15:05:08 Thank you. Seems It seems that I'm the only person standing between you and cheese. 15:05:13 So I tried to make this as concise. But again, it's been great, being able to be in this program so I'd have had really good discussions with a few here. 15:05:21 And those of you who I haven't had a chance to speak with you in the next few days. 15:05:28 So, I am a physics of living systems Fellow at MIT for those of you who don't know, and have a broad range of interests in the field. Some of you have heard, Sergei and Martina talk about a book that I've been part of. 15:05:43 But today I'm going to be talking to you about equal evolutionary dynamics, and what role perhaps strains may have to play in him. 15:05:53 So, a lot of us here are interested in studying the equal evolutionary dynamics of complex diverse communities as they are in nature or as they might be brought to the lab, and there's sort of two pieces of conventional wisdom that I think guides most 15:06:09 of us to not all of us when we're thinking about the dynamics in these communities. 15:06:16 First is that usually evolutionary changes when they occur in these communities occurred in single individuals, and these single individuals are usually part of, you know, strains within some species and but usually when we think of the ecological dynamics 15:06:34 of these complex communities. 15:06:36 We tend to think that they emerged from the interactions, you know, between individuals but we tend to think of them, occurring between groups and conventionally, whether it is that we're making models or whether whether it is that we're doing our analysis, 15:06:51 our analysis, we think that this happens at the level of, you know, like a more concrete example of this is that when we are writing consumer resource models, a lot of us will write, and one dot and will say, this is the rate of change of species one 15:07:04 for this is the data species bar. 15:07:08 And this is usually okay, but I would say that there's some tension here which is that there's something according to the level of spins and something according to the level of species, and maybe it is true that, that one is to ask why this strange also 15:07:24 play a role in in ecological. 15:07:28 So, I would say that the reason why we haven't used even thinking of doing this, it's not so much a reason of complexity, but it's because there's very little that's known about the dynamics of strain and their interactions and complex community. 15:07:41 So that's sort of the gap that I'm going to try and fill in this talk today. 15:07:46 So I would say that, you know, what does this really mean we need to think of some sharp questions. There are various degrees of sharpness that these questions can arise and privacy the overarching broad question, to my mind is, does this find scale strain 15:08:03 diversity that Daniel is talked about many other things, think about does that influence ecological interactions, especially in complex diverse community. 15:08:12 And I would like to break this up into three questions. 15:08:16 But maybe I'll take a question before that. Yeah. 15:08:28 So we want to stress be in the same species therefore subject to a lot of recombination however ecologically different, or is it more in the in the old fashioned sense just anything that that was isolated we just call it a new strain or yeah some of that 15:08:32 clarify what you mean by stream here, either because we introduced species so Australia mean anything. 15:08:45 say yeah so that is an excellent question what is a strain what is a species is a sort of little bit loose. 15:08:51 I would say that during this talk and a little, if you give me a little bit are defined precisely what I mean in this talk with strange, but to sort of satisfy your curiosity for now I would say that, basically, I can end at the end of the day, whatever 15:09:04 isolated from, from my communities I can make a tree out of those. And basically the deep, the sort of shallowest levels of this tree I will call it strange, but as I sort of coalesce the more and more I can think of species and G and Canada and so on. 15:09:18 I will define more concretely what I mean by species and strains in the context of what I'm talking. 15:09:25 In, but to give you a teaser, I'm going to think of as vs species, and we can argue about that. And I'm going to be talking about strange which has the city, which, which are the same 16 as sequence the same ESB as strength. 15:09:42 Okay. 15:09:44 So I look back to the broad question, the strain diversity influence ecological interactions, and I'll break it down into three questions that I'll try and answer in the storm. 15:09:56 The first is well before we get, get to strain diversity influencing interactions. We need to ask, Well, is it really true that multiple streams within a species persist and coexist for long periods and communities, or is it just that when you know there's 15:10:10 some evolutionary change there is a new strain that comes up and just quickly fixes or goes nearly to fixation. 15:10:17 Maybe it's not so much of a problem, problem anymore so this is one of the first question that we need to address. 15:10:23 Sure, there is, there are strains, they persist and coexist for long periods are the next question that we can ask, is our strains within species actually doing anything different, or they ecologically distinct for the ecologically equivalent. 15:10:41 And there are many ways in which you can think that strains are the same or different. 15:10:46 I'm going to be focusing on the dynamics of these strains. 15:10:57 And in principle one, one can also ask, Is there any predictor that we have for when strains are equivalent of an strains are distinct, for instance, is there a characteristic genetic distance beyond which strains start to display distinct dynamics. 15:11:06 That's question number two. And then finally, question number three which is, let's say that strains are actually just distinct strange but within species, showed this distinct dynamics to these distinct dynamics between species indicate distinct interactions 15:11:22 between strength. And by that I mean, if are the interactions really defined at the level of strains if that's true, then, one might see for instance a strain in a species interact very differently from a strain of another species then with this one part 15:11:41 part. 15:11:41 And all these three, three things will become more clear and precise as I go through my talk, but he's a sort of broad questions that I wanted to put out. 15:11:49 And I would say that the, I would argue that the key reason why we sort of don't have the answers to these questions. Till now is that we're really data limited in this respect. 15:12:01 One can go out and study national communities, but really international communities if you're investigating strain dynamics for instance if you're looking at the human gut microbiome, there's all sorts of things that are going, going on this sort of post 15:12:14 post effects and so on and so forth. So that makes it really difficult to disentangle what is really happening because of interactions between Spain's, and what might be happening to all these expensive. 15:12:27 So, I would argue that what is really needed to address this question is that we need to bring fairly complex communities to the lab that has been to domesticate them. 15:12:37 And then we need to grow, grow them and control conditions and track the dynamics of strange species what what have you. 15:12:45 So this is what we try to do. 15:12:47 I should mention at the outset that this work was done in auto cord arrows group with two of his former postdocs Leonora Bertelsmann, who is now at Boise State, and give you 11,000, was not you who is now word farmer biome. 15:13:06 And so here is what we decided to do. So, like I said before, our approach was to domesticate communities complex communities and bring bring them to the lab, and then track the dynamics of strange but then we did this in two steps. 15:13:23 The first step was actually done by Leonora in this previous paper, paper, which is to actually domesticate the communities and bring them to equilibrium. 15:13:33 And here is what she did. 15:13:35 So she went to Harvard forest. She really likes pitcher pitcher plants, and she took. 15:13:43 Now, you know, I can tell you a whole bunch about future plans but the key thing that you should know here is that these pictures. The, these, these sort of beta clients have these sort of modified leaves in the form of pictures, and they hold this particular 15:14:02 acidified fluid. They have nectar on the surface of these sorts of pictures which attract praise such as insects, and so and so they come and then this sort of fall in and there's a whole Mike microbial community that forms inside this picture fluid that 15:14:19 helps the plant digest. 15:14:22 These this insect pattern. 15:14:25 So okay, we have 10 different plants. They're from the same ball. She took out the pitcher fluid from each of these plants and brought them to the lab. 15:14:37 And then she put them in a medium in one is one dilution in what one is to one ratio. 15:14:45 One part of the pitcher fluid from each plant, and one part of a medium made of essentially acidified water and ground cricket powder, and maybe now that you know the these in these pitcher plants these microbes are really digesting insects such as liquids, 15:15:00 it might make some sense as to why we actually wanted to ground ground crickets, you can just buy these off Amazon. 15:15:10 In, in each of these sort of buyers there's about a 50 at ground cricket, if in case that information. Interesting. 15:15:21 And okay so she had a ton of these different parallel serial dilutions running at each stage, every three days, he would take one part of the community growing, growing inside this medium and transfer it to fresh pitcher pitcher. 15:15:36 This fresh medium. 15:15:38 Yeah, and she kept doing this for 24 for 21 transfers, that's about 21 generations. And after that. 15:15:47 She performed 16 s, and click on sequencing, to track the dynamics of PSP within these communities. 