 Proposition for this work. So let's begin with the history context. The first modern humans who arrived in the Southeast Asia arrived 4,000 years ago. And they were the ancestors of the aborigines in Australia and Papua populations in New Guinea. And they are called Melanesians. These people are identified as the Melanesians populations in many historical and archaeological papers. And in those time, 4,000 years ago, the sea levels were different. And the bold line represents the minimum sea level. And when they see what at this level, there were more land for humans to migrate in the islands. Because that's why they came from Asia down to Australia. The second step is that about 4,000 years ago, another migration, big migration events began. It was the beginning of the Austronesians people coming into the islands of Southeast Asia. And the main hypothesis today is that they come from Taiwan. And this hypothesis came from genetic studies. But 4,000 years ago, the boundaries of the islands were not alike today. So for this population, the landscape was really different for them to migrate to move and to settle down to build new villages. So this is a historical context. And we are more interested in this second step and the mixing between these two populations of Melanesian and Austronesian. And we are especially interested about the genetics behind this mixing. And this is a figure from a paper in 2010. And it is the Austronesian percentage of the populations on different points on the map. And you have the population with 100% of Austronesian ancestry in Asia and in the islands near Asia. And the population with very little of this ancestry in the east. And so when you look at this data versus the longitude, you have the black curve, which are the observed data, which shows a steep climb, which is, although we are expected, more something like the red line, which is more gentle climb along the longitude. Because it's something you have in geographical region with no such a landscape as Southeast Asia, which you have a mainland no geographical obstacle for people to move. And the real question of the work is why. Why do we have this, especially this strange and mixed rates, mixing rates in the central region of Indonesia on these islands? So the main hypothesis we have to explain this are throughout the geography. Because we have a landscape of islands, you need to have some kind of selling technologies to move efficiently from one island to another. And this should have a big influence on migrations and on the genetics. You also have some other technological issues about population growth and farming. Because the population which came from Asia 4,000 years ago were more technologically advanced. And something that is not shown on the picture, but you can see in the genetic data that there is something strange in the mixed rates if you look at only the sexual chromosomes, X and Y, and the non-sexual chromosomes, that's the 22 chromosomes we have. So there is a difference in the mixing rates. So this tends to show that there is a preferred migrant rules between the Austronesian and Melanesian populations. And the way it should have been is that the Austronesian or Asian women should have preferred to marry with Melanesian men. But it's an hypothesis. So we decided to model this with an agent-based model. So it's a quick explanation of what an agent-based model is. It's a system of autonomous entities that are interacting with each other and that are living on environments. So for my model, these entities will be the humans. The environment is a geography. And we have interactions between agents. That's it, men and women will marry and form families. And between agents and the environment for the migration because we have villages with a limited amount of place, of room, and the agents will have to move where there is some space. More visually, this is the model. So we have our human population divided between single agents, which can be male or female. And when they marry, they are transformed into a family agent. So these agents live in villages. And these villages are connected in a network that is points taken on the map of Indonesia and Southeast Asia. And the edges are weighted. So in the first networks, just by the distance, so that agents will prefer to move to a shorter place, a closer place to the village where they were born. But we want to add some other weight to be able to make a difference between a migration on an island and between islands. And to be able also to take into account the sailing technologies at a certain point. We have, as we have two populations, two different groups, the Austronesians and the Melanesians. We want them to have different dynamics, so different migration rates, different fecundities, and different test rates, too. And we also need some kind of measurement of genome mixing rates. So we need a genome-like structure for the agents. So that's the model. Now let's play with it. First, we decided to start with a test network which is very neutral to be able to know if the model is behaving as we wanted. So I can show you a few videos of this. Sorry, it's not the good one. But you can find the network here where each village is growing or the population is decreasing depending on the distance and birth rate and the distance between nodes. Here you have the, we can follow the population, the black line is a total population. In green, you have the families. In blue, the men, and in pink, the women. And we, in this special simulation, one of the very first to know how the system is behaving, we also can follow the age structure of the population. So here it can be strange