put all these pieces together. And by predictions here I mean, which species is going to be president or surviving which species go extinct. 13:12:26 And, and the benefit of this approach is it doesn't require making parametric assumptions about the nature of interactions. And it also doesn't require ensuring parameters which in and of itself can be can be quite challenging acquiring large amounts 13:12:41 of data that are not always available. 13:12:46 of data that are not always available, and the drawback or the price that we pay for the simplicity is that the first we make qualitative predictions about just presence absence and we know nothing about the dynamics, this is purely steady state. 13:12:59 And Mr. We wanted to know how well this approach works. And just to be clear, it doesn't have to work, right, even though it sounds simple, and we know that higher order interactions can exist. 13:13:12 And for example, if our species a inhibited or killed off species by producing some antibiotic. It could be that species B is an antibiotic the greater and, and it modifies this interaction between them in the presence of species be maybe a no longer 13:13:28 is able to kill see. So this is a higher order interaction and affected by the presence of all three species and. And in this case, our simple approach might just fail. 13:13:40 Right, so we know these higher order interactions exist, we know that they can affect community structure. We just don't know how often it happens. 13:13:50 So that's an empirical question. And to answer empirical questions you need experiments. And so we we use the following experimental system. We took a set of eight sewing material. 13:14:04 This is the father genetic tree. It's not really important, all you need to know is that these two or three letter abbreviation so just names of species. 13:14:12 And the nice thing about the system is that you can easily distinguish between species by just looking at the requirements one place to another, and you can see the bird distinct mythologies and colors. 13:14:25 And, and then to measure interactions between species or outcomes of competition. And we did simple a batch partnering with the periodic solution so you put together a set of species in this case to, and you let them grow and interact for a while, then 13:14:41 you dilute into fresh media, and you repeat this for a few cycles to let the system and calibrate and then it then. 13:14:49 We played them we look at it, which theses are there, and which are extinct. 13:14:56 So to give you a sense of what the data looks like. And this is one particular pair of species, and here, and PV and PP are the two species, and I'm plotting the fraction of one of the species as a function of time these growth cycles, starting from five 13:15:14 different initial conditions so when both species are present that equal abundances 5050, and when one of the species is the majority, or the other species, the majority and the idea is to test whether there's dependence on initial conditions, or whether 13:15:28 the system is a global attractor or multiple thing, steady states, and in this case you can interrupt for a second. Yeah. 13:15:37 Excuse me. Can you remind me what the nutrient environment is here I don't know. Yes, thank you. I didn't say, so this is a minimum media am nine, where the limiting resources carbon, and we hit two carbon sources Gulf chronic acid, which is a slightly 13:15:53 oxidized version of glucose, common in soils and Syria and amino acid, as a carbon source and and Was there some reason for those two in particular. 13:16:05 And was there some reason for those two in particular. It's semi arbitrary so we tried looking for things that are somewhat common in soil, and we tried several different things looking for combinations that allowed for coexistence have enough of our 13:16:19 strains. So, as a side note, because it's a bit interesting. So, this is builds off on previous work it was done in Jeff cores lab, and where they use them in nine also minimum media but containing 31 carbon sources, because they wanted to get her existence 13:16:38 of multiple species. So they thought, many carbon sources would facilitate coexistence. And I preferred simply find them on the media, and such that we might rationalize what happens in terms of who's doing what and 31 carbon sources seemed like a lot. 13:16:54 And so I started reducing the number of carbon sources and quickly found out that you actually don't need that many carbon sources to get the existence of multiple species. 13:17:03 And so, in part, the set of carbon sources is based on the original 31. 13:17:11 And do you have data on how individual strains are utilizing the nutrients, I assume nitrogen and phosphorus are not limiting. 13:17:18 They are not limiting for this data I do know, for the next, the next week I'm going to show you we do know how individual species, utilize the nutrients or the carbon sources. 13:17:33 I don't know why or what exactly you're asking I'll just say that we have not been able to make sense, or doesn't seem to be in an easy connection between who's utilizing what carbon source and what kind of interaction. 13:17:47 They have with the likelihood of them coexisting. 13:17:50 And I'm just trying to get at whether the outcomes of these experiments are understandable in terms of resources. Yeah, so what you're looking in, not in an easy way. 13:18:00 I would say, related question. 13:18:03 If you do change the media and repeat all of this, would you expect that these two body interactions change but still your broader resolve that complex communities can be broken down to two body interactions, would you expect that to be. 13:18:18 Exactly, yes. 13:18:19 So the particular interactions change but the overall picture remains robust at least in the media that we try. 13:18:31 Other questions. 13:18:37 Great. And, yeah, please do keep asking questions throughout this is great. 13:18:43 And so right I was showing you that the this particular case show competitive exclusion where this top species PP excludes the bottom species PV, and this is a global attractor It doesn't matter what the initial condition is you always converge to it. 13:19:14 And another qualitative and outcome was coexistence. Again, this is a different pair of species, the same top one but a different in a different competitor PA, and starting from five different initial conditions as a function of time now, and the community 13:19:29 converges to a state of coexistence we're both species are present. Again global attract are no dependence on initial condition. And these are the only two outcomes that we've seen in the system within all the pairwise competitions, and no movie stability, 13:19:47 no limit cycles fluctuations, nothing, just these two simple things. 13:19:53 And so this is promising that predictions might be feasible. 13:19:58 And so as I said, we measured all the paradise, and outcomes, indicated by this network. And, as I know two interesting things about it one as I said there's a lot of coexistence here, and you might think that this is easy because there are two carbon 13:20:14 sources. And that seems to be independent of the ultra Sanchez group is shown that you can get rid of coexistence on individual carbon sources, presumably because of a lot of cross feeding, and in regards to step is previous question I think the reason 13:20:33 it's hard to make sense of, of these outcomes in terms of usage of our primary carbon sources that they're not the whole story or not even maybe the main story. 13:20:45 I think a lot of the secreted metabolites are important as well. And, but anyway. Here you can see a lot of coexistence with the cyan arrows. And the other thing is if you notice them, whether it was the competitive exclusion where for example this species 13:21:02 exclude this one. They point upwards. I organized it this way. And so there's some directionality from poor competitors like this bottom one two good competitors like this top one in particular, there are no cycles and No. 13:21:22 in Rock paper scissors or in transmitted interactions and, and we found this to be the case, even in a much larger data sets so competitive cycles are an interesting theoretical possibility whenever we looked we couldn't find any almost any. 13:21:37 And, but this was a side note, so now we have, we have this network, and we want to make predictions based on this. And so again, here's a particular trio, and the edges, show you the pairwise interactions so this talk species, EA excludes this left species 13:21:54 SM and data to coexist with the stars indicating the existence for action. And now when we need to make some predictions. So we said we'll try the, maybe simplest, most naive assembly rule, one can imagine, which is that species will survive in the trio 13:22:15 if and only if it survived in all the pairwise competitions. And so in this case, the prediction is that will end up here, where the square indicates to coexist and so these two species were this left species SM is excluded because it loses out in competition 13:22:33 to EA. 13:22:36 And so, to test this prediction, we did the experiment. And so this triangle is also a simplex, it shows you the fractions of the three different species. 