15:15:54 And this is what she found. 15:15:57 Here I'm showing, or 10, or 10 of those communities because they were 10, different plants are these are stacked bar, bar plot showing the relative abundance. 15:16:08 These are stacked BB BB plots showing the relative abundance. Over time, this is over 63 days, there are 21 transfers every three days, and each color here represents a different ESP. 15:16:17 Yeah. 15:16:22 That is an excellent question. Um, that's exactly what I was getting to for the rest of this talk, I'm going to define each of these as fees, that is, each exact 16 s variant type as a species. 15:16:35 Okay, and strains are going to be two or three distinct isolates with the same 16 as sequence that is the same as is. 15:16:46 So what you're seeing here is essentially what I will call the dynamics of species. 15:16:53 Yes, question, how diverse are this ASAP so again you for your practical definition I'm okay with it but are they have the same sort of the Geno's have the same so that is an excellent question they're quite diverse actually spend several Fila, I'll show 15:17:09 you a fighter genetic tree in a moment, but they can be quite diverse as beta, the sort of beta pro to bacteria gamma protocol bacteria, back, back to the IDs, and so on and so forth so it's not that they belong to a particular genes, per se. 15:17:25 Second question is how confident. 15:17:29 Can we be that the features have been eliminated in this model that is an excellent question. Um, so we cannot be fully confident that phases have been eliminated. 15:17:38 So we cannot be fully confident that phases have been eliminated. I forgot to mention that in this process of taking the picture fluid, or two, or to the lab. 15:17:45 There's, there's, importantly, all sorts of things in this picture fluid there are protozoa there are there are rotifers and so on and so forth. So one step that Leonora did was to filter to filter this fluid through a three or two or three micron filter. 15:18:01 And that is usually, we hope, only let's back bacteria through but it can certainly let phases through. 15:18:09 What I can tell you is that I have additional evidence to think that there that pages aren't doing something important here, because we eventually sequences communities using meta genomics and I look for pages and I didn't find any, but it's a good point 15:18:24 that this protocol doesn't necessarily eliminate page. 15:18:30 Yeah, I know it so standard to use relative abundance as the quantity of Matthew use absolute abundance would you see a very similar trend, like, is that. 15:18:40 That is a good question again, I thought I like. 15:18:43 So, Leonardo did quantify the biomass of these communities and they were fairly similar. So I would say that one can think of the relative abundances as being indicative of absolute absolute abundance given the total volume of these communities. 15:19:04 Okay, I'm ready. 15:19:07 Yeah. Yeah, I got to that. So one more of these. 15:19:12 Have you checked questions. 15:19:15 What do you mean by that, have you. That's an excellent question. It doesn't look like they really reach some sort of steady state there. So, what is it about you. 15:19:26 Yeah, so. So, by incremental be limited to two things to look at a equilibrium. One is that we looked at NMDS plots and I should really have that and MDs plot here, which basically shows that if you imagine this MDS on some space that all of these communities 15:19:43 start from some faraway point very quick, quickly reach, reach some point and then this this keep wiggling around. 15:19:51 The other piece of evidence that we have for it is that we looked at the breakout is the similarity of one of these communities over time and to be to see them become roughly, same over time. 15:20:02 So in that, too. If you were to do these experiments for 10 more days. You're saying that it's the last day that it's 6363 so if you want to do 10 more days. 15:20:15 You're saying that you probably, I show you what it looks very similar to the 63 both would be very different from the zero day like era in relative terms. 15:20:24 Yes. And that's what you then correct and i and i show you what it looks like in 10 and 10 more days and, in fact, I'll show you what it looks like for 300, more, more days. 15:20:33 Okay, thanks. 15:20:36 Okay. 15:20:38 All right. But this is an excellent question of how are we know the dirty equilibrium, I would argue for now that we know it because we measured break record is the similarity and we looked at the MDS plots, kind of look like there's an equilibrium but 15:20:53 but we will get to see more of that. 15:20:56 Daniel had a question of what the phrase here mean the greys here are usually things for which we don't have a good taxonomic assignment. 15:21:06 All of the colored ones here on ESPN, for which we do have a good tax economic assignment. 15:21:20 And so, yeah, so, so that there are some cases where certainly you, you can see that some sort of equilibrium has been reached by Dr certain other cases may not be super obvious, though, most of them I would say to my eyes at least. 15:21:27 Look, look like by, by the end they're fairly have hovering around me. 15:21:34 Okay, so, so that was Step Step one, bring complex communities to the, to the lab. 15:21:41 Get them to eke equilibrium there about the order of 10 to 20 SV is that sort of coexist in these communities. And so now we wanted to do step two, which is to start tracking strange, strange dynamics, Eddie, Eddie equilibrium. 15:21:57 And so now I just take one of the communities as an example, here's a, here's a microcosm six. For instance, we had to remind you 21 transfers that were done every three days at a oneness to one delusion. 15:22:13 Now we want to track these communities at this news at this new steady state for a much longer time. And so, and so what we did was that we changed over to tracking them for 45 additional transfers. 15:22:29 But instead of diluting every three days to convenience, we started diluting them every week, every seven days. And we change the dilution factor of our dilution regime to want us to 100, to allow more and more generations to pass but between every, every 15:22:46 student and this is about one, one generation. This would be about six or seven. 15:22:54 And you can see in these communities that basically, even though, you know, things, things are changing over time at the level of, as, as we sort of, you know, nothing super dramatic that's going going on. 15:23:09 Things are going up things are going down, things are going up things are going down the sort of some sort of dance that is occurring in these communities. 15:23:16 And so these are the communities that we have 10 of these to remind you that we decided to track. 15:23:24 In terms of strange, strange Dinah, what do we do for Shane dynamics, we can't just use 16 s, and click on a sequencing. 15:23:33 Um, so what we did was that we also did shotgun meta genomic sequencing at eight evenly spaced time points. Through these 45 transfers. 15:23:43 And so, if I focus on those eight different time points. I can show you the dynamics of one of these as these are species. 15:23:53 Or, for the rest of the stop here. Here I've shown you ESP 27, which is a, which is of the genius sila fellas. Here I've shown you what we call as the 14 which is Gina Pseudomonas. 15:24:06 And now when the six year. Now with this shop shotgun method genomic sequencing, we can actually sort of pure under the hood and see whether or not there are any strains that are underlying these species, and I'll explain how I how I do that in a moment 15:24:22 but I just want to appreciate what that can look like. Here are two examples for these two as a space in in each of these as vs we can resolve them we are the meta genome and get the order about two to two strains. 15:24:40 Here strain one is trained to for this xylophone species. And here's phase one and phase two for this Pseudomonas species and sort of immediately what you can see is that these strains are not widely going to go over, it's not like what one of them is 15:24:55 widely going to fixation and the other one is sort of going away they're sort of persisting and coexisting for long, long periods. And so this is going to be now the meat of how we're going to be able to track strain in Dynamics. 15:25:10 And the other thing that I want you to look at is that, depending on which species we are looking at the behavior of the underlying strains can be quite different. 15:25:19 So for instance here in this xylophone as you can see this strain to the sort of minus train is almost static and abundance, not really changing. 15:25:28 And if you compare that to this other strange statement that's fairly, fairly dynamic compared to the two. 15:25:37 If I instead look at this, this other species, I can see that both of the strains of quiet time and this depends on how you normalize the rate because if you look at the second one in the fraction of the total species is constant, basically, if you look 15:25:53 at the red graph right string to is a constant fraction of the species, what is the other one to normalize by the abundance is actually changing more yeah yeah do you make that decision. 15:26:10 Um, I would say that you could choose to look at, to look at it both, both ways. One is one is to look at the frequency, the frequency of the two strains within the species, and the other is to look at the at the actual strains is two different units. 15:26:22 And both of them would be valid ways of doing the analysis for now I'm looking at the two strains is different and, and just the very fact that you have a father of two strains is is interesting right yeah mines them for instance when he looked at the 15:26:40 slide away from me. 15:26:42 He didn't see one. Absolutely. So, like always do i mean this is how you select them. 15:26:51 Give me a minute but yes it's almost all always one or two. 15:26:51 If you look at these species which are persisted in the original community. 15:26:57 Do you also only see two streams. 