 So as I come from Massey University, I'm finishing my PhD right now. And I'm going to talk to you about modeling the influence of human social rules on population ethics. First of all, just to get rid of it, but a few other instruments, I'm working with Mary Cox and Martin Hazelton from Massey University. And I'm also working with an anthropologist in Singapore, Steve Lansing. And I have data from Indonesia, collected by the Aikman Institute, processed in the University of Arizona. So a bigger benefit of collaboration. So I'm going to talk to you about human populations, JLX, and I assume you don't know much about it. But basically, we are using JLX to reconstruct history. Looking at JLX pattern of different population, we can tell a story of migration, settlement, and make sure population size. And we can even time them with molecular dating. However, they're all based on basically one framework called the quiescent framework, which is really, really simple. Everybody knows the model I came from, the kind of working, not all the time. And they have a really strong assumption which is random mating. It's an assumption that works maybe well for animals. However, in humans, we all know mating is not random. We have social rules. So that's where my work's coming. I'm looking at the interaction between mating system and genetic patterns. So every study around the world has rules of marriages. Always have. It's kind of typical of human studies. They are made of prohibition and prescriptions. Typically, all studies have a taboo about incest. You cannot marry your brother or your father. Still in New Zealand, you cannot marry your uncle, for example. And then there are some prescriptions. You have to marry your cousin, or you have to go to this other village and marry that person. And those are all social rules. So we're gonna defend marriage system as a set of migration plus choice of mates, because very often we have immigration in this. When I say marriage, I just think in terms of reproduction. So what matters at the end is that there's a baby. So that's a bit of a shortcut, but just keep that in mind. And not about the genetics. Well, there's a lot of genetics study, of course. And there's also some variation between genetics of different population and their mating system. Typically, there's a high migration scale with women moving from population to population. You also have different things than another population. But there's been no quantification of any of it, basically. Once again, in terms of genetics, assuming you may not know much, we have four kinds of different DNA. Myocontrol DNA, it's from our mother, and all of us have one. X chromosome is different in male and female. Autosomes are the ones that are not sexually linked. So that's a big bunch of our DNA. And then we have the Y chromosome only transmitted from father to son. So they're gonna tell different stories about what our lineages are. And we will be able to use those different data and what they tell us to reconstruct social roles. So first of all, there's this question. Why did the mating system emerge in all human societies? That's been studied by anthropologists and sociologists. The process could be economical, social, those have been modeled. It could be biological. You don't want to keep on marrying your brother because there's an impregnated effect and you're gonna have genetic defect. But there's no real penalty on that. We don't really know how it works. So are the mating systems driving the genetic change? Can we model it? And are the changes making a biological process that may influence the emergence of the mating system? So for the first time, we're gonna quantify complex interaction between marriage roles and genetics. So it's a complex system as you must be aware. And so I took a simulation approach. Had to code my own simulator because there's a lot of things to program. It's an individual based model could in C++ from Squatch. Basically, it's very fast and I can model all my multi-layer complexity from the individual having DNA transmitted from his parents, living in a population, marrying certain kind of people, migrating, evolving through time. So that's quite complex in the within. Can also have some demographic effects. It's quite computer demanding, but then I can do all the mating systems I want. So in my simulation, once again, my marriage system is gonna be a migration plus a choice of mate. I'm gonna impose on my individuals. I have independence of communities with migration rights. And I do for it from one generation to the next, I do one migration per individual or non-migration. I do my mating only within my community. So if I move somewhere else, I'm gonna marry there. I have a polygyny, meaning the male have several female that corresponds to the population I'm studying and what their past social rules are like. And then I have to think about dance because that's true for most human population. And I'm gonna look at a very specific and interesting mating system called asymmetric prescriptive alliance. It's been found in historical population and all those read that. So you can see it's all over the world except Europe. And it's kind of an old anthropological mystery because this system tells you you have to marry your mother's brother or daughter and you have to move to a specific village which is a very strange pattern, right? It's always the mother's brother's daughter which is only one of the four cousins you can have and you have a migration pattern. And no one really knows why they emerged and specifically why they emerged in so many populations. So it's made up of