13:22:47 So the point in the middle will be equal abundances of all species, and the edges would be sorry the nodes would be 100% the 100% SM, and some combination of the two of them on the edge. 13:23:01 So we started competitions, from three different dimensional conditions dominated by just one species, and then you can follow the trajectory of the composition over time. 13:23:12 And, and I highlighted. 13:23:14 The final composition by these, these colored dots. And you can see that our predictions were great in this case we need ended up with coexists of these two species as predicted in exactly the predictive factional very close to it. 13:23:29 And so this is very nice. And, but of course it could be an example the diet, cherry pick for you, and which is not representative, and, and, and to know whether this is typical the pairwise interaction or outcomes are are predicted, and you need to look 13:23:48 at more data or more examples so we looked at all. 13:23:58 trios of our data set, you can see much accepted the, I like the colonies and have done a lot of competition experiments. And so summarize the results. 13:24:04 So for each one of these trios again we will look at the pairwise outcomes. We use the simple assembly room room to make predictions about who's going to survive and who's going to go extinct. 13:24:10 And then we compare this to them to the measure the outcome of the trio competitions. And it turns out the correct prediction, and happened around 90% of the times, and we made it, we fall asleep predicted to the species will survival of the species go 13:24:33 And so 90% accuracy is quite good thing and but I do want to highlight two of the most interesting cases of surprising outcomes were simple assembly will fail. 13:24:45 And so, this is one case where again this top species out competes this left species, when it's just the two of them, but the other troopers Paris coexist. 13:24:56 And, and we would predict that these two species will coexist and the community will end up here. 13:25:03 But actually what happens is that the top PP species loses out to the combination of these two, even when it starts out dominating the community, 95% that quickly goes down in abundance and you end up with just used to. 13:25:16 So it's a case where neither one of the individual bottom species can exclude the top one, but together they can. 13:25:25 And the other interesting case is alternative stable states, or by stability. It's, again the same type of pairwise interactions, but now you can either end up in this state with a lot of PP when you start with a lot of PP or this bottom state with very 13:25:41 little tip when you start out with very little and pp. And this is the only case where we had the alternative stable states or dependence on initial conditions. 13:25:54 And now what's, I think interesting is that if you, if you look at this in, when I first looked at this I said okay, these are clear cases of higher order interactions for pairwise interactions will not do this. 13:26:07 And, but I was, I was mistaken, and even simple are simple local Tara model with just linear or pairwise interactions and can actually recover these dynamics. 13:26:20 So, this is these are simulations and parameters with the parameters fit in from these data. And, and you can you can find a particular set of parameters that gives you this is surprising outcome. 13:26:37 And you can also get by stability like I showed you before. And so these surprising outcomes are not sufficient evidence to rule out that it's just fairways and or even leader repair was interactions and an overall. 13:26:53 If you and simulate communities from this generalized local therapy community, you make predictions using our simple assembly rule, and then they're not always correct right because you don't have all the or you don't use all the dynamic information you 13:27:05 just use the the outcomes, which constrain the values of the possible parameters to some extent but doesn't give you the exact parameter values. so you don't make it perfect prediction every time. 13:27:17 And, and actually in the simulations that this is the distribution of accuracy is that you get in local Tara and simulations. And you can see that are 90% accuracy is not unusual. 13:27:34 And with the, with just simple local Tara dynamics. 13:27:39 So from this week concluded that, at least in this system. If there are higher or the integration interactions they're either sufficiently rare or sufficiently week to not really affect the community structure, at least in terms of who's there was no. 13:27:57 So there's a question, hold on a second, 13:28:02 would you expect that if you have species with dramatically different growth rates, such that the time taken to compete in different pairs are dramatically different that you would actually see different outcomes. 13:28:15 Yeah, that's an interesting question. So, so I'm not sure I understand Do you mean that the intrinsic growth rate the exponential growth rate of different species this our parameter is different between different species, or that somehow in the context 13:28:31 of or in the presence of other species, and the growth rate changes. 13:28:38 So I guess like it's maybe a much simpler question so if you could go back maybe two slides, I think it's easier. 13:28:46 Yeah, maybe this one. 13:28:51 Okay, so in this case, the time taken to reach either of the fixed points is the same five dilutions. But suppose one of the fixed points it takes 10 dilutions right like they're both close enough to each other, that they're not actively competing in 13:29:04 some way or they're competing slow enough that they both can survive each other for a much longer, but if you give it longer, they reach a slightly different fixed point, that makes sense. 13:29:14 So you can say that empirically. 13:29:18 Some peers do reach their equilibrium point or they're fixed one much faster than others, and doesn't seem to correlate with the accuracy of our predictions, I guess, it is possible that if there are very different time skills that maybe you first converge 13:29:39 to some quasi steady state, and if you're not careful, you would think. 13:29:43 And this is the the actual steady state but if you wait longer than you'll see the change to the real attractive or fixed point, and the time it might take you to get there or the dynamics of getting to the fixed point where the really fixed point whether 13:30:03 you get stuck in this quality was a steady state or not can depend on initial conditions. And we haven't seen that empirically but I'm not sure if that answers your question. 13:30:07 Yeah, I guess like it just seems like all of these species are close enough when you separate them and growth rate that that may not happen but, no, thank you. 13:30:17 And, yeah, so there are differences in their growth rate of. 13:30:34 So it's not orders of magnitude but there is some variation. 13:30:38 And so I'm going to switch gears just a bit and say that then. 13:30:45 All of this was done on fairly short time scales. And so, roughly, 60 generations, something like that. And, but then wondering what or will we be able to make predictions and on a longer time scales where evolution like like play an important role in 13:31:05 community structure where in these short time skills, presumably it's mostly evolutionary changes. And because you things don't have enough time to arise and spread. 13:31:18 General. 13:31:27 Yeah. And so definitely we can you know hand-waving Lee say that these carbon sources are found in soil but but yes, there are definitely not well adapted to our lead conditions. 13:31:34 These are this media the media us are all things that none of these organisms see naturally right they're all very different than what they. 13:31:40 So sort of a pro one think there's a lot of potential for rapid evolution there absolutely is. Yes. 13:31:48 And so they experimented the, they'll tell you about now, and is done where the species are not adapted at all to these conditions, and yes I think that's why we get a lot of rapid evolution. 13:32:03 And right now we are conducting a follow up experiment where we use or create adapted species from this experiment, then we'll see. We hope to see how, how much does it make a difference whether species are already pre adapted to the environment or not. 13:32:23 Okay. And right so we want to know what happens on longer time scales and, and the first things we wanted to ask, we realized we don't even know over what time scales would community structure be stable, and we don't know when changes happen how repeatable 13:32:43 are they, if we run the same experiment and multiple times when we get the same changes or not. 13:32:49 And, and if things change how we'll do our prediction rules apply if these evolutionary time scales. 