15:27:00 Are you asking about the original picture pool. Yes, that is an excellent question. I talked to the best of my knowledge, I have not looked at it and I don't even know if the sequence to sequence that marriage. 15:27:13 But that is a good, good question. We don't know how many strains there were in each of the speech of foods, would you presume we're selecting on some of the strains versus others in yeah was leading that that is a that is a good, Good point. 15:27:25 What I will say is that what I know from Leonora his previous work is in the first three days, or there's a ton of things that die. So if you actually look at the diversity of the fluid at day one, versus a day three or already tons of stuff has died. 15:27:50 So, I'm presuming that the pitcher fluid is much much more diverse in terms of strains than our lab in Britain will be there's two possibilities that one is that it's species that die with all the sub strains in it. 15:27:56 I'm describing species but I suspect that a similar thing is going on for strange Western. 15:28:13 Can you tell us what Pseudomonas species are there, um, that is so so so the answer is sometimes we can sometimes we can't some of the Pseudomonas species are poorly characterize but some of them are not. 15:28:24 So I think there's a sort of manage. 15:28:27 So, in in in here. 15:28:29 But the other Pseudomonas species that don't, I don't. 15:28:34 Do you see particular species become dominant in across replicant communities. 15:28:40 SB. 15:28:47 Know that usually doesn't happen like I was saying there's not a lot, very dramatic that goes on here. Usually these communities are sort of big wiggling around and compositions. There's not very dramatic x extinction events that are going. 15:28:59 I think we should just pass this around. 15:29:05 Oh sorry, across across communities. 15:29:09 No, that's not true. It's not true that any. In fact, in these communities we don't even see family level can convergence so it's not even as if we're particular families are always consistently dominant. 15:29:23 Sorry I misunderstood. 15:29:25 Yes. 15:29:26 So in order for me to appreciate what it is that you're showing here and significance of it so the right most of figure. 15:29:37 I really need to know what it is that you mean by stranger how you measure it, you said shotgun. But yeah, I, how do you, how did you do it because it is already the highest resolutions you get 16 s level, and then you do shotgun Did you construct, managing 15:30:01 I wanted to show you something provocative before I showed you how I do it but maybe. 15:30:07 But maybe now you're interested enough that, that, that you can see right. 15:30:09 them assemble genomes and then do whole genome comparisons, how did you get to the strains, give me two slides and I'll describe exactly how I did. 15:30:15 Okay, so I hope that you have picked enough to see how we got this and what we want to do with this but this is what we have we have this sentence for all of these things are now we can begin to answer the questions that I laid at the beginning of my 15:30:30 talk so he has an outline of what you're going to hear. 15:30:34 The first thing that, that, that we find is that even highly related strains, ensure very distinct dynamics. 15:30:43 And it's going to be through, through this plot is going to make a lot of sense. By the end of this. 15:30:49 Second thing that we found was that actually the genetic distance between strains of the same species can be a fairly decent predictor of whether or not these strains will show distinct dynamics, or whether they'll be copper, and that's beyond a characteristic 15:31:03 Stop. 15:31:14 And the third thing that I'm going to show you is that even strains in the same species can show sort of stronger more stronger interactions or more coordinated dynamics, with strains of completely different species often in completely different Fila 15:31:21 distance about 100 snuff. So just about 100 snips seems to be enough to be coupled strains of the same species in terms of their time, mix. 15:31:30 than with their own closely related part, but you are you going to tell us like 100 snips relative to what the whole genome the core genome. 15:31:40 So this is just going to be a plain absolute measurement of the number of set. Okay, genome wide. You know why this is called core genome like recording. 15:31:49 Okay, okay. And what about number of genes and the accessory genome difference in your Do you mean to ask, how many genes these hundred steps are spread across, no I mean for the strains that are decoupled versus coupled how variable are they in terms 15:32:06 of their accessory genomes. Okay, that is an excellent question. So it turns out that we can't we don't have information about the accessories you know because we didn't actually sequence the individual strengths. 15:32:16 But the accessory genome domains behind the curtain. 15:32:22 Um, how many data points you have in the upper left. It looks very small is so this is smoother. 15:32:30 But we have the order about 300 points. 15:32:34 Okay, um, alright so let me try and answer stallions question. 15:32:41 So the way that. So, like I was saying, we resolve strains by looking at snips within these metagenomics reads, let me describe how we do it. So here is sort of one one community where at eight evenly Thumper time points, we're sequencing the meta genomes 15:33:03 of what what we sort of individually did was that we got across all of our communities, 34, isolated genomes from from the same communities, which beard sequence and assemble into genome so these are not met a genome assemble genomes these isolates that 15:33:22 are derived from the communities and then sequence will assemble. 15:33:26 Now we have sort of reference genomes, as sort of a reference genome library for our communities. 15:33:32 That is what we use to map the reads that we get from our from from our magazines to these reference genome. 15:33:42 So, this is a cartoon of what what that looks like here's you know one, here's, you know, here's a bunch of leads that pile up on on to them. There's a whole set of problems that we need to solve to make sure that if a read maps equally well to two or 15:33:57 more genomes we throw it out if you read maps to a gene that is found with a very high identity across two genomes we throw it out as a whole bunch of things that I'm not going to bore you with of of what we do to make sure that we are mapping leads uniquely 15:34:13 to these genomes. And that's going to be okay. 15:34:17 In, in the end I have a bunch of leads that have that are piled on the genome one, a bunch of leads or a file on the genome to and so on and so forth. 15:34:26 And then what. 15:34:29 And then what we do is that we're going to look for snips at particular positions in these gene genomes. And remember we had eight different time time points. 15:34:38 Oh, I, so I can actually now make trajectories of these snips or each my genome. And so here are our two exam. 15:34:48 There are two examples. 15:34:57 What, what we often find is that on the y axis is the Allen frequency of the snip that is at each position, how what what fraction of reeds do I get that map the alternate genome versus the, the, the alternate base versus the reference base, and on the 15:35:09 x axis I have time, I have eight is distinct time points, and each line here on each project to rip represents a particular. 15:35:32 And what we see here is that it's not like all of these snips are doing their own thing that they represent different new mutations, they almost always appear to move very tightly together in clusters that are there are some examples like I was showing 15:35:34 of strains that are very very dynamic, but there are also some examples of strains that are relatively static and fee. 15:35:41 So, these are sort of two examples of what the strain trajectories Look, look, look like. But the key thing for us to take take away is that virtually all of the snips close to about 98 or 99% of snips that we see are of this type, they have two features. 15:35:59 First is that most of them have a non zero frequency at the first time time point. They're not de novo mutations that arise during this experiment. 15:36:10 But at least we don't think that they are. 15:36:13 And the second is that they're not present in all sorts of different genomes this whole sort of a pigeonhole principle that you can apply here to say that are these, these sort of are all snips that are linked on one. 15:36:27 And so the influence that we make is that these snips actually belong to a few distinct genotypes or strains. 15:36:34 And this is what I'm going to mean by strain when I go on here, and a follow up question that we can ask is how many of these genotypes are strains within each as be your genome twice. 15:36:46 And this is the question that Sergei was asking. 15:36:49 And the answer to it is that we can usually resolve up to two strains. With each genome or for each species. 15:36:58 And there's, there's something to be said about the fact that we have. Each time time points and all sorts of classic sequencing errors that that create sort of wide bands that limit resolution. 15:37:12 So, here's sort of another example of what these snip, snip clusters Look, look like again on the y axis is the earning frequency of on the x axis is the time in days here. 15:37:27 And so what you can see again is that often when we do see another cluster, that's sort of here and re, it's sort of almost all nearly two fixation. 15:37:39 And so the inference here is not that there are three distinct strains, but they're actually again. Two strains that we, we can resolve. 15:37:47 One of the strains as has sort of this frequency of this cluster and the others are all sort of what I call a fixed fixed snips, these are sort of telling you that both of these strains have this change, relative to the reference. 