 Yes. Okay. Great. Thank you for Sylvia introduction. Today I'm going to talk about an ongoing project in my PhD thesis. And this is this work is about a pair wise competitive network of bacterial communities. To start with talk I want to give you some about the motivation of this work. So, as Martina pointed out, the micro we are living in a very diverse microbial world. And these microbes are virtually everywhere. And they perform important functions in their habitats. For instance, they are important decomposers in the in the ecosystems. They can also help degrades the wastewater from the human industrial processes. And of course they also produce fermented food that we like to eat like cheese or wine. And there has been great interest in trying to assemble synthetic microbial communities. Which can perform the functions that cannot be easily performed by the single strength for instance if there's a hard to degrade toxin or like lignin it's hard for a one strand to to fully degrade the complex that he needs a full community to do it. So, we as a human want to apply that into like using the ability of a microbial communities to perform particular function. But before a community can do a function it has to coexist first. So recent interest has been in predicting the coexistence of a of a species set of community from the pair wise competition. So, for instance, Friedman and Gore, they proposed the simple assembly rule. So basically they competed a, a set of eight bacterial isolates, but there are species, and then the competing all the pairs and trios. And what they found is that a trio can coexist only when all of its pairs coexist. So, for example, in this trio of ABC, it will degenerate into a pair of AB, because in the pair wise competition, a complete see. So, in this case, ABC will degenerate into AB. However, many of the works in such microbial community has have been with species that may not coexist together in the community, or the environments that we use to culture and and compete the species or measure the pair wise competition. They are the same as the environment where the community was first assembled. So it has been hard to see how these effects, the new structure that we observed. And a natural microbial community is usually highly diverse. So a soil of a grain of soil to harbor more than 1000 bacterial texas. And the number of pair wise combinations increase in quadratic. So it's really challenging to measure all the pair wise combinations if we are, we try to disentangle the pair wise network from these kind of highly diverse natural communities. If we are able to assemble a microbial community that with a decent number of species that we can manipulate and measure in the in the lab. In this community, the species survival is mostly different determined by the competitive in actions. And in principle, the network of their pair wise competition competition should tell us something about their coexistence. So, quite arguing here is that in imagining the species pool there could be many possible topology among species, and during ecological assembly of a community. Those network that fail to persist will be purged by the selective force imposed by the community assembly. And the other the rest of the network that can persist, or the network of consistent community will be favored any and there will be present in a final coexisting community. So, to test this idea, I'm taking advantage of the previous work in the central slab, which shows when we start from a very diverse species pool. So these species can be from leaves so from a soil basically very diverse, diverse sources of bacteria. When we grow these species in a simple environment in the lab. So it's basically has only sugar as the only carbon source. So at the end of 12 serial dilutions. So you can think about it as a periodic system know environment. And in the end, many of these communities have converged to a similar taxonomy composition. So these communities not only have the same families, at least they are in the family level. So these families not only have the same family composition, but they also have the similar proportions. So here I'm showing you the relative abundance of 13 representative communities that we have assembled in the glucose environments. And, of course, this pattern has been observed also in the other 96 glucose communities and the similar pattern also appears in the other kind of neutral environments. So what we we can take advantage here is that we are able to isolate the strands from these communities. So we know these communities coexist by themselves, and we can take apart the community and then isolate each strand. These strength will allow us to be in a good position to ask a question here, which is to stable communities that show similar taxonomy composition. Do they also show general network structure. So, they have been several ideas being put forward to what network structure we should expect to see in a coexisting community. So the first one is what I just mentioned from a simple assembly rule. So it's basically saying everybody should coexist with everybody in pairs. So, so that's the trio can coexist or higher diversity community can coexist. And the other idea is that intensive cycle should be expected. So even though the species would not coexist in a pair, they are able to coexist in the trio or higher, higher diversity communities. So to test these two ideas, I'm doing experimentally measuring the pair wise interactions. So take one pair from the first in the least community has a force strands and take one pair for example, we manipulated the initial frequencies of these pairs from 95% 5050 to 595. So we can grow these pairs in the environment, which is exactly the same as the community was assembled because it's a very simple just and like glucose medium. So after also serial passage. In the end, we, we played it's these final cultures on the solid alcohol plates and then because they're morphology at different we can count the colonies and determine the final relative frequency. So we have the relative frequency in the beginning of the experiment and after their competition. So, one example I'm showing is, I'm plotting the relative of frequency of a three different initial ratios in the beginning and in the end. So if each species can invade the other from rare. We need a lot these community, sorry, this pair is in a stable consistent relationship, then we draw a mutual link to connect these two species. On the other hand, if only one species can invade from rare, and the other cannot, then these pair is in a competitive exclusion relationship, then we draw a certain arrows to connect the winner toward the loser. So we can repeat it for all the possible pairs from the 13 communities and only pairs we've seen the community are measured. So, I think maybe before I jump into the result I can take a post here to make sure that people are fully understood with with the method. Does anybody have questions. Yeah, and I can I can proceed to guess. Cool. So, after we've done with this experiment, we can test the first