13:32:59 And so to answer these questions we took again instead of certain species this time 16 strains. Some of them are the same strains as before they're in from a culture collection at CC and some of them we have isolated and from from soil, and this is their 13:33:16 father genetic tree a bit more diverse this time. And, and we propagated all 16 species in model culture as well as about 40 pairs of species and 40 trios with several replicates each so we can see how repeatable things are, and. 13:33:38 And then, as before we do this growth dilution cycles, and it's similar media, this time, there's three carbon sources where that acetate, and it allowed us to to culture, this wider range of species. 13:33:52 And again we grew them in cycles of 48 hours, followed by the illusion by five to 1500 and we repeat this for 38 cycles giving us about 400 generations. 13:34:04 And in through these 400 generations we measured the total absolute abundance, or proxy of it. 13:34:16 The ODM and n od increased almost in all cases, again, probably indicating that the species are not well adapted to, to our experimental conditions our priority, and Jonathan. 13:34:28 Yeah. 13:34:29 Yeah. 13:34:30 So just a quick question you said that here the dietitian factor was 1500 Correct, yes. 13:34:37 What was the direction factor in the previous experiments that you were talking about. 13:34:42 Yeah, sorry I didn't mention but exactly the same it's exactly the same statement. Thank you. 13:34:51 Okay, so we measured the OD so absolute abundance and we also periodically plated to get the relative abundance. So, we have these data. 13:35:03 And then, said the first question we had was over what time skills does community and structure change, will they remain stable or the same composition after they've reached their theological steady state. 13:35:17 Will they stay there for hundreds of generation will they quickly change. 13:35:21 And so I'll just show you some examples. And so this is a pair, and where you can see it started out from some arbitrary initial composition, and it quickly over something like 6070 generations, which is some fixed point and then stays there for the rest 13:35:45 of the experiment for 500 400 generations, and the same happens in this trio. And so there are some combinations of species that equilibrate and then stayed in for the rest of the experiment. 13:35:54 And, but this was actually not typical what was more typical or dynamics, like this. And so here's another pair, again, from are arbitrary the initial condition, rapid convergence to some steady state. 13:36:09 And by the way, I say convergence but you can see here because they all started from the same initial condition. 13:36:15 But we've done some other experiments where we did start from different initial conditions to show that it's actually some effect for, and then it stays there for a while, and no longer time scales in the composition changes, you can see that the cross 13:36:33 replicates so each one of these lines is a replicate this variability in the timing that this takes and the exact trajectory. 13:36:41 And, and the same happens here in this trio. And here it's less obvious but they they converge to some steady state and then they change. 13:36:51 And now some of you by the way might be wondering if if these are actually evolutionary or heritable changes, or just maybe some long evolution and ecological transition. 13:37:01 So I'm not showing it here and but in case you're wondering what did we isolate strains from the end of the experiment, we started the competition's over the CO culturing. 13:37:17 And this time, species didn't again follow this trajectory, but rather quickly converge to these new fixed points almost always indicating these changes are indeed. 13:37:27 Some period that will change. 13:37:30 Jonathan here, you're estimating relative abundance using the same method that you were describing before just by plating colonies and yes okay and so and so, over this entire time the, there's no obvious change in how these colonies look. 13:37:46 Yeah, some of them do change in morphology, and they don't change enough that it's. They look like the other species, and there were some cases where we were not quite sure. 13:38:00 And so we either exclude them or we send them to sequencing and but I'm not showing any cases where we didn't include any cases in the analysis where there was dumped on it. 13:38:12 Thank you. 13:38:13 Thanks, 13:38:17 and Martin Can I ask a question. 13:38:19 Yeah, please. 13:38:21 Oh, for the media that was used to make plates, good they have the same composition like same carbon source and and nine. 13:38:43 No, so the plates, the other places, on which we play that these are rich media these are nutrient broth. And the idea is, they live in, or sorry you played in the loot enough conditions to see individual colonies so the goal is that they don't interact, 13:38:52 they just each single cell that lands on the plate grows rapidly it makes a colony. So it's better to do this on which media grows faster and the colony morphology is more distinguishable. 13:39:08 Okay, that makes sense. 13:39:11 I mean, do you think I'm plating would have had some effect on the quantitative connotation of counting more days, these bio mats for the, for the endpoints, the steady state. 13:39:29 Yeah, so I'm not sure if that's what you're, you're asking but they could be different plating efficiency, where different species the fraction of cells that are in the culture that are able to grow in the faith can be different. 13:39:45 And. 13:39:46 But, so that's certainly possible and that could change through time. And I think that's that's, I mean, probably happens but it's a bit unlikely that this affects the results because everything is done by plating so we're always comparing clothing for 13:40:03 plating. So unless you know during evolution. 13:40:06 One of the species changed its plating efficiency dramatically and the other didn't. 13:40:14 I think we're safe. 13:40:22 And so I showed you some examples here. And we can also quantify this for all of our communities all of our replicates so each dot here is one replicant have one community and what the show is over time. 13:40:38 And the change in composition since generation 70 which is roughly here the time with a equilibrate. 13:40:47 And so you can see that the community is changing their composition. And over time, and maybe they slow down, at the end, and maybe not, we don't know what would happen if we kept going, whatever, that would be after changes went on like what this is 13:41:04 the final state this community or will change again. 13:41:09 And. 13:41:12 Okay so, not surprisingly, and now if we want to know how well our predictions of community structure work over these times skills they don't work very well because community structure change. 13:41:24 And what I mean by that is if we only know what the composition was it the ecological time skills in predictions are going to deteriorate over time. 13:41:34 And to illustrate that. Here's again a simple, simple acts of three species, and the stars give you the composition of pairs of their generation 70 so it is a logical times game. 13:42:01 And so I wouldn't really go into how we do this but it's again some phenomenal logical 13:42:21 and modern where we say that the fraction of some species will be proportional to the geometric mean of expression in each one of the pair in the pairwise competitions. 13:42:23 And, and, in the red square here tells you what our prediction is. And, and you can see by this is quite good, and over, political time skills, is consistent with our previous result and just for reference, you can say well if I don't know anything. 13:42:38 I don't have the pairwise interactions. Maybe my best guess is just yes equal abundances of all of them. 13:42:45 And this is the.in the middle. 13:42:48 And, but then a generation. 13:42:52 400. And so our prediction, based on ecological information remain here, but the composition of the trio change to over here. So our prediction is, is no longer accurate. 13:43:04 Again, not surprising. 13:43:06 And, but just quantify this over time. And so this is a function of time generations and this is the distance between our prediction and observation, so low distance means accurate prediction and large distance means not accurate. 13:43:22 And, and the red line tells you how accurate our predictions are for each generation, if all we have is the ecological information and distraction generation 70, and the green line tells you what will be our prediction accuracy if we know nothing and 13:43:39 just guess equal abundances. So you can see that the the accuracy and the two rates over time and filled by generation 400, knowing what the ecological fractions are is almost in his poor predictors not knowing anything. 