Multiscale model of exploration-exploitation tradeoff: from genes to collectives

 描述（由申请人提供）：许多动物，特别是人类，依赖于社交网络来获得他们的一般福祉，在许多情况下是为了他们的生存。社交网络对个人的生物医学影响可能对控制肥胖和药物或酒精滥用产生重要影响。这些后果还可能通过军事指挥和控制对任何组织的运作产生重大影响，无论是小型还是大型公共和私营组织。在许多情况下，对群体的中央控制必然是松散的或不存在的，组织产生于每个人所采用的一套规则。因此，了解适应性的集体行为是如何从一群几乎没有中央控制的个体中产生的是很重要的。该提案中的研究旨在通过将模型与三个生物尺度的实验相结合来理解群体行为：从大脑神经网络中的基因表达效应，到这些网络如何影响行为，再到在没有中央控制的情况下运作的协调群体的集体动力学。我们将从一个不同于通常使用的视角来研究群体组织的一个重要问题。与大多数基于代理的方法一样，我们建议研究由行为规则不同的个体组成的群体，而不是让个体在一组共同的规则下运作。后一种情况在人类和许多动物群体中更为典型;因为个体在许多方面--经验、体型、年龄等--自然地存在差异，这影响了他们对各种情况的反应。我们建议将蜜蜂作为这类工作的动物模型，因为在一个大的（约10万只蜜蜂）社会群体中，任何个体的生存都取决于群体作为一个整体的表现，而这个群体的运作没有中央控制。此外，我们可以在多个组织尺度上研究蜜蜂生物学。我们可以通过实验操纵已识别基因的表达，监测和操纵大脑中的神经网络，并确定殖民地中不同基因型蜜蜂的组成。我们将专注于蜜蜂群体如何解决人类也面临的寻找食物的核心问题。也就是说，如何分配资源，以利用已知的资源与探索新的资源。几千只觅食的蜜蜂如果不能有效地完成这两项任务，就有可能使殖民地失败。我们专注于一个基因位点，已多次隶属于一个或另一个觅食专业。我们建议研究这个基因座上的不同等位基因如何影响 个体的行为选择，然后研究这些个体是如何融入群体的策略来解决这个觅食问题的。我们将使用一种新的多尺度建模方法，该方法集成了三个生物尺度，使用标准的基于代理的建模，平均场近似，大脑的决策模型和基因调控模型。通过建模和实验之间的来回相互作用，我们的方法将确定关键参数，使群体面对环境挑战。

 DESCRIPTION (provided by applicant): Many animals, and particularly humans, depend on social networks for their general well-being and in many cases for their survival. The biomedical impacts of social networks on individuals can have important implications for regulating obesity and drug or alcohol abuse. The effects can also have a large impact on the functioning of any organization, ranging from small to large public and private organizations through military command and control. In many cases the central control of groups is by necessity loose or nonexistent, with the organization arising from sets of rules that each individual employs. It is therefore important to understand how adaptive, collective behavior emerges from a collection of individuals with little or no central control. The research in this proposal is aimed toward understanding group behavior by integrating models with experiments at three biological scales: from gene expression effects in neural networks of the brain, to how those networks affect behavior, to the collective dynamics of a coordinated group of individuals that operates without central control. We will investigate an important problem of group organization from a different perspective than is commonly used. Instead of having individuals who all operate under a common set of rules, as would be true of most agent-based approaches, we propose to study groups composed of individuals who vary in their behavioral rules. The latter condition is more typical of human and many animal groups; because individuals naturally differ in many ways - experience, size, age, etc. - that influence how they respond to various situations. We propose to develop the honey bee as an animal model for this type of work precisely because the survival of any individual in a large (ca 100,000 honey bees) social colony depends on the performance of the group as a whole which operates without central control. Moreover, we can study honey bee biology at multiple organizational scales. We can experimentally manipulate the expression of identified genes, monitor and manipulate neural networks in the brain, and determine the composition of honey bees of different genotypes in the