a migration system which is a wife-giver, wife-taker. So if you have five populations here, you give your female, your woman to another village and you take your women from a different village. So you don't take and give to the same village which is once again quite strange. And within that, you're gonna look for your mother's brother's daughter. If you do that every generation, this migration, your cousin is gonna end up being always in the population where you moved to. And the way I model it is that I'm gonna force people to do that, the female to move and try to marry their cousin. But because it's not very clear in the literature how strange it is, I'm gonna have two regression parameters, P and P-mate. If they are zero, everybody must follow the rule as much as they can. Meaning if you don't have a cousin, you can't marry it. But if you have a cousin, you have to marry it. But if this is one, so P-mate is one, I'm only doing the wife-givers, wife-takers. Everybody is relaxed in terms of who you should marry. And same for the migration, in which case if you marry, if you migrate wherever you want, you mate with whoever you want. It's a random system with female migration. And I'm gonna have a continuum of value here. And first I look at what's the influence? Is there an effect of the mating system on my genetics? I have my four kind of DNA, the Y chromosome, the X chromosome, allosomes, mitre. First of all, you see the Y chromosome here. I vary my two parameters, the regression parameter. There is no difference all across it because we don't affect anything in the male lineage. The male always stay in the village in all my kind of model. That's what we expect. Where we see the biggest difference is in the mitre-controlled DNA. So that's the one transmitted from mother to children. And basically when the system is the most stringent, the lower diversity in mitre-controlled DNA. While when the system is the more relaxed here, we have a higher diversity in DNA. And that's definitely observable. And here we see the same kind of pattern at the much lower scale. So then I do a study case on the property. Yes, so the migration is definitely interesting much more than the mating. So those two plots you don't see much because they are all on the same scale. This one I kind of zoomed in on different scale and you can still see that there's a correlation but this one is not very influential compared to that one. Also if you migrate to the wrong population, there's no cuisine. So the rule about marrying your cousin is kind of obsolete. So then I study Rindy. Rindy is a population in Indonesia here on Sumba which is in the middle. And it's famous for having an APA mating in history. We have the genetics and we're gonna use the genetics as a footprint of the past structure. And we want to see, can we reconstruct the past marriage from the DNA? We saw there's an effect. That seems significant. And we have data, genetic data for all kind of chromosome. Can we see in this DNA and reconstruct what was the mating system? So that's what my data looks like. I have 28 men because they present all kind of DNA. I have, for mitochondrial DNA I have sequences which means I have a set of basis ACTG. That's the base that we can have for 28 individuals. For the rest of the DNA, four samples failed. So I only have 24 and I have something called sniff chips. Basically for a site, you're gonna see whether you're mutant or not. You're gonna have a huge chip which is an array of those snip. And you're gonna have binary data, are you a mutant or not? For each individual, for each site. So you have typically for the otosomes, 600,000 sites. So that's huge data, as you can imagine. And within this data we have signal. That's the number of sites that show polymorphism. So at least one individual is different from the other on this site. That's for the otosomes. If you only look at one chromosome, we still have quite a lot of data. The X chromosome, same. The Y chromosome, we can see we don't have much data. So unfortunately we're gonna have to kind of drop the Y chromosome because the data is not good enough. And that's our mitochondrial DNA sequence. To do my inference of my population, mating system for my simulations, I'm using an approximate Bayesian computations framework. I can answer questions about it. It's a bit complex of a complex Bayesian framework, but basically it's simulation based inference without likelihood, which makes it quite nice for this problem. It's often applied in population genetics. And I can infer either parameter or model choice, or both. In this case I'm gonna do an inference on parameters to estimate my best mating system. And I'm using a rejection algorithm. The way it works basically is that I do a ton of simulations and I choose my simulations which are the closers to my observations. And I do it first using all my kind of data and inferring three parameters, population size, which modifies a lot the genetic diversity. And my two parameters, P and Pmate, which are the way that I model my mating system. And if I do it blindly, I have a big inference of my population size since I can infer it. And two big blob on my other parameters, which doesn't seem like I can infer any of it. But then I looked into details. That's a typical plot where I have, that's just per population size. My diversity in black, it's all the simulations. In red, it's the observations. And in green, it's the one that have been, the simulations that have been selected in my ABC model. And you see, well, first my simulations are within my observations, so I must simulate something right. But the selected