13:43:57 Jonathan I'm a little confused by the plot is are you making a prediction, based on the pairwise interaction inferred at some generation at a future generation. 13:44:08 Ah, ok, sorry. And so, what I'm doing is I'm, I'm using the same prediction it's always the same prediction, based on the fractions that generation 70 so the psychological study states, and I'm saying how accurate are these predictions at different time 13:44:24 points. So it generation 70 they're pretty accurate generation 100, they're not as accurate in the generation 400, they're even less accurate measured as a distance on the simplex I suppose. 13:44:36 Yes, exactly. 13:44:38 So, this will be the distance from the red square to the centroid of the orange dots. We're here it's a very low short distance and here it's a very long distance. 13:44:51 Have you tried to isolate the three from the final ones and see if you then the pairwise interactions give you a view of the whole community. 13:45:02 And, yeah, so we have isolated them, and we're working on it now, we haven't measured all the affair wasn't directions of the both one says it's just a lot of work. 13:45:11 Hopefully we'll know soon. 13:45:13 And the presumably maybe substantial variability within the volt strength and maybe no longer anywhere near. 13:45:19 Yes. 13:45:20 So definitely replicates sometimes, like this one, they go to very different directions so presumably the clones the orange clones from here behave very differently from the orange ones from here. 13:45:34 And a couple more questions I have one more and Ben has one to kill shots question about the plating efficiency. I'm also concerned about that have you validated these relative abundances by an independent means. 13:45:48 So the short answer is no. And, except for a few cases where we've done sequencing, and it's roughly consistent but then, I'm not sure maybe that they understand the concern because. 13:46:03 Right. Every quantification method has its limitations. And what are the sales or if you do Fox or something whether the sales are viable or not viable. 13:46:14 And I think the point here is that we always compare and see if used to see a fuse. 13:46:19 And so, and also see if use of the same species so let's see if use of species a in a pair and see if use of species a in a trio, but but I guess my concern is that Imagine you have one strain where the plating efficiency is going up as a function of 13:46:34 time. So then it looks like an ecological interaction when in fact, all you have is a single strain adapting to being plated. 13:46:41 Yeah, I mean, that yes that can happen, and we cannot rule it out, or we have not ruled it out. Though the selection is not for plating. And we propagate them without playing directly from the MIT Media to another MIT Media planning is just for quantification 13:47:01 so if if some strain is better at playing it doesn't give it to me selected advantage. 13:47:08 I guess have related question building on Daniels. 13:47:12 Do you have any data to indicate that the environment in which the strain evolves has an impact on what the sort of shifts in ecology are like the same species evolved and monoculture still has the same impact in Thai culture, and so on. 13:47:27 Sorry. Can you repeat that I'm not sure I understood. Yeah, I guess I'm wondering whether your data can can tell you anything about the strength of CO evolution this experiment where let's say that the purple and orange species in the bottom left here, 13:47:40 you know evolve this shift and ecological structure over the course of these 400 generations, you know, is that because they were present in the same class together or with the purple species on its own, have acquired the same kind of mutation that would 13:47:54 lead to such a shift. 13:47:55 Yeah, so I'll show you some data pertaining to that in a few slides. So let's say we don't know the genetics yet, and we are sequencing these but don't have any results yet. 13:48:10 What I can say is that if you take, because we've, we've also evolved these species in monoculture in isolation, so you can take the say purple species in orange species that have evolved in monoculture and then compete them, and they they reach very 13:48:24 different, almost. Most of the times they reach very different steady states than the ones reaching co culture. 13:48:31 Cool, thanks. 13:48:33 And how does the absolute number of cells, change. 13:48:37 So the absolute number of sales increases over time, and the cultures started with a low initial number of sales increase more than the ones that start with a higher number of sales originally, and somehow consistent with this picture that this may be 13:48:57 an optimum some peak, and the closer you are to it, the less you have to improve and the further you are from it that initially, the more you have to improve. 13:49:11 Other questions. 13:49:14 Gentlemen, can you explain one more time. How you the GOP model, how we fit the gym, I didn't see a gap model here at all. 13:49:26 Are you, are you generate a prediction belt. 13:49:30 You mean the previous. 13:49:35 And, like here. Yes. 13:49:38 And, yeah, so I took all the data from the pairwise. 13:49:47 These ones. And for each of these you can you can fit the GOP model. I mean, we can go into technicalities but if you want that maybe that's that's better than offline, because there is some finesse to how you do that. 13:50:04 And, but the, so you can do this for each pair and get some alpha values so some interaction coefficients, and then you can also take the monoculture data and get some growth rates and caring capacities and. 13:50:20 And then, and I didn't say that. But, or you can talk about it but you can you can use these values to make predictions as well. And in this predictions are actually slightly less accurate but then if you use the simple assembly room. 13:50:32 And, but for the purpose of the simulations, and done here. And then each simulation is done by randomly picking growth rates and direction coefficient, from a distribution that's similar to the distribution of of actual limb or if the infrared interaction 13:50:51 coefficients. 13:50:54 Hi Jonathan. 13:50:56 So to two questions I guess the first question is what the economy plating efficiency in the Lenski experiment, we've had the same issue of strains that are evolving over time, they even though they're not adapting to the plates that you played them on 13:51:14 for competition experiments. They either increase or decrease in depleting efficiencies in a way that's, you know, completely irrelevant other adapting to minimal media, some pie topic effect of mutations that getting to the minimal media. 13:51:28 So for some strains have just stopped doing competition experiments on plates. So it's just something to I guess consider sequencing has become more reliable now. 13:51:40 Yeah, okay. 13:51:42 The second question is. So you said that include them in mono culture, and they were not predictive of pairwise interactions, but when you these pairwise strains are predictive of the trio's. 13:51:55 So it seems that then competitive exclusion ability is completely different from how it spreads evolving because there's other nutrients at play, but it seems when you add a third layer. 13:52:09 It doesn't really change things too much. Do you think that's a function of the media because of the niches that are available in minimal media, if that changes things were different or do you think it's more general property. 13:52:20 Yeah, I don't know because this was indeed surprising to us that there's a big difference between evolving in mono culture and culture but then culture will tues with one other partner or two other partners seems to not really make a big difference. 13:52:37 And yeah I don't quite know what to make of it or what kind of maybe simple logic and can account for that. And because of that I also don't have a sense of whether this result will be robust to changes and say, media, and yeah so I really don't know. 13:52:55 But if somebody has ideas or thoughts that would be great. 13:53:00 Thanks 13:53:05 to iron and the question to Jonathan. So, these, these blood survey nice brand notice that you say you're making a clear distinction between ecological time scales and evolutionary timescales right. 13:53:16 Is it a, is it a Claire, is it a clear way we can see that in these plots like how, how do you discern between these two timescales. 