at the first look because these agents, single agents, really disappear when they form a family. So the age are just shifted in this diagram. But the humans always live in the system. And you can see that some single agents can remain above 20 depending on if there is someone you can marry in the village and if there is also somewhere to go, somewhere it's possible to go. One thing that is special in the migration pattern is that we only authorized an agent to move once in its life. Let's say the way we imagine the population at this time is that people will grow up to, let's say, 15. At the age of 15, they can marry to someone else. And at its age, they also can move to just one village that is connected to their village where they were born. So this is a very specific dynamic. So an agent cannot cross the whole network in its life. And as we are interested by the genetics, this is the percentage of Austronesian ancestry in each village over time. So if you remember the network, I begin with the left part in light blue, which is fully Austronesian villages, and the right part in dark blue, which are fully Melanesian people. And you can see that on this network, every village has a mixed population. And as we want to simulate this black line is the end of a regular simulation after 4,000 years. But if we let the simulation goes, we reach an equilibrium point around 0.5% of Austronesian ancestry in the genome, which is what is intended to be found in such a system, just a mix of two populations with a very neutral environment and no specific constraint on the network. And the second one will be shorter. Here, the network is a bit different because my Austronesian populations are source populations. I said Melanesian people cannot go in the West. So the genetic admixture we have with this kind of simulation is really different. I can speed up. It should be in time. And in this simulation, we also get rid of the edge effect and the corner effect we add in the previous simulation by controlling the size of the population, the size of every villages, along the overtime. So this is really a good rule because it's behaving as we intend it should. And we can make the difference with different starting conditions and different constraints on the network to test different hypotheses that we have for the model. Yes, it was just for the specific controls. So the next step is to work on the real network, the islands of Southeast Asia. So this is really recent. So you may have some criticizes on this. It's not a so big network, it's only 100 of nodes. But we are trying to decide now how we can connect this. The first thing we try is to just connect it by the distance. And it seems to be around a cutoff of 600 or 700 kilometers. We have something that is quite realistic, not too much connected, not too low connected. And this is starting in the early steps of the simulation. You can see some villages that have a population that are really decreased, other ones that have increased, especially in Borneo, which is highly connected. And at the end of the simulation, we have some places in the network where we have a mixture. And if you look at the evolution of the genetic admixture, you can see that you have Australian ancestry that have reached almost all nodes, all villages in the network. And we want to have the same curve as the first figure, and we can fit the data quite nicely. Here it has the financial situation we want, because it's shifted to the west. But it's probably only a matter of starting condition as a speed of genetic spread out. So we have a lot of things to do for this model, especially playing on the starting conditions, which because we have many assumptions, many possibilities on these conditions, we also at the very beginning of the setting down the network. So the number of nodes may be important, especially per island. We are trying to fit with the length of the coastline depending on the islands. And also when we are building the connections, trying to analyze the connectivity at the neighbor of each node. It's also quite slow with the big network, so we have an optimization phase to do. And obviously, exploring the parameter space from this network. Thank you. So you have the red dots there seem to be pretty clearly that you have two things happening there. Do you know what those two things are? Sorry. It's because you have one line which is from Taiwan there, and one line which is from Sumatra there. If you could go and construct one of these networks, given enough genetic data, you could infer how people move back. Do you have any idea how much genetic data you need to be able to build that sort of network? Is it anywhere close to being realistic to say we're going to connect to genetic information and try and infer a complete migration map and infer all of those large weights in there? For the migration? So I'm watching the genetic spread. I don't understand really your question. What amount of genetic data you need to build the model? To build this network. So you're using a network as your sort of starting condition, then you run some genetic simulation, not the way the genetic's here now. But you could look at it back, because you could say, I'm going to go out to the real world, connect all the genetic information that I can, and then build a network out of it. Do you have any idea how much? The data set is already big. It has been done by our Indonesian collaborators. And we have, I mean, 100 of individuals. I don't remember. Sorry. 2,000 across Indonesia. So it's quite big. And maybe it's a good idea to have a look at this in this way. I don't know if it was something we have thought about.