simulations don't seem very close to reality. But I actually look at much more than just this diversity on my control DNA. And if you look at all of what I observe and what I simulate, all of that is wrong. I simulate something which is really far from my observation. And that comes from a problem of data, which is the Assertion and Bias problem. It's a very famous problem in population genetics without simple solution. So what happened is in my steep trip I was talking about where you look at whether you are a mutant or not. They are chosen for medical purpose and they want to find sites with mutation. So you're gonna have a lot of sites that mutate much more than any site on your DNA. Basically we don't sequence where there's no mutation and we know a priori where there's gonna be mutation. So what I simulate in terms of just sequences that mutate doesn't match that. And the problem is that because those snippet chips are made by companies with copyrights, we don't know how they are made and they will not tell us. And those snippet chips change all the time as well. So there's no way to correct for it, there's no way to simulate it. So we tried a few corrections. Basically we had to look at a ratio of diversity between two, the autosome and the X chromosome with both at bias or the difference in diversity. And basically so that's our inferred posterior density for each parameter. We still cannot have the population size. We have a small idea of the PIMIC which seems not to be zero. But then we can't really infer the migration parameter. And we can do a cross-validation of our ABC. So typically that's our estimated value and that's our true value for the mating parameter. Doesn't really work. We don't estimate anything near the true value. For the migration, we have much, a bit better. It's not perfect but we kind of have a signal and same for the population size. But it's really not great. So the question is, did we lose the signal just because of this assessment bias? Is there a signal at all? We can do a correlation of our three inferred parameters and observations and there's high enough correlations that we should be able to infer those from the observation if we had the right data. Once again, we did the simulations without the observation. If we had the perfect data, could we infer anything? So that's our cross-validation which looks much better in terms of population size and PIMIGRATION. Doesn't seem like the PIMIGRATION mating. Sorry, we have much signal at all. And indeed if you look here, that's why we don't have much correlation. So as a conclusion, we can first observe and quantify the effect of the mating system and that's a framework for people who wants to study more population or what's this genetic effect of my mating system? Can I correct my current genetic model from that? In terms of inferring the mating system from the relics, the data we have is not good enough and we can't correct for it for now. But we can do full genome sequences soon, pour in a couple of years, five years top, that we become very cheap and it seems that with the full genome sequence which is what we simulate, we will be able to do it. So basically the framework is now here, we're just waiting couple of years for the data. Also the real mating system is quite unknown. People have great ideas, the anthropologist file about it. So that's why we moderate with some kind of pre-legislation parameter but that doesn't help with trying to reconstruct because there's so much unknown about it. We have this complex link, mating system and population data which is really interesting and I hope we can get much more out of it. And for just a bit of time, I'm gonna talk about what I'm gonna, I like to do further from that because that's a whole new world that can open from the genetics, how we can maybe look into social stuff, local migration. Because we've been able to do kind of quantify the penalty of, in terms of biological factor, how much this mating system affects our biology, we should be able to use it in terms of how did the mating system emerged, what was the weight of this biological factor and do more in terms of evolution of mating systems. So we can add the biological cost to the current social and economic network models of how society arise and I'm sure we can get still much more out of the genetic, trying to look at it from a different angle and just because I'm here, I'm actually looking for a puzzle. So feel free to pin me if you can. Thank you. Did you also estimate the probabilities for migration and parameters in models? Did you know what fraction were able to follow with it? Ah, so that's what I was trying to do with this and because of my data it seems, we can't infer much, like our power of estimation is pretty low but we can see that they don't follow the migration rule much because there's no probability it was there and if it was there we should be able to kind of see it and if they don't migrate according to the scheme they can't marry who they want because the cuisine won't be in the right place. Yes, so we could do dynamic but that would add much more parameter and because that would be very interesting. Unfortunately, the data is really poor in time of what people think happens and even what the genetic tells us. So for now we can't really look at it. What we can do is look into simulations how that would affect differently and we can look at how if you shift for mating system when do you lose the signal typically? So you have a mating system, you shift, there's a signal in the genetic and after a while it's lost. So there's a lot of things to look at here.