13:53:37 Yeah, so, indeed, it's not maybe is such a dichotomy. It's maybe just easier to refer to them this way. And in these clouds, maybe here you can you can sort of see a hint that there's a rapid change at first and then plateaus, and then then subsequent 13:53:46 changes are happening at more random times. And, but the, the, I think the real way of seeing it is. 13:53:57 I have some slides at the end, maybe I'll show them later but instead, you can start with them with different initial conditions. And then you see the convergence to turn the tractor to a fixed point that's independent of your initial condition. 13:54:12 And that I think is indicative of reaching an ecological equilibrium. 13:54:16 And, and then on a longer timescales after you've reached a deck Olivia. When you start seeing additional shifts, then these are likely evolutionary. 13:54:30 And instead, the way to verify that these are indeed, some evolutionary change, other than sequencing is to isolate strains from some sometime point and then restarting the culture and seeing whether they go through the same long trajectory or rapidly 13:54:48 converge to the new to the new fractions, which is typically the case. 13:54:57 Right. Okay, thank you. 13:55:04 So maybe maybe I'll skip this a bit, and for the interest of time, and I'll just say that 13:55:05 Okay. 13:55:14 the variability between replicates them to be low. And so, I showed you. 13:55:22 And maybe these two cases where this one the, there's a large variability between replicates and this one is the low variability between replicates. This one is much more typical, and this one is maybe the most extreme case of variability between replicates 13:55:39 we quantify this in a variety of ways. And, but the other thing to say is, again, this, this case which is maybe the most extreme case of variability, even here, you can see that in all cases, cross all replicates is the orange species that goes down 13:55:55 in frequency and the purple species that goes up. 13:56:00 It could didn't have to be could be that the orange species would would increase in exclude the purple species but that never happened. 13:56:10 And, and again, even when there's variability in the exact quantitative fractures, which he sees increases in which species decreases tends to be quite conserved, much more than you'd expect by chance. 13:56:22 And this is just some, some tests joined it. 13:56:28 And so, because things seem to be repeatable, most of the time, then, then you could imagine that maybe 13:56:45 these repeatable trajectories of the current pair would be predictive of the ones that the current trios. And so, I guess you all know the end of the story by now, and because it said is, and so what you can do is before we made our predictions, just 13:57:03 based on the this information from generation 70, maybe, psychological times they do. 13:57:09 And by the way, maybe it's a good time to say that I don't think that this number 17 generations is some magical number that should be conserved across biology. 13:57:19 And I think it is dependent on the specifics of the experiment the growth rate of the species the population size and dilution factors so bottleneck size, and what mutations are available, etc. 13:57:35 But in our case we use generation 70 Woods, and most communities seem to converge to to a fixed point by then. And so previously we only use that information and try to predict what happens in trios in the future. 13:57:52 And, and now we said, if we know what the fraction of species of pairs is at some given generation maybe a generation 400, and then we can make predictions about what happens, or what the composition of the trio is at the same generation. 13:58:07 And so this is this green x is again our prediction, based on on the simple geometric meme, and. But now, not the done by by test composition of pairs, but by present one so at the end of the experiment. 13:58:23 And you can see the screen x is now much closer to the actual measure composition of the trio's, which are these orange dots. 13:58:33 And so the same graph has before showing you the accuracy of predictions now I've added the blue line, which is the accuracy of the predictions of to composition based on the composition pairs, at a given time. 13:58:49 And, and unlike the red line, based on the ecological information only which keeps going up and becoming worse, and the blue line doesn't even comes down a bit. 13:59:01 We can discuss why we think it comes down if you'd like. 13:59:06 and. 13:59:08 Right, so let me just summarize quickly this part and say that we've shown that the pairwise interactions can quite accurately predict the presence or absence of the disruptor over ecological timescales we show that the structure of these communities 13:59:25 changes, and typically within 400 generations that any less presumably because they're not well adapted this environment. 13:59:35 And, but the core evolution of Paris is pretty well predictive of the coalition of two years. So, if before our conclusion was that the higher order interactions are not strong, or rare, and therefore the effect community structure over evolutionary time 13:59:54 skills than it seems like something has similar happens also on evolutionary time scales. So higher order evolutionary interactions, also seem to be rare or week. 14:00:07 Jonathan Sorry, can I interrupt you for a second, go back to the previous slide. 14:00:12 Yes. So in this slide, given how the prediction with the present generation pairwise compares to pairwise a G 70 remains consistent closer to what looks like 200, maybe 150. 14:00:26 Is it reasonable to assume that this rate at which the evolution is actually changing the interactions is slower and the initial adaptation is much more monoculture growth rate, and that's where your predictions hold up for longer before which at 200 14:00:43 to 400 is where the real differences start showing in terms of evolution is that the way to think about the deviations here. 14:00:52 So, actually think about it a bit differently. 14:00:56 And I think what happens is that changes tend to or similar changes tend to happen in these cultures but at different times. 14:01:07 So you can see again in this example the orange species goes down and sometimes it goes down quite fast, and sometimes it goes down quite slow and. 14:01:18 And this variability in timing happens in the trio's as well. And initially the accuracy of predictions goes down I think because there's these mismatches in times. 14:01:30 And, and when you get to later in the tech course will, this is not the greatest example but in others, it's more apparent. And then you're only past the time we're ready for species increases in abundance. 14:01:44 And it happens at some random time between whatever generation 150 and generation 300, then by generation 400 in all the replicated increased. And that's true in pairs and trios. 14:01:57 So I think the fact that this decreases here, the accuracy increases it's because the transients or some of the trends in which is quite stochastic and maybe when you find it beneficial mutation is only test. 14:02:12 And now all the communities and replicates in pairs and fields are starting to converge to to maybe this new and steady state. That's how I think about it, but then based on what you just said you should be able to been sort of timescales of transition. 14:02:30 So, so, exactly the way you just said if you separate the trajectories and the ones that find the sort of evolutionary trajectory earlier versus later. 14:02:39 And you separate your prediction in that way, you should be able to see them. the distance change accordingly. The way you expect No. 14:02:46 Yes, I we were trying. It's messy. So, I see him because the reason I guess I was confused is the fact that the variance. Also isn't increasing because like I thought what you just said, which is that the timescale being different in different trajectories 14:03:02 might be why it surprises initially, but then I'd expect the variance of the prediction to also rise in the initial stages where everything is starting to separate from each other, and then it to converge back later whereas that's not what's happening 14:03:16 it looks like or unless obviously the graph may not be the best way to view that. Yeah. And so, maybe a better graph, and which I can dig up later is, is exactly this variability between replicates over time. 14:03:32 And this one, indeed, it initially shrinks. When you get to the maybe political point of view, then it builds up again. So replicates become more variable, and by later in the experiment. 14:03:45 So generation 300 and, and on, then it narrows again. 14:03:52 And so that happens within replicates of each pair or trio. 14:03:57 Awesome, thanks. 14:04:02 So, how we doing for time. 14:04:08 We have 20 minutes, something like that. 14:04:13 Okay, so I'll tell you one more. Sorry, I don't want to go over this too many times but if you could just go back to the plot the story was asking you about this one. 14:04:24 Yeah. So that's aggregated across all pairs and trios is that correct, yes, missy. and does the plot. 14:04:33 I suppose we would have to talk about it later, but how does that plot look you know for a specific set of pairs and trios Does, does it does the blue curve, you know, really dive towards zero, I guess what I'm asking is, are those standard errors or 14:04:47 standard deviations. 