15:38:03 So for instance, maybe my reference that I sequence from from some community had a parity in a particular position, but both of my strains of this position have a. 15:38:12 And so those. All of those are going to show up as snips, but all of those snips are going to show up at a frequency of nearly, nearly one. And so this this this characteristic is generic for all of the strains that we observed for all the snip clusters 15:38:27 that, that we observed, and that is what leads us to make the inference that there are between one and two strings, the genome that we can talk. 15:38:36 So, like Sergei was asking this, this thing is not unique to what we found. This is actually seems to be now a fairly generic thing, that, that, that, that others have found. 15:38:53 And it's actually a puzzling to me, I don't know the answer for why why that is, but what I'm showing you here is data from the Lenski lines, which again are are very different from these picture planned communities they were started from an Isagenix 15:39:08 eco life strain and and what you usually find is again by tracking the frequencies of various snips through time, that most of the times are these populations cluster into two into two distinct creates, or you're the authors call it a major create and 15:39:26 a minor clay. And then these claims processing coexist for long periods of time. And again, almost all the times, but not all, always. 15:39:35 There were two persistent lineages. 15:39:38 This has also been seen in the human gut microbiome and Ben was talking about this a couple of weeks ago, there is usually between a small number of strains between one and three strains, they're able to see in in the samples. 15:39:53 Some of Tami's work is also Sean. Sean this again in the, in the context of the human got Mike microbiome. 15:40:00 So I think this isn't interesting puzzle. I don't have an answer to why there's a really small number of strange. 15:40:09 But, but this seems to be a fairly generic, I would dare call it a phenomenon that I think may need an explanation. 15:40:18 Think of all of these, the sequencing depth, not that deep or what you can resolve right. And so if you have if they tend to be spread over a wide range of abundance is, you're likely to see one or two. 15:40:29 Yeah. Correct. So, it's true that things that are at very very low frequencies let's say a percent or two to just not be able to resolve them from noise or to see sequencing errors, but I guess I'm more sharp version of the claim would be that it's very 15:40:46 rare to find two or three strains that are at relatively high frequency, let's say, one at 21 and 30 and 150. 15:40:56 That still seems to be, to me, I think that we should be able to resolve from from these from these trajectories but it's very rare that we are able to see three and almost never I've seen something that's forum. 15:41:09 Just showing the Lenski data there there's one least one case of astronauts I'm straight in measurably low level. That's very long time and then comes up but with enough. 15:41:19 Yeah, yeah, for that is one case in the Lansky lines where they have three different plates and one of them eventually come comes up but like I said that's only one. 15:41:33 Do you think you can make a sort of parallelism so also the species in the communities you have one or two they're very abundant and others that are very rare. 15:41:39 and it's kind of the same thing for strings. 15:41:42 That is a good point on I would say that yes it's true that in many communities, you see a small number of species dominant but that doesn't seem to me like a generic phenomena for instance in these Pitcher Plant layer like communities. 15:41:56 There are many many as bees that are roughly 10 10% frequency. 15:42:02 Yeah, but it could, it could be that and it could be that we're fooling ourselves. It's really just how much more structure. 15:42:13 But I move on from from this I wanted to point this out as as an interesting phenomenon phenomenon that maybe needs explanation. 15:42:22 Seems to be fairly generic, but ok now we have strengths. We have species. 15:42:29 We can delineate both these strains, just by looking at the frequency of one of them. I'm going to think think of them as a major and a minor strain may may may be in some, some cases but strain one and strain two is what we're going to track with every 15:42:45 cluster. I can sort of measure. 15:42:47 A coverage weighted snip frequency as the frequency of that strain. And that's what I'm going to use for all of my communities, going, going forward. 15:42:58 Okay, before I get to the questions I also wanted to point point out that once we can resolve these strains. We can then go ahead and ask what is the average genetic distance between these strains Is it very large, is it very small. 15:43:10 And what we usually find is that it actually spends many orders of magnitude, which is to say that there are some strains of the same species that have a very small number of snips that are differentiating them, but then some, some others which have more 15:43:25 than 10, and thousand snips are differentiating them. And it's important to point out that all of these persistent coexist in these communities for close to 300 to 400 generations, sorry I think I forgot to when you, you did like some strange isolation 15:43:41 earlier on and you said you made a meta genome to them so can you also measure distance beyond just snips or, for example insertions and different between okay that is a that is a good question. 15:43:55 Oh, we can and we try to but again we found that the insertions and deletions have very very rare. 15:44:03 Close to maybe one or 2% of the, of the changes that PC, compared to the reference are actually in social authority. 15:44:13 And I would say that that's surprising, we were actually initially expecting to see maybe a fair bit of insertion delusionary the combination. So just to clarify, then to you, you isolate, let's say two different strains, you do a method genomic assembly 15:44:26 and you're saying, and then, what are the reeds can be assigned to that meta genomic assembly. Okay. Yeah, I can describe it again. So, um, so we have for each species. 15:44:47 Let's say I have one genome that I've isolated in seek sequence. Now, now I have all of my leads that come from my entire community. 15:44:52 Now I can ask which of the leads that are in this community map to this one genome. 15:44:58 Yeah. And that is what this card, cartoon shows this genome one is the is the single genome that isolated from this community and got a genome for, and all of these colored things or reads that I got from my community, not from my single genome that pylon 15:45:17 division genome. And now I can compare the reads that I detect from the community with with with the particular position on on the genome and detect the snip. 15:45:31 What I then do is that I track the snips through time. 15:45:35 And the snips through time Tell, tell me that there are two, two strains, if I'd seen, maybe multiple clusters break break off here, then I would say that they will more than two strings, that's sort of how I'm making the influence of the number of strange, 15:45:50 strains, it's detached from how many strains I sequenced the number sequences just giving me a reference backbone on to which I can pile my 15:46:03 little technical question. So, I've done similar things and I like my experiments that we do this kind of alignments like this whole genome. Get I run alignments I started as a response on the whole genome and for example of a lie all the reads from everybody 15:46:16 like to say a student was referenced, you know, unlike certain locations on the genome will get recoveries just like from like, like somebody totally irrelevant so yeah that correct for this kind of things. 15:46:31 Ah, that is an excellent point that is often something that happens and so we have ways of raising 00 paper that is in your method section. 15:46:42 Yes. 15:46:43 So yes you can read the method section but also Ben describe it a couple of weeks ago that's basically all sorts of things that you do to avoid regions of the genome that are very common across multiple, multiple genome. 15:46:55 You just don't don't want to look at home Julius question. 15:47:12 I would say that the percentage of reads that that's mapped on to our genomes is roughly the percentage of the relative abundance that is occupied by those genomes in this community. 15:47:15 Okay, I saw so yeah okay okay so that is an excellent question. 15:47:24 It's of the order about 50 or 60%. 15:47:29 But you could be missing a lot of strain level diversity, simply by virtue of with which references, you've selected right. 15:47:37 Yes, we could be missing a lot of that, I would say that again the references that we selected span close to span, mostly everything but this great gray, part. 15:47:48 And so, about 50 to 60% of the abundance in these communities, we are able to get strange, strange level diversity for the rest we can't. Yes. 15:48:02 Okay. 15:48:07 Alright so, like I was saying there's a wide distribution of genetic distances between these strains, this will become important in a, in a few slides. 15:48:09 I will move on. 15:48:17 And again, like I was telling in response to a question. 15:48:21 The general or that have strains are very are quite diverse diverse and not particularly limited to one genius or the other, the parentheses, the numbers in parentheses here. 15:48:35 Dr Porter, the number of strains that we could detect across all of our communities that belong to each other. 15:48:40 And so you can see that there's a fair number. 15:48:43 Okay. So, question one. 15:48:47 Now that we have many streams within species and we can track their dynamics. 15:48:52 We can ask whether strains within the same species show similar dynamics across time or whether they show different dynamics. 15:49:00 And here's a card, or tone of what I what I mean. 15:49:05 Say that I have a community with a whole bunch of species these different colored peoples here. And let's see I focus on a particular species, and I looked at the strains within it. 15:49:15 And again, I'm looking at the abundance of overtime. And I plotted the trajectory of both strings of person that says chain one and three and two, I could call them, copper, if the dynamics of strain one are very similar to the dynamics of strain to. 15:49:35 Or maybe I look at another species and I see that actually both the strains are doing completely different. 