hypothesis, which is a everybody courses with everybody and these hypotheses will give us a very intuitive expectation, which is all the pairs should be coexistence we shouldn't expect any exclusion happening in a community. Because remember these communities coexist by lens ups. Here I'm showing you the poor the result of all the 186 pairs I have from the 13 communities. So, what we found is quite contrary so the coexistence is rather a minority of the pairs, because only less than 40% of the pairs are coexistence, but more than 60% of the rest of land are mostly exclusion. So, these same slots, the simple as very rule may not apply in our community, or at least they may not fully explain why our community coexist. So to explain why they are very high, there is high fraction of exclusion links, we turn to the second hypothesis. So, the intensity should be expected. And we can measure intensity for a network, or for one community by this ratio of intensity intensity the number of intensity triads over the number of old child possible trades in a community network. And I'm plotting one intensity motif, for example, so this mock community has three only three species, and because he only has one possible trial so, and it's an intensity trials. So, this network has an intensity equals to one. And for the for all 13 communities I have. So, surprisingly, that we don't see any of the intensity found in the network. So, um, similar to the simple assembly rule the intensity also cannot fully explain why our community coexist. We have tested to high policies that have been put forward to explain why network, why, what they will structure should be present in the coexisting community. But they all fail to explain why they coexist. So, can we have a more comprehensive way to describe neural structure maybe we are missing some neural structure that is present but we didn't measure. So we use a idea of a network motif, which is basically a recurrent sub networks that's frequently observed in a large network. So these ideas has been the network motif idea has been used in a large number of different kind of biological networks. So this isn't a new idea. So, never motive has been used to describe networks such as full webs, or transcriptional networks or neural networks, and depending on the network types. The number of network motifs are different. And in our kind of competitive network, they are seven of them. So there are three motifs, one and seven, which are I just mentioned before. And there are the other five types of motifs, they are in the intermediates between motif one and seven. I explained these motifs in more detail when I show the data. But here I will just walk through how these motif analysis works. So, basically, we can compare the observed network to the randomized network to determine whether the observed motif is most statistically significant. So here is a mock community of 10 species, and all the pay-wise interactions is either pay-wise competitions, either exclusion or coexistence. So we go through each of the triads in this network. For instance, the red, I know it's hard to see but the red triads is the motif that has two, coexistence and one exclusion. Then we count this motif one, and we go to the next one, and we repeat it for all the triads and match them to each of the seven motifs. Then we come up, we can come up with the observation for this network. And in this case, motif one is not observed, and motif two is frequent, etc., etc. Then we randomize the network just by randomly swapping the pairs by like, we keep the number of coexistence and exclusion the same, but we swap in the links. Then we get a randomized network. Then we repeat counting the motif for this randomized network, and randomize the network again and then count motifs again and again and again for a thousand times. Then we can have a distribution of the motif counts from these random, these 1000 randomized networks. Then what we're doing here is that we can compare the observation to the randomized network. So, if the observed motif counts, for instance, motif one is very low, is very rare. So if he falls within the bottom 5% of the randomized network, and then it's considered that this motif is underrepresented. First, motif three is very frequently observed. If he falls within the top 5% of the randomized network, then it's considered a overrepresented motif. So, for these overrepresented motif we shaded by green and for underrepresented motifs we shaded by red. So we can repeat it for all the motifs. Then, before I go to the next slides, I would just want to emphasize that this is just one network. So we can repeat states for all the possible, all the 13 networks I have and counts the motifs. So in the next slide, what I'm going to show is a poor result of all 13 networks and counts the motif from the observation that also randomized networks. So, the first result that we can see is that the intensity motifs are underrepresented. So motif one, that forms an intensity loop, and this motif is never observed in our network. So it's consistent with our previous results. And of course it's also underrepresented compared to randomized network, which expect that we should see some of the intensity. And motif four is an interesting motif that is kind of half intensity, because it has a bbbc but ac coexist. And these motifs is also very rare in our networks. So the total number of motifs we have 288, and among then we have observed less fewer than five of these network motifs. And the third also kind of intensity, but it has two coexistence, and these networks also underrepresented compared to rendered. So some of the motif seven, the simple assembly rule, and it's more enriched than expectation by random. So it seems that at the Paris level, the simple assembly rule does not apply but here it seems that if you look at the network structure, then these motifs are, it's more frequent than we expect by random. So it seems that the simple assembly rules works to some extent to explain our coexistence, the current is coexistence. But this isn't the only enriched motif. So from the chart, take it now on this. So yes, you said this results are pulled over 13 networks. How reproducible are these observations slash statistics across the examples. In other words, what are the error bars on the plot on slide 18, as in how big are the error bars. Very for the clarity. So the first question. We. So across all the communities, there are some are really small, for instance, there's three species. So for last communities, if all the pairs are coexistence, then there's no way to randomize the network. And in the large communities, there are more ways to randomize the network. So, yes, there's a sample, the sampling has different like effects, depending on the community size. But in general, for those communities larger than five or six species, we do see the same pattern in those large communities. And for the second question. So the, the bar in these plots means that the range of the top five to the bottom 5%. So if a observation. Yeah, I understood that I understood what the