14:04:49 These ones. Yeah, Yeah, these are standard deviations. 14:04:53 Okay. All right. Thank you. Yeah. 14:05:04 Okay. 14:05:04 Right, so I want to put the list bit to them keep short I want to switch gears and talk about the environment. Right. So, so far in our big scheme of trying to build up complexity, and we we focused on, you know what the species interactions are, and 14:05:24 what will be the community structure, and that's all true in a given environment. If we change the environment then like, like somebody is all the interactions are going to change. 14:05:37 And our predictions are not going to be accurate. And so we wanted to know how the environment affects species interactions. And, and, In particular, we know that. 14:05:48 I hope I convince you that, to some extent, if you know what the interactions are, you can make some predictions about the composition or maybe even dynamics of the community. 14:05:58 But, as you all know, if you know the distribution of interactions maybe not the detailed network but just how many competitions you have, how many mutualisms you have. 14:06:09 And then that can also affect the overall properties of the communities like maybe its stability. 14:06:16 And so we wanted to ask what is the distribution of microbial interactions and. And then as we said it depends on the environment because could be that in one environment you have lots of competitions, and maybe the community stable and. 14:06:31 And in a different environment, then the distribution shifts and you have lots of mutualisms and maybe the communities then listed. 14:06:53 So not only do we want to know what our typical distributions, look like, but how does the environment, change these distributions what kind of distributions, can we expect in different environments. And so for that, again, you need data. And so we took 14:06:58 a we isolated 20 soil, bacteria, and we fluorescent be labeled all of them. And so we have a labeled and unlabeled version of each species. And we measured all pairwise interactions between them across 40 different carbon environments. 14:07:16 So these are 40 different carbon environments, they're composed of different types of carbon sources mono saccharine and sugar alcohols amino acids and so on. 14:07:24 And, and some variations in concentration as well, but mostly just different carbon sources. And so if you want to measure all of these so all interactions between how species a effects VCs be across all carbon sources with enough replicates to actually 14:07:41 infer things. This requires about 160,000 measurements and, which is quite not. And so we teamed up with Portland is left over at MIT, and the recently developed. 14:07:56 This is nice. device called the KHF, which I think also had the pleasure of getting to work. And, and I don't have time to go into the details of how it works but suffice it to say that. 14:08:12 What it enables you to do is to take a library of species, and in carbon sources in our case, and compounds like you encapsulate them each one of the species or carbon sources into the same nanometers size, color coded drops. 14:08:44 actual device this key chip, which has etched into it is the micro wealth and the micro wealth as a god heavy geometry that allows some pre specified number of droplets and to fall into them. and. And this allows you to create many many random combinations 14:08:47 of species or environments and pretty easily. So, this is an image of one particular field of view on the chip when you can see these micro wells. Here they have room for two droplets of combinations of two things, and fluorescence is from a fluorescent 14:09:07 the label and species. 14:09:10 So you can do something like 30,000 of these microwaves on a single chip, you can think of it as a glorified movie well plate. 14:09:19 And, but what it allows you to do is to measure, many many combinations, which is how we could measure so many interactions across so many carbon sources. 14:09:30 And so what we do is we we take first the, the wells that happened to have two droplets of the monoculture of a fluorescent the label species, and then we can quantify the growth of the species, and in the carbon source here and across all carbon sources. 14:09:49 And when you do that you see that here are 20 species, who are the carbon sources or problem environments, and the color indicates how well species grow. 14:10:00 And so we have some, some variability and in how well species growing these different carbon sources. And one thing that's important to note is there's a lot of white. 14:10:11 And so species that just don't grow in monoculture. 14:10:16 And in this will be important later on but I think it's an important feature of our study, and maybe not by design but just because you work with so many species across so many different environments that you do include some species that don't grow on 14:10:30 your own. And traditionally in other experiments like the ones I told you about before when you work with a small set of species in a particular environment. 14:10:39 Then, most people don't include species that don't grow on their own. It's considered wasteful. So they're just not there. 14:10:46 But here we do have a species that don't grow or just grow very poorly on their own. 14:10:52 And so you have this information about how well each species grows on its own in its current environment. And then we can start looking at how well it grows and co cultures. 14:11:02 So again, just one of the strains is fluorescent to label, and we calculate we only look at Wells where it's coupled with some other species it's not for recently label, and we look at how well it through the forest and labelled species grew after a while, 14:11:16 and we can compare it to the monoculture growth. So if we have more fluorescence than will say that they're unable species facilitates, and there's a positive interaction and within the label the one. 14:11:30 And, and, in this case, it grows more poorly in the presence of the species so the species inhibited or has a negative interaction and. 14:11:41 And we can take the since we have labeled the not label versions of all the species, we can we can look at the at the reciprocal interactions. And then you can classify them as competition and mutualism we're both species benefit from each other's presence 14:11:58 parasitism or one benefits of the others expense and the neutral interactions. So for the purpose of today's talk, I'll refer only to these qualitative types of interactions we have of course quantitative information by how much did they grow better or 14:12:14 worse, and it tells a similar story, but but I won't go into it today. And what I will say is that, if you look across the whole data set. And the first thing that you find all pairs of species or carbon sources, is that there's quite a lot of positive 14:12:32 interactions. 14:12:33 So I think oftentimes we used to think of interactions between microbes is predominantly competitive, or negative, and that turns out not to be quite accurate. 14:12:45 And so this is the fraction of the different types of interactions across the dataset. You can see the true mutualisms for both species benefit or indeed quite rare they're only 5%. 14:12:57 And, but the interactions were at least one of the species benefits let these mental isms or exploitation pacifism together right they make up over 30% of the data, so about 30% of the cases, at least one of the species benefits from the presence of that. 14:13:17 So, positive interactions not so bear. 14:13:21 infractions not so rare. And the other thing that. Can I can I just interrupt for a second. Yeah, it's absolutely amazing data. 14:13:32 I guess I don't have a no expectation, like, what, where does your expect, what is your, what is your expectation from there it's a mostly competitive. 14:13:39 Yeah. 14:13:41 like it yes yeah How do I get there. 14:13:45 And you mean on theoretical grounds, or empirical ones. 14:13:50 And either would be fine, I think. Yeah, I think the radio grounds, depends on your assumptions right you can assume that the for example in resource consumer models that all the species just consume resources, and don't secrete them or modify the environment 14:14:09 in a way that's beneficial to others, then everything is competitive, right. MacArthur model or competition. 14:14:18 And then empirically I think many times the experiments that people do involve rich media, and we're all species or the pic species that all grow well, and indeed when all species grow well, you get a lot of competition, I'll show you that in a second. 