15:49:41 And so, 15:49:44 and so those are strains that are called DD copper. And so, one. And so one question that one can ask and that and and that is what we asked is how many cases do is give this type of how many cases do I see this. 15:49:59 Yes. 15:50:03 This is a little bit confusing how you plotted that list. 15:50:07 So just to clarify what is going on here for instance. 15:50:11 What is blooded is absolute abundance of both major and minor screens right so it's not like 15:50:20 if it was a stacked I would say that the major strain is is moving in sync with a minor or something like this. So, This is absolute abundance. So, in order to get the abundance of the species I just add up the to curse that is absolutely correct thank 15:50:51 you for clarifying that. Here what i'm talking in these calls are the are the abundance of each strain in the community, ask a question here. 15:50:51 Yeah. Objection, I guess you identified strange to begin with, based on their different dynamics. 15:51:05 The way you identified a new strain was. Oh look at this thing that looks very different from this other group of snips that behavior different from this other group of snips so 15:51:09 Like, I do so we understood it. 15:51:17 now asking whether whether strains exhibit different dynamic from one another, it's a bit already conditioned on them, exhibiting at least certain differences to one another, otherwise wouldn't have to pick them as different strains so that is a good 15:51:32 point maybe I miss misled you a little bit because I showed you these interesting dynamic examples, but I wanted to point out, there are also, it's not common that I identify strange, just when the snip trajectories are moving a very often. 15:51:47 Very often I actually see frequencies of that are fairly static, and then that static cluster sort of stays stays the way it is. And so, that is, there are many, many of these clusters that we also identify. 15:52:06 Could it be for example, in this lower figure that you actually have three strains there. 15:52:13 They all behave the same, but you based on on the way you designed your, your analysis you actually just lump them together as one straight and then, and then you wouldn't even look at this synchrony that exists, but in fact it's actually three strains 15:52:29 moving in synchrony. 15:52:30 Correct. So, this is a this is a good tech technical point that if there are three strains there and if they're all not doing anything interesting I'll see them at relative frequencies a point three, and then all of them will stay at point, point three, 15:52:43 three, and then maybe I'm calling them to two strains but maybe they're three bright, but I guess in that, like, in that case, my point still remains because even if I were to coalesce two of those strains into one. 15:53:08 the exact same thing would the the key thing that's happening there is that the frequency of those strains is not changing over time, they're behaving the same way that this could be changing but all the same like, I mean, I have bounds on how much they 15:53:15 could be chained changing by looking at the width is only let's pick the first figure. Maybe those are two strains actually and they do move in synchrony. 15:53:21 right. You mean 333, just couldn't be three and let me argue what could not be three. 15:53:42 Because if there were three, three strains, then their abundance, you know this is about between point six and point point eight, but the snip frequencies that I see, so if there were three strains and I would say okay one of them is is frequency point 15:53:51 six. One of them is, is a frequency point six, and then one of them is a frequency some something else I already have something that adds up to more than one. 15:53:55 And so, it just, it just can't, can't be more than more than two are here I can argue that it can't, can't be more than two, but here I can tell, that's that. 15:54:06 That is correct. 15:54:07 So, so, so I have some constraints. When I have four strains, and I shouldn't have, you know, and if they're all the same. And I shouldn't have my snip cluster sitting at point four. 15:54:23 Just, just because point four times four is more than one. Like just about, Is there a mathematical way that you're, you're determining the number of strangers this by I. 15:54:30 Oh, so it's a little bit of both. 15:54:33 What I do is that I cluster the snip, snip trajectories using a clustering algorithm, but then sort of a look at them by I. and that's sort of how I'm telling you these stories of what, what happens when I see more than two clusters. 15:54:49 But, but I have to live by to see that almost all the times, an extra cluster look like. 15:54:57 Okay. Yeah, I was, I was kind of doing the same, the same direction just when you're clustering you're not using only dynamical trajectories, because if you're just calculating correlation coefficients across time points, then you can have this artifact 15:55:13 for instance you have two strains one maybes 10% The other is 30%. They will clearly be different bands, but they will have the same dynamic so they will cluster together. 15:55:26 So you, you have to have some sort of, if they're if they're 1030 and if I can resolve them within the band, then I have that, then I would call them to different bands but if, but of course if the smell was too large and I wouldn't be able to tell God 15:55:39 so your clustering algorithm takes into account, not only the relative changes but also absolute abundance of. 15:55:50 Okay. So, let me move on. 15:55:53 So, let me make this more precise, we, we wanted to ask the question, how many times do we see this distinct dynamics and similar dynamics. And so what we're going to do is that we're going to measure the temporal correlation between the trajectories 15:56:08 of the same bit between con specific train strain trajectories that, that, that is strange that are within the same ease. 15:56:17 And this is a broader I'm going going to make. 15:56:22 I'm going to make a distribution of what I call strange, strange coupling. And that strain strain coupling is the magnitude of the correlation between strains of the same species. 15:56:31 I'm looking at the relative abundance of each strain, strain one and strain to and finding the absolute value the correlation with between them. And I'm going to be showing you a distribution of this. 15:56:42 Yeah. And just to orient ourselves. Bad news, close, close to zero, that is bad news which have a really low, low correlation, or what we call the decoupled, and where, and values that are very high, that are close to one are what we call proper stance. 15:57:00 And this is what the data looks looks like. 15:57:03 This is smooth and, as if there was asking me, but we have about 300 different strain pairs that we're looking at. And so, what we see is that there's a clear by, by modality in this this distribution. 15:57:21 There are a whole bunch of strains that have really high cup coupling that are what we call copper, and then a bunch of strains about 20% of strange trainers that are the peak of course you see nice transition anti correlated or only positively on when 15:57:34 when it's very very rare that we see strains that are anti correlated, but those are cases where, for instance, or the abundance of the species stays constant, but one of the strange sort of fixes and that sort of that, that would be kind of this example 15:57:49 that I that I was showing over one of the Spain sort of fixes we believe. 15:58:00 Okay, so, So about 20% of strain parents can show very different dynamics. 15:58:05 We wanted to then ask is there is there something predictable, do we do, can we know something about which strains will leak decouple in which strains are covered. 15:58:16 And so for that. We thought that maybe the genetic distance of these strains remember they span over a wide range wide order of magnitude. 15:58:24 Predictive. And so what we're going to do here is that on the y axis again I have the strange, strange coupling that was on the x axis on the previous plot. 15:58:33 This is the correlation between strange the scenes, he sees, and on the x axis is the genetic distance in the absolute number of snaps between those strengths. 15:58:44 And again, these, these are what the date data look look like each of these points is a different strain pair. And in red I've shown you a moving average across across all of these points. 15:58:57 What we see is that if we look at the moving average, average. There is a fairly sharp decline in the strain strain coupling as you go beyond about 100. 15:59:09 And so, in some sense, it seems that, yes, strains can be decoupled, they can show fairly distinct dynamics, and even about 100 snips or enough to decouple. 15:59:23 And that's and that's just to orient ourselves about point 1% of genome, roughly if you assume some, so some of your profit. 15:59:33 Okay, um, what, what can have happen often Dude, you don't need decoupling often what, what happens that it seems that only one of the strains in influences the species dynamics. 15:59:44 And so again, this is a slightly, maybe can confusing with the plotted given what I was showing before, but on the y axis here is a relative abundance and the x axis here is again time. 15:59:56 And what I'm showing you is in green, the abundance of one of the strains and in blue the abundance of the other strains. The blue line shows the abundance of a species. 16:00:07 And what you can see all over time is that even though this major strained doesn't really move to too much, suddenly this blue minus train grows. 16:00:17 And that's mostly what's responsible for the increase of the, of the species about. 16:00:22 This is what for instance, a pair of dB copper strange looks looks like. 16:00:28 Okay and one, one can also ask, what are those snips What are those roughly 100 200 snips doing. 16:00:35 And again, this, this is an example of of where those snips are sort of look, located. 