bar here means I just meant to ask what are the, if you added error bars on this plot, which would be how irreducible they are across maybe your large network examples, how big would they be like in this, for example, in your motif seven, you know, it's observation is just slightly above that range, if I put air bars on the red dots. Yes, Martina is helping me there. Yeah, what are the air bars in the red dots. And I, I'm not sure if I can draw the error bar for observation because they just one count for the total. But this is a poor result so there. So it's not the mean of the of the old networks is a total count of all communities. So I don't think there's a way to draw a robot, but if it's a the mean of the total communities I seen these pattern should be consistent with at least among those large networks. Thank you, we can maybe discuss later as well. Yeah, yeah. Thanks. Thanks for the question. So, for the other kind of of motifs. Motif two, three and five they are all over represented. So these three motifs share a same. Common is which is they are all hierarchical, which means that a we see them what if they're hierarchies. So what if two is a pure transitive train. And what is three is that one species beats the other two, and the other two coexist, and what if five which is the AB coexist together, but like, each outcome piece the third species. So these three motifs are all transitive, and they are all over the sentence in our communities. So, I think from this part, I can, I think I have shown some data that support the idea that stable communities. Exhibit generic network, combating network, and which has the features of species post from these communities tend to be exclusionary land coexistence, at least in our simple environment. So the motifs, then interest in motifs are underrepresented, whereas the transitive motifs are over represented in a in a coexisting communities. So, with these data, we can ask a second question I think is very informative, which is whether these network topology, a property of community assembly, or is it more constrained and the emergent from the ecological So, in the interest of time I just want to highlight the scars scarcity of the two interesting motifs. So motif one is never observed and, and, and the motive for is also rare as well. And I think part of the explanation for these motif can be attributed to the species pool. So, if we consider a species pool, and we connect all the possible networks in the species pool, then we have this global network, which is determined by ecological interactions and ecological interactions affects the species survival which shapes and structure of this global global network, and somehow, if these global network has the motif distribution, different from the expectation by random. It's possible that the interest in motifs are also rare in the species pool. So, even though communities assembly, maybe favor interest in motifs, but they are so rare so it's unlikely that the community assembly could favor the enrichment of these motifs. So to test this idea, one way is to go through all the species pools and then try to map out the network in the species pool. And it has been so much work so I resorted to use a modeling approach. And I'm using this microbial consumer resource model, which is basically captures the microbes will behave. For instance, what the resource competition among microbes and what new chance that the micro prefer and and the metabolism that each microbes has. And then metabolism will determine what metabolic by products that the micro produces and release to environments. So we could include the cross the common cross feeding features of the microbial community into the model. And with this model. We could simulate simulates a experimental scenario, which is star from diverse species pool and then culture all the monoculture in the in the environments that we provide with them is a simple environment with only sugar. And these materials some of them will not be able to grow, because there's a minimum media. And for those that are able to grow, we can take a sample of three and compete all the pairs, exactly the same as what I did for experiments, then we can match a network for these communities, sorry for this network of three. We can repeat it for 1000 times to get a sense of what the species pool network would look like. Because, of course, is we cannot map out all the pair wise combinations in the species pool if a point so large, but we could easily get the data that is more rich than the experiment can do. I'm showing only motif ones and motive for for clarity. And as a control I also ran a top down assembly, which is how we assemble the communities tested in experiments. And it seems like the model can could capture the result of of those experiments quite well because motif one is never observed and motive for is also rare. And each dot here is a mean of a 10 communities that I tested and I repeated for 20 species for so there are 20 points here. And these top down assemble communities. Motifs interest, the two interesting motifs are underrepresented. Especially like regarding their absolute motive counts they're pretty rare. Similarly, the species poor also lack also lacks interest of teeth motifs. So motif one is never observed and motive for is also rare and there's no significant difference between the species poor and community in a motive for motive ones never observed. These kind of support the hypothesis that the interest tip motifs is already rare in the species poor. And, of course, we, I haven't exhaustively explore all the possible model parameter choice of the model. Guess, there will be too many of them, but by tuning some of the parameters we sense that these features is general. There's not contingents on the parameter choice. And this pattern appears as long as the new chance competition is considered in the model. So, the supposed idea that's ecological interactions, in particular new chance competition would limit the appearance occurrence of the intensity motifs. This is my final slides. So to stable communities assemble in simple environments so generic network structure, the answer is yes, and the structure is generic, at least in the simple communities that we considered. I guess I think, in the more complex community with all the other ecological mechanisms in play, I will be, I will expect that the structure will be different, and will be less predictable. It's an able topology of poverty of a community assembly or is it emerging from ecological interactions. I think part of the answer is, yes, and could be explained by ecological interactions, at least for the intensity that we have the lack of intensity we observed, and these could be mediated by by ecological interactions, and in particular resource competition. In fact, I would like to thanks for ICTP for inviting me to give this talk, and I also like to thank for thanks everyone in a central slab, especially Georgia Bayek who helped me with like experiments, and I also like to thanks for the funding fellowship from the Ministry of Education in Taiwan.