14:14:36 And then there are previous papers like Kevin foster newspaper where they've looked at them some fairly large number of interactions and they found that, indeed, almost all of them are competitive. 14:14:53 And can I ask you. 14:14:55 Yeah. 14:14:57 Sorry. 14:14:58 So, there were a lot of traits that not grow in one culture for certain carbon source. Yeah. When you grow them pairwise, can, can it flip that result. 14:15:11 And can it can it contribute to the increase of the positive interaction. 14:15:17 Absolutely that's most of the story. Yes. Ah, 14:15:23 ok. 14:15:25 Question I was curious if you have two species that are actually the same and you mix them. 14:15:33 Yeah, that's great so what what do they expect is competition. We're ready mix one species the strain that's labeled with the same species of strange It's unbelievable. 14:15:45 So what you'd expect to see is that the, the label strain grows to 50% of what it does on its own. 14:15:53 And in the majority of cases that's exactly what you see. 14:15:58 And there is a small fraction of cases where you you don't get that you actually see that if you compare it to growing the label species on its own with nobody else around. 14:16:15 It has the initial density because there's no partner, then it grows worse than when you provide, and in an additional in initial density of the same species unlabeled. 14:16:24 And we think that might be due to Allah effects, or they work together to break down, or do something else. 14:16:36 How closely related to far apart strains are you talking about. 14:16:41 And, yeah, I'll go back for a sec to the other genetic tree. And so they come from. 14:16:47 Come to families. And so they're not dead file genetically diverse and they're all from two families within the protobuf here. Well I guess I meant in connection with the next question. 14:16:58 You said you mix those ones which are identical except for one of them being labeled or there yeah exactly there. We took the species and we have one strain of it that we just stuck a gap in and another that we did. 14:17:23 And, right. So, And this is maybe not surprising but interactions can be very different across carbon sources so this is the network of interactions nodes or species edges their color Tell you what, quantitative kind of interaction, this is and this is 14:17:40 on one column, this was smelters this isn't proline you can see the networks are very different. 14:17:58 And so knowing that the carbon sources and mono saccharine doesn't tell you much. And what the distribution is not going to be more similar to other motorcycle rides them to say when growing the amino acids. 14:18:14 And so, and what Jonathan do. Yeah. 14:18:16 Can you say something about od, or productivity, instead of interactions, does the carbon source tell you about that. 14:18:25 Yes, yes the carbon source does tell you a lot about that, about the pairs, I guess right. 14:18:32 And if I understand your question so you're asking if I know the carbon source, can I, does it tell me something about the overall productivity, the density. 14:18:43 Yes, absolutely. 14:18:50 Okay, maybe skip that as well and go straight to what I think is the most interesting result. So, as I said, these are these are very different, and but if you stare them a bit then you can find it. 14:19:04 Actually that there is a pretty strongly regularity. 14:19:08 And, which is what I haven't told you is that the size of the node, representing a species is how well that species grows in modern culture on this common source. 14:19:19 And so very different species grow well in monocultural mottos than on proline. 14:19:25 But what's conserved is if you notice the red edges, and the ones that represent competition, they tend to connect species with large nodes. So once they grow well on their own, whereas the, and the purple. 14:19:43 The our citizenry location lines they tend to connect large nodes with smaller ones. So species that grow well with species that don't. And that's true in both networks right which species girl well or not difference a lot between the carbon sources but 14:19:55 this pattern is the same in both of them. 14:19:59 So, to quantify this a bit more what you can do is now you can you can break down your data into looking at in pairs of strains and carbon sources were both strains grow poorly. 14:20:14 And then when they start growing better and better. So if you look at the carbon sources and periods that grow poorly, and then most of the time you get mutualism, so both of them don't grow on their own, and both of them don't grow together, so they 14:20:26 don't affect each other. There's some positive interactions, but not many. 14:20:31 And as you start looking at the situations we're both strains species grow well on their own in one culture, then you get a lot of competition. 14:20:41 This is consistent with what I said these previous results we get in the distribution is dominated by competitions that's one species grow well. Again, most people, I think who do these experiments they bias themselves to working only with the species 14:20:56 because it's more convenient. 14:21:01 And then you can you can also look at cases where there's a disparity between the ability of species to grow and monoculture on these carbon sources. So again, we start with the an equal ability to grow, sort of disintermediate equal ability to growth 14:21:16 we do still have lots of negative interactions. And then when we go to the right we start shifting to more unbalanced situations where you have a species the grocery pulling on its own couple of with a species that grows very well on its own. 14:21:30 And now we start getting a lot of positive interactions, most of them as part of pacifism but also some incremental ism and mutualisms and almost no competition. 14:21:43 Can I can ask a quick question. Yeah. 14:21:49 So, parasitism it would be like if I had a crush feeding interaction like the, you have a primary for mentor and then something that requires the product is that. 14:22:00 Is that right, that's parasitism this one mechanism. Yes. 14:22:04 So I wonder if one of these sort of rate yield trade offs that Terry has investigated is sort of suddenly encoded in here. 14:22:12 Oh. 14:22:22 And, sorry, I cut you off I got excited, I'll just say it out to other people follow us any interest that that you're you're showing your showing that the parish where we're one grows poorly and the other grows well on its own, have these positive interactions, 14:22:40 which you know one mechanism for that would be that the one that grows well is, you know, incompletely metabolized in the carbon source such that it makes some product that is edible by the second one, but not, but the initial carbon source wasn't edible. 14:22:55 And so that sort of suggests that growing well has something to do with incompletely metabolized in the carbon source, which is, which is something like a real trade off. 14:23:05 Yeah, thank you. 14:23:07 Thank you for clarifying just, just say about that as well that we, we also have a few cases where we included the same common source of the lower concentration. 14:23:20 And when you look at these they have much less of these positive interactions. Then when did the concentration of nutrients is higher and I think that's also consistent with the same idea. 14:23:33 Another question. 14:23:34 I was curious, there's a kind of really stupid no model that I'm trying to compare this to in my head, because let's say I just drew all these growth rates completely randomly independently. 14:23:47 Then if I got, you know, somebody who got lucky on their own. 14:23:53 You know it can only go down in some other conditions, how do I compare like is there some. 14:24:02 Yeah, such a such an old model, we can make here. 14:24:05 Yeah, I think that would be a nice exercise to do, and I think it's true that if you grow very well you can you can only go down but I think the positive interactions are maybe the interesting ones because they require them. 14:24:23 There is some, some beneficial effect to having another species around if it's all for example this who grows faster any nutrients faster than you will not get any positive interactions. 14:24:36 Jonathan. So, I, I might have missed it but it you say, when we classify them into competition. 14:24:49 If you build them by taxonomic distance or metabolic distance is their enrichment for either of these interactions in those pins. Yeah. 14:24:58 You know, which is negative and positive interactions which could be commenced lives and mutualism. 14:24:59 Yeah, absolutely. But, but I think it's all driven by by this, so. So, in this case, so I guess my question is, and maybe it's sort of touches on Hill said, like, you know, so he sort of extrapolating from this and saying, positive interactions are common 14:25:14 out in nature, but couldn't sort of simply be that, you know, if you take any streets, even if you take the same two clones, and you provide a huge fitness deficit to one of them, and no one can grow much better than the other. 