16:00:42 I'm not going to belabor the point here I I mostly what we find is that the snips are concentrated in either transporters transnational regulators, or certain central government of metabolism related genes like more, more a. 16:01:00 Yes. Sorry. We talked about this before but to echo Marin's question so that's snips in jeans, but you're not taking into account the accessory genome, we are not taking into account the accessory genome that is a big cap carrier. 16:01:14 So, I have a question for the so basically if you went into NC VI and you were able to pull like identify you know strange, like identical strange I wonder if there'll be enough of anything to then look at the accessory genomes and say like is 100 snips. 16:01:25 That's kind of like a timekeeper or distance keeper in a way right yeah yeah so I did, I did. 16:01:32 I didn't look at this. 16:01:34 Typically it. the answer is it really depends from species to species. 16:01:39 But, but there can be some substantial differences in gene gene content of the, of the order of Part Five ish percent i would say, even for like of the order of 1000 snips. 16:01:51 So I would say that it's not necessarily true that the accessories you know that I hadn't. 16:01:59 Perhaps you're, you're getting to this point and building up but I'm going back to one of your first points and one of the early slides in your talk where you were, you pose a question of whether or not the dynamics that you're seeing could be due to, 16:02:13 for example in this differences and selection versus just neutral dynamics and so I've been thinking about this question now for. 16:02:21 I'm waiting to see if you're going in that direction we are going that okay well I'm just gonna sit down, I, I'm not entirely sure if I get that into my seat waiting for the anti. 16:02:33 Okay. 16:02:36 Excellent. 16:02:37 All right. 16:02:38 All right. Um, let's move on to question, question to the next question that we sort of asked was, okay there are strains that are doing different things that can be be copper. 16:02:52 But what about community interactions are they, they're stronger between strains, they're stronger but between species and what can we say, can, can we say anything using just dynamical date, data, the answer of course is going to be strictly speaking 16:03:09 no, but we're going to try and say something anyway, with the caveat that time series was don't exactly give you an interaction. I'm going to keep hammering on this caveat again and again. 16:03:22 But I wanted to put that out also to start. 16:03:26 What is our motivation for doing for doing this analysis, our motivation is that all of these communities that that we have are growing the same control. 16:03:36 A biotic conditions that are in the same medium, same time, temperature, same vials same aeration. 16:03:45 Then, very controlled conditions where other things can really come come into them, and more importantly this, it seems like they're mostly sort of dancing around and equilibrium is there's nothing really wild going, going on in terms of species going 16:04:01 at staying in the wild fluctuations from zero to one. 16:04:05 And so in these conditions. There are two things that could be going going on and complex divers. 16:04:13 One is, if there are ecological interactions between the various mem members that say strains, or species. 16:04:21 Then the increase in one should usually let, let, want to see coronations or, it decreases during increases in the other in a predictable way. 16:04:33 That is, if there are interactions. On the other hand, if there are no interactions, then we should just see random fluctuations, the increase or decrease of any particular species of strain should not be particularly associated with the decrease or increase 16:04:47 of any other particular strain. We shouldn't be expecting to see any correlations between different members in these in these factors. 16:04:58 So, this is going to be the motivation for us to look at these dynamics. And what we're going to do is that we're going to look at dynamics between the correlations with between at this at the state level, and coronations the species level me describe 16:05:28 Let's say I have two different species species alien species be. 16:05:32 I have two strains within each of these bases, a one. A to B one and B, we do. I can measure the correlation with between a pair is across species, but between these strains I can measure the correlation between a one and B one, it will be too. 16:05:51 A one and B to an eight one B, one and important to point out that I'm not measuring the correlations between a one and a two one B one and B to, these are things that are already looked at when I was looking at top, top thing and the decoupling master 16:06:05 question here. 16:06:25 Obviously, very strong interaction, one, one influences the trajectory of the other in the, even in the most important models, you're going to have a P, half face difference between the two, you plot them. 16:06:27 Are you also looking at time blacks. So or Are you hiring them always at the same time points, and the others because, for example, classical consumer resource dynamics okay you have a predator or prey. 16:06:37 You see a cycle, zero correlation. 16:06:40 So that would be an example textbook example of interaction. 16:06:46 Zero correlation if you do not account for the time lag or the face lack. 16:06:50 How do you deal with that here. Yeah. So, that is an excellent question I was thinking what exactly that when I was doing these measurements. So I did look at time timeline. 16:06:59 It turned out that, that, that, that most of the times the time lag was not important, that is, I didn't seem much higher correlations if I introduce your time. 16:07:08 I'm like, and I think that's mostly because between any of any of my transfers there's a significant number of generations that passes there's about 50 generations between any two time or you don't have a bar resolution but yeah I don't have a temporal 16:07:22 resolution to be able to tear it apart otherwise I would love to do. 16:07:28 you're still looking at relative abundance and spend looking at it. 16:07:33 So, if you're not interacting and one of them, and they're just both doing non interactive random walks and there are still correlations right. That is true for communities which have low diversity, that is to say if you had just had a community of two. 16:07:46 Two men, a members and one would go up the other word necessarily go down. But if you have a community with like 20 different men members and that's not necessarily true. 16:07:58 Okay. 16:08:01 So, what am I going to call strange strain correlation is going to be the max of all of these pairwise correlations between feigns across the different these. 16:08:13 And for the same species I can also measure what I would call a species species correlation. Here I'll coalesce, the strains from a from a species so I could call as a one in a and b one and B two and measure their court. 16:08:30 And that's sort of what I'm going to call the species, and can ask the question, which of them is stronger. 16:08:40 I'm going to put, put that on this plot here. 16:08:44 So here so here I'm going to compare these interspecies correlations for every species bird on the x axis I'm going to plot the species species correlation of each pairs of species a&b on the y axis I'm going to plot the strange, strange correlation for 16:09:00 the same pair of species, there is now that I can resolve things at the level of strange. Can I see increased or decreased correlation. 16:09:09 And I'm going to make up, and there's gonna be a whole bunch of points, one for each species pair. 16:09:16 And usually if they're above this x equals y line that is an indication that strain correlations are stronger, and is a below the x equals y line that is an indication that species correlations are stronger. 16:09:29 And here's what the data look look like here, each point to remind you is one species pair in one community. And I have gone across all of the species were all of the communities. 16:09:44 Question is, how are you controlling for the multiple testing serve you take the maximum of four point, you expect the court Yeah, to be higher. 16:09:52 Correct. So I'm not controlling that here, but we also did another, another model where we shuffled all of the strains across species bar boundaries, but at the same comparison of for of having for versus having one. 16:10:07 And that doesn't show the same, the same bed. 16:10:13 So that's how we control the flat, here we are not explicitly controlling for that. So, so it's true that some things are just coming from that but, 16:10:25 Here again, I was saying that each point is is a particular species pair. And what we see is that the overwhelming majority of points about 76% or above the x equals wireline, the influence that we make some lat is that strain level correlations, typically 16:10:51 higher than species level correlations. It suggests that the interactions might be strained dependent, with the caveat that correlations are not interactions. 16:10:59 Question. 16:11:02 Yeah, so they're not they're not Martin gives us something that's close to 64. 16:11:08 That is, that is to say if I shuffled strains across species boundaries and I do the same comparison. 16:11:15 Then I get 6040 and sort of 7624. 16:11:19 Yes. 16:11:21 I think you just answered my question. All right. 16:11:25 Just Just to clarify, I mean, I'll ask you more about this later, but when you do the shuffling Do you read normalize the correlation, you really normalize the fractions per species, or you just shuffling. 16:11:35 I didn't vote relative fractions. When do you do the normalization. I did both, both versions one in which I don't know, or normalize things on until the very end, and one in which I do nom normalize as soon as I shuffled my strains across both both those 16:11:52 cases, I'd never see a fraction as high as I observed it. 16:11:57 Yes, maybe you could just say, like, maybe you said this already but can you say a few words about why the max correlation is the number of interest. 16:12:06 Yeah. 16:12:07 That is a good question. the idea here was that. 16:12:11 Maybe it's the case that with that in species and species be. It's really a one and B two that are interacting and so they'll have a really high correlation, and a two one B one don't really do anything. 