Machine Learning and Individual-Based Simulation for Theoretical Biology

我们研究了模拟生态系统中物种的进化过程和出现。我们设想了EcoSim，一个基于个体的进化的捕食者-被捕食者生态系统模拟。每个个体的行为模式都是独一无二的，是进化过程的结果。这个工具的一个主要和独特的贡献是，它结合了一种行为、进化和物种形成机制。我们已经证明，我们可以有效地应用机器学习技术来提取最重要的特征，并提供一个易于理解的模型来解释问题的主要特征。我们将在模拟中添加几个功能，以允许通过大小、强度或感知能力等特征的演变来产生更真实的生态位。然后，我们将应用我们的方法来研究两个新的困难的开放问题，使用机器学习方法来提供语义丰富的模型和规则，供生物学家使用。选择一个人作为配偶是受性选择的支配。根据雄性的特征，雌性对雄性有特殊的偏好。仍然存在争议的是允许女性偏好进化的潜在遗传机制。众所周知，有两个主要的遗传质量好处：好基因和相容基因。然而，它们中的每一个在由此观察到的遗传多样性中各自所起的作用仍然不清楚。我们将使用几种择偶方法来研究多种假设。我们将分析一些不同的运行，不同的许多因素和环境条件，以了解在哪些条件下，以什么代价这种交配策略可能会受到进化的青睐。虽然无性繁殖在真核生物中相对罕见，但它比有性生殖有几个强大的优势。鉴于这些优势，有性生殖为何如此普遍令人费解。我们感兴趣的是研究有性繁殖是如何以及为什么出现的。这些模拟的目的是确定资源水平和环境应激源(捕食者或病原体)和繁殖模式之间是否存在关系。我们将能够检验许多不同构型的假说，并评估哪些环境有利于有性繁殖的显现和扩散。

We study the evolutionary process and the emergence of species in a simulated ecosystem. We have conceived EcoSim, an individual-based evolving predator-prey ecosystem simulation. The behavioral model of each individual is unique and is the outcome of the evolution process. One major and unique contribution of this tool is that it combines a behavioral, an evolutionary and a speciation mechanism. We have shown that we can fruitfully apply machine learning techniques to extract the most significant features and provide an easily understandable model to explain the main characteristics of the problem. We will add several functionalities to the simulation to allow more realistic ecological niches to emerge through the evolution of features such as the size, the strength or the perception capacities. Then, we will apply our approach to investigate two new difficult open questions using machine learning approaches to provide semantically rich models and rules that be used by biologists. Choosing an individual as a mate is governed by sexual selection. The females have special preferences for males according to their traits. What still remains controversial is the underlying genetic mechanisms that allows the evolution of female preferences. It is well known that there exist two main genetic quality benefits: good genes and compatible genes. However, the respective role of each one of them in the resulting observed genetic diversity is still not clearly understood. We will investigate multiple hypotheses using several mate choice methods. We will analyze number of different runs, varying many factors and environmental conditions, to understand in which conditions and at what cost such kind of mating strategy could be favored by evolution. Although asexual reproduction is relatively rare in eukaryotes, it has several strong advantages over sexual reproduction. In light of these advantages, it is puzzling why sexual reproduction is so prevalent. We are interested in studying how and why sexual reproduction emerged. The goal of these simulations is to determine whether there is a relationship between level of resources and environmental stressors (predators or pathogens) and mode of reproduction. We will be able to test many hypotheses in different configurations and evaluate what are the circumstances favorable to the apparition and diffusion of sexual reproduction.