14:25:30 By say knocking out something and you put them in the media and by virtue of eating things from the dead cells, because there's so many of those cells are metabolites that are being excreted you naturally have a positive interaction just because you're 14:25:44 in the vicinity of someone who grows much better in a certain environment, but doesn't necessarily reflect that out in nature is positive interactions between ecologically relevant micro Yeah, No, I think that that's absolutely on point. 14:26:00 So, yes, I do think that what it indicates is that whenever you have a large fitness difference. And then, then yes the species that grows well maybe secretes or dead cells or creates some, some resources that the species doesn't grow well in this environment 14:26:18 can utilize, whether that's true nature or not is a good question but to me, it suggests that also in nature, whenever you find species that can grow well on their own. 14:26:30 And then they're probably relying on on other species that do grow, grow well, and because it's so generic right it doesn't matter what species, it doesn't matter. 14:26:41 At least which common source. And then it's just that it could happen very frequently in nature. 14:26:47 Question about intensity of competition, like, how much can they hurt each other. Is that, is there microbial thermonuclear war can they just devastate each other. 14:26:59 And so, in principle, yes right they can secrete toxins or use an array of other weapons, and so it doesn't all have to be even mediated by by resource utilization. 14:27:11 In our case, we don't know that it's all utilized mediated by resource utilization. 14:27:17 And we haven't seen any cases of like, mutual described destruction or something like that. It could be that some of the competition is mediated by interference competition mechanisms and not by resource competition but we don't know. 14:27:33 Okay, Jonathan I have a question I asked in fact I have two questions so you measured positive and negative interactions in terms of yield. I just wonder if you measured in rates, do you get similar results or have you attempted. 14:27:46 Yeah. So, we have not. And I think one of the, one of the nice features of this k chip system is that you get high throughput, and in terms of many conditions but it comes at the expense of not having good temporal resolution. 14:28:05 and unless you really. 14:28:07 Lower the throughput, so you can look at few co cultures at high temperature resolutions or medical cultures that low temperature resolutions, which is what we chose to do. 14:28:16 So we just don't have information about growth rates, I see. And the second question is that you mentioned right if there's a large difference in monoculture growth in yields as either to detect ignore these positive interactions I just wonder, Is it 14:28:29 because if you have one culture that has very low yield, so even little bit, increasing the little in the solo yo be very noticeable is just, you know, the detection of sensitivity. 14:28:41 So, I don't think so because we do look at these things and account for them but more importantly, we've, we've also validated them, and smaller set of these outside the chip with the, with the economy counting and things like that and it holds a great 14:29:00 thanks. 14:29:01 Jonathan I have one more question to in the story from Josh Guildford and Alvarado, the sort of punch line as I remember it is you've got to sooner mode and growing fast. 14:29:15 So creating something that the entire of actor than zooms opposite Sorry, excuse me, Josh, in my ear opposite. 14:29:23 So, So you got a fast primary carbon source to greater than Crosby's if you break these interactions out by family I noticed you have the same to families. 14:29:34 Does it recapitulate that basic pattern is that the majority of these mutualisms. 14:29:40 Yeah, or parasitism that you're observing, I suppose parasitism is would be. 14:29:45 Yeah, right. So, so it is. It's a bit hard to be conveying these things because more similar or final genetically related species, they also consume similar resources, and therefore their growth monoculture growth is similar. 14:30:07 And for getting these positive interactions you need disparity in growth rates, so that tends to happen between, and, Pseudomonas in intro and. Okay. 14:30:17 So, so we don't know if it's really something specific to to this interaction between these two families, or it's, it's more generic to the fact that one goes well and one doesn't. 14:30:29 I would say that when we look at, even within family interactions in the few cases where there is strong disparity between the growth rates of safe to the moment then we get a similar thing. 14:30:45 Can I ask the question. 14:30:50 I'm sorry, I just want to follow up on a point that you made about like on when the carbon sources lower, there's a lower amount of positive interactions. 14:30:59 And I was wondering if you looked at the same two organisms but just at different carbon concentrations or different sensory concentrations and if you saw any differences, then like the percentage of positive interactions versus like competition. 14:31:12 Yeah, so So, all of this is exactly the same set of 20 species where we could measure their interactions in the higher concentration of say glucose and and then lower concentration of glucose and. 14:31:27 And when we do that again for the same set of species, then the percentage of positive interactions we get is lower, and they tend to be more competitive. 14:31:37 When the carbons concentration is low. 14:31:43 Maybe a couple more questions the cavalry are getting restless for the cook. 14:31:48 Sorry, I'll be I'll be quick, not double down but I'm just. This is really cool data and I'm just really trying to resend, what's the right way to think of it. 14:31:58 So, trying to classify my head things that are really striking and surprising versus things that are kind of sanity checks, especially if you're discussing this in terms of percentage of positive interactions. 14:32:08 So like here, if I'm looking at the plot that's right now on the slide, the last column, right, what are some observations from the last column, there's, you know it's purple and light blue is the dominant color there's nothing that's red. 14:32:21 Right. 14:32:21 But then if I given the two mixed together as VCs that strain that doesn't grow it all and the species as a string that grows a lot. 14:32:30 All of those are, like, those are the only colors you could have there, right like of course a lot of them will be parasitism because the dominant one will only typically go down the one that couldn't grow couldn't grow the only thing that can happen 14:32:44 to it is it'll grow more or still nothing right. So, all of that column, the interesting thing to me is that there's blue the mutualism that although the one that was growing very, very, like a lot can still grow a lot a bit more. 14:32:57 So like what's the sort of right variables in which the root plot this to somehow take out the know the fact and highlight what the striking things are. 14:33:09 I'm not sure that's, that would be my now, because it's true that some things you couldn't have but you could certainly have mutualism in a mentalism so you could have the species that grows very well, just not be affected by the species that doesn't 14:33:23 grow, or, or just killed or overwhelmed the species that grows the tiny bit right so what I thought would happen is that these two colors will be much larger 14:33:39 When you think of species that was extremely women a species that doesn't grow I think what I would expect as the species that grows extremely well just takes over. 14:34:00 Any more questions Jonathan Do you have anything else you want to show, ask a question. 14:34:08 And because this is actually the question was asked before is if you take strange that just don't grow it all. And you couple of them to strange the girl well. 14:34:16 Wait, it just, I just want to ask Jonathan if he has anything else he wants to show cuz his time's running Yeah, no, I think this is this is a good point to stop, maybe I'll just show this one less slide. 14:34:21 So if you and this is the percentage of strains that don't grow it all in monoculture that are now able to grow when you couple of them, strains that grow, and like increasing the levels of, well, and the striking thing is that when you complete non grower 14:34:36 to a very strong grower then in 85% of the cases the non grower is now able to grow. And right so I think this was with us, the beginning what did this, this sort of obligate facilitation happens was important and I think it is dominant part of the story. 14:34:54 And I think that's that's a great place to start. 14:34:58 Okay, thank you.