16:12:25 And so, a priori I don't know which of those strains is going to be the most interesting one and what I'm most in interest rate is to know whether or not there's any stranger in there that's across species boundaries there that has the highest rank order, 16:12:39 that's sort of what the last. 16:12:52 Mac. This is the max absolute carnage. Okay. Um. 16:12:54 Alright, so this brings us to this question, which was OK, I can actually before I come, come to that I wanted to highlight two interesting regions of this plot. 16:13:06 One is sort of the region here, where there are typically very low species species correlations are very high strange, strange correlations. And this is sort of what we found as a surprise here What do you often see is that there's virtually no correlation 16:13:23 the species level. What I'm showing here is the abundance of these two species of promo factor and delta Delphia. 16:13:30 But if you focus on just their minor strains. Then there. Then, and they have a very high negative correlation. 16:13:37 And so these are sort of what we call hidden correlations bit, that, that, that, that we find sometimes that we would have never seen if we were just looking at the data itself as is or the second thing is sort of this interesting region here, where, 16:13:57 where you see sort of species level correlations being being higher and then in this case sort of strange, strange within a species of kind of doing their own thing. 16:14:04 That's sort of why your species. 16:14:09 But again, dynamical correlations or dynamical correlations, they're not in interactions, and maybe the, the correct way to really address the question of it, interactions is to really do more, more experiments. 16:14:23 In this case we didn't have the option to do do that. But we were still asking sort of which of the sense of like which scenarios, is this dynamic or pattern consistent with. 16:14:34 This is where selection versus new neutrality sort of comes in. 16:14:37 So we asked the question which of the scenarios, could this pattern be consistent with, could we have maybe a mathematical model that tells us what we should expect this pattern to look like under different scenarios. 16:14:49 And what are those. 16:15:07 interactions across pieces boundaries with others. 16:15:09 The other is that oh, this is all just, this is all just random fluctuations, and the fact that you are doing your you're doing things like Max, really it's true that the interactions are concerned with the species level, and you're just seeing random 16:15:17 fluctuations within strength and that then and that sort of this idea that strains are actually identical there are no fitness or selective differences between them. 16:15:34 But what you're seeing is only a result of random fluctuations. 16:15:38 And so these two hype hyper these two hypotheses hypothesis one and a hypothesis to our what we decided to test using a simple resource competition model. 16:15:50 There's no reinventing the wheel wheel here, a lot of you have heard what these models look look like the only thing that we did, did here was to include this idea of having multiple strains and explicitly testing whether or not these strains are different. 16:16:06 And I'm going to the to this of the idea is that there are, let's say a bunch of research resources and our growth growth medium. These are what are there in our ground cricket powder. 16:16:18 And we have species that consists of two. To start, two strains. And these strains have various random preferences for these, these resources. What we're going to do in this model is that we're going to be able to tune the difference in the consumption 16:16:33 rates between strains of the same species. 16:16:37 And that's how we're going to make these strains of the same species different verses identify 16:16:44 what what we then do in our consumer resource model is that we simulate serial serial dilution, much, much like the experiment that, that we did. And the idea here was that of course in a simple consumer resource model once you fix the environment, you're 16:16:59 almost guaranteed to fix the set of species of a set of strains that are going to coexist, you're not really going to have the kind of temperature fluctuations that that we have. 16:17:09 So what we do in this model and this is partly empirically motivated, is the idea that once you settle to a particular steady state from transfer to transfer if you have growth growth medium em in one transfer. 16:17:24 You couldn't principal have a slightly different growth growth medium in the next transfer. And the idea of this is that we're using this ground cricket powder, which has all sorts of different trees resources in it. 16:17:33 And we're adding the same amount of it but maybe sometimes we get a little bit more of some amino acid and sometimes you get a little bit less of it, but on average, the resource medium calm composition is the same. 16:17:47 And so once you start sort of fluctuating this resource medium a little bit, we start to get them temperature fluctuations in the abundance of these communities. 16:17:56 And now we get trajectories that we can perform the same tests. 16:18:00 And what do we do with our two different type hypotheses, or false hype hypothesis, to remind you is that strains are different. There are strange specific competitive and interactions. 16:18:12 And I represent that you're with a with a consumer, with a consumer consumption made matrix or each entry or each row here is a different species or a different strain here species as a strange name. 16:18:29 They want to do is be screen strain B one and B to. 16:18:34 And in each of my columns I have the consumption rates of each Midas resources. I haven't drawn, all of the resources here and principal of the order of 10. 16:18:44 And I'm going to live in sort of net, net land here, and I'm going to assume that there's a fixed enzyme budget for all of for all of my strings. This allows all of them to coexist and since I'm not really interested in color in their coexistence but 16:19:00 in the temperature fluctuations and coronations this sort of saved. Saved My Life. 16:19:05 Okay, so what am I going to do, is the straight to make these strains, different or the same. 16:19:11 I'm going to say that each that that each species, sort of vector here of consumption rate is completely random. 16:19:21 But for strange EA one in a way to, there's a universal distance de that there, that I have that tunes the Euclidean distance between strain, a one in it. 16:19:35 This is, this is to say that if I have stream new one here in my resource base train it a two can only be a particular distance du, and this distance di, I can tune. 16:19:49 And the second hype hypothesis will be one where the interactions are species specific restraints of the same species will be identical. And so that's sort of shown, shown here. 16:20:00 And the only thing that's going to happen is that once I simulate the trajectory for a particular species I'm going to add random stochastic fluctuations between these, these random, or stochastic fluctuations the magnitude of them comes from work is 16:20:16 sort of similar a similar distribution to what we measured in an experiment. 16:20:21 Okay, so what what happens when, when we do, do that. The first thing that we see is that by tuning this distance between strange di can actually recapitulate the decoupling but between strengths. 16:20:35 What do I mean by that remind me that a few slides ago had shown you, as part of the strain strain coupling as a function of the number of snaps of the genetic distance between strains. 16:20:46 Now I can make a similar plot but with the strain strain coupling versus this competitive distance be between strains, one here will wrap, well, so all this is normalized to the average distance between any two random vectors that is the average distance 16:21:01 but between species. And so, stuff that smaller and smaller, is all measured and ready for that. And these are what the data from the model of look like this sort of a similar deep decline of the strain strain coupling with this competitive distance, 16:21:17 the farther and farther to strains become in terms of their phenotype, the less and less coupling between them becomes. And just, just to remind you this sort of looks a bit similar to what we signed the date data. 16:21:33 And now we can also ask about the scatter plot of the strange, strange correlation species species correlation that we look, looked at in our two hype hypotheses. 16:21:44 In a minute hypothesis one was when strains of this distance and there's a whole variety of distances that I have in my, in my model, much, much like in the XX experiment some of the strains are 10 steps away some of the strains are 100 snips away some 16:22:02 And when I heard that in this is, this is the pattern that I expect to see in the model. 16:22:10 I can also do that for hypothesis to where my strange within the species were identical, but I add random noise simulating stochastic fluctuations within my strengths. 16:22:22 And this is what the expected temporal patterns should look look like I'm sort of something clear here though maybe it somewhat looks looks like cheating, which is that more often than not, Bs motley supporting the idea that when strange or different 16:22:44 phenotypically. 16:22:46 We will often tend tend to see strange, strange correlations that are higher. And when strains are identical, and you have you have essentially only random fluctuations but between them, then you can, and to see kind of the opposite. 16:23:03 There's a I have about 1010 minutes so maybe I will zip through, through this, but there's sort of a simple Jim geometric interpretation of sort of see how this arises and the model and resume geometrical interpretation sort of relies on this idea that 16:23:18 there's actually high dimensional things simple acts of these resource consumption rates on which each strain or species six. Now for some simplicity I've made things on only for three three resources button principles as a much higher number of these 16:23:34 resources. But again, the closer you are to one of these points the higher and higher consumption rate for at least. 16:23:41 And so, this is sort of what the model says, each strain is sort of a point in this resource space. This is let's say strain a, a one. 16:23:53 I cannot draw a sort of circle around this strain of a one of radius D, and any point on on that circle is a possible concept called specific strain possible a two, if you, if you will. 16:24:07 And now I can make. 16:24:10 Now I can make another species exactly like this, another species starts off with a completely randomly chosen point B, be one again I go a distance do way, and pick pick randomly a possible BB to an OK now I have four four of my friends, even a to b 16:24:29 b one and B two, what do I do when I'm thinking about my species what what is my effective species which combines a one and a two look look like, and my argument is going to be and I can show it here is that is going to lie somewhere in between a one 16:24:45 and a two one this space, how, how much closer to a one or a two it's going to be. It's going to depend on which of them is more more dominant. But the idea is that I can sort of make an effective species consumption rate here in this sort of pointy thing, 16:25:01 and ineffective species. Be beef. 16:25:06 And if I do that, or this kind of explains why you often get stranger and correlations being in higher. 16:25:14 Remember that I showed you that as a function of the competitive distance be any to stay a bit between any two strange, strange, strange coupling goes goes down. 16:25:25 So there is some sense in which the distance between any two points in the space. 16:25:30 Rep represents the correlation between, between those, those two strengths. 16:25:35 And so what you can see here is that it's very often the case that you will find for two species, or 232 strains that are much closer than the constituents species. 16:25:46 And so typically what what what what happens in this gym geometric picture is that you have one pair of strains which is much closer. and that has the highest level correlation than the constituents species themselves. 16:25:59 And that is what we think, typically leads to this generic consequence of us being this sort of status. 16:26:07 What is the sort of conceptual model that Yes, question. 16:26:12 Yeah, sorry, I was just gonna ask, how much does this rely on the total resource consumption of all the species being the same song What relies on that I've tried versions. 16:26:23 Whereas tried, try to break that up a little bit. The key thing here is I can't break it too much, just because then it's very hard for multiple strains to coexist. 16:26:32 But, in as much as I recall the enzyme budget around a bit. This result is robust. 16:26:39 Okay so, so 16:26:43 how does this depend on the amount of total amount of resources. 16:26:49 It isn't very sensitive because I'm looking at the relative abundance. 16:26:56 In the end, in the end, what, what my whatever my starting resource constellations are the species kind of deplete them all at a certain rate and they get to certain. 16:27:06 And so in the number of going to turn it up upon the total biomass certainly is going to be dependent on how much I'm sure there are a lot of questions, but let's let's hold off until the end because there's not much time left. 16:27:17 Okay. I'm liking the question. 16:27:21 But yes, there is less, less time. 16:27:23 Okay, so what is the conceptual model that sort of emerges from, from this on, here's what we think might be going going on. 16:27:31 Typically nature. 16:27:34 You know there are species or this is you know some sort of broader tax taxonomic unit some, some people think it's fam families. There are various clusters that exists in phenotypic space. 16:27:48 And so what I'm showing you, for some simplicity is just a phenotype space, which consists of the consumption rate of resource one consumption rate of these, these species and, and each of these points here is a different state. 16:28:03 This is sort of maybe what's going going on in nature. 16:28:06 When we domesticate communities from nature and bring bring them to the lab. 16:28:11 We're picking a subset of these strains and having them coexist with each other. 16:28:17 And once we pick the subset of strange from the species clusters. 16:28:22 Then we start to see correlations within them and this has this has implications for their dynamics, around the equilibrium. What do I mean, let's say, we take this, this pair of strains that we happen to that happened to survive in our community and 16:28:39 the closer phenotype space there tend tend to be strongly correlated, even though they come from across different species bound. 16:28:47 closer phenotype space there are 10, tend to be strongly correlated, even though they come from across two different species bound and end if they're in. 16:28:54 Insert farther and resource base, or then they will tend to tend to be caught. And so, this is what they. This is sort of, I think, one, one way to reconcile why there are still there still cohesion and phenotype at the species level, but still in these 16:29:05 communities, you can tend to these TC these behaviors were strains of different species are often much, much more coordinated than strains of the things is. 16:29:18 I have a bunch more of a show you about the genetic basis of this thing. 16:29:22 But, do I have a couple of minutes. 16:29:25 Okay, so I just end up with the maybe the most interesting thing that happens here I've kind of already told you. There are a bunch of things that are, that, that we find that are that are typical snips that distinguish strains that typically transporters 16:29:44 But there's one more really, really interesting thing that happens in these communities that I haven't told. 16:29:50 Up until now, we've only been looking at that, 98 99% of snips that I said what present at the first time point in these communities. 16:29:57 What about the other one or 2%. 16:30:09 Those are the one or 2% are typically what we think of as de novo new mutations there, they are about 400 generations that these communities are going, going through. 16:30:11 And so certainly one might is expect a small number of mute mutation that take up, and the most interesting class of the novel mutations that that we see is that a lot of them, generate what we call solutions. 16:30:26 And so what is a pseudo gene trajectory Look, look like. 16:30:30 Here is sort of what what that looks looks like on the y axis again I have the alien frequency of each snip the x axis I have time and again in the gray I've shown. 16:30:41 Again, another cluster that I had before Each of each of these are snips that have a freak that have a non zero frequency at the first time point, or the trajectory in red is the trajectory of for dinner mute mutation, why do I call it a de novo new mutation 16:31:03 because it had frequencies zero at the first three time points, and then suddenly arose. 16:31:04 And typically what we see happen, and and so this new mutation corresponds to the pseudo generalization of a particular GL described which, which, which in this region has been up a bit. 16:31:15 And then what we see typically typically is that some of these genes go up and go, go down all of those we throw away because we can't be sure well there were a sequencing errors or not, but some of these actually eventually come up and sort of seen him 16:31:30 to go into one one string. One of the two strings is sort of management of cluster, sand, and this, and this is what we think of as the mutation fixing in one of the strengths. 16:31:45 And I won't belabor the point here but typically it seems like there are, there is a steady fraction of new pseudo genes that are created, even though there are a bunch of pre existing pseudo genes that are found in the strings on the y axis here is a 16:31:58 a number of regions that we found. Po Po community, and on the x axis here is time in the number of generations, and most of these new solutions that can be fine. 16:32:11 Are belong to categories such as taxes proteins or various PAGE PAGE pro proteins that may be one can argue, or not that important in this community. 16:32:24 So maybe what we think this could be some small preliminary evidence for is this idea of use it on. 16:32:34 And I'm not going to hammer on this a lot, but this is what we think is going. 16:32:40 So, let me summarize what I showed you today was that we could track the strain dad dynamics and the standard ecological replicate communities drawn from pitcher plant. 16:32:53 And we find that often highly related strains can display, very distinct dynamics. 16:32:59 They typically seem to decouple beyond the genetic distance about 100 snips. 16:33:05 And importantly, and this is this to me was the most surprising thing in this study that often strains of the same species are much strongly correlated with strains of other species. 16:33:17 Then, with their own strains, or with the whole species. 16:33:23 There are a lot of open questions that are left there. 16:33:27 And only dig, dig into some of them this is my last. 16:33:31 One of them is that a lot of us have been talking about course of course training and functional groups, and so on and so forth. When, when, when I've been part of this conversation, and often what we tend to think of is sort of clustering things on a 16:33:47 phylogenetic sort of saying always families the right level is generally the right right level. And what this study sort of is trying to maybe tell us is that because strings of different species of strains of often different general can be much, much 16:34:01 more similar to each other and terms of their dynamics, maybe we need to go much final. 16:34:07 Before we go much course. 16:34:11 The second thing is that how typical is this idea of string stranded decoupling we showed this in this one, one system. But is it you know fairly typical is this genetic distance of about 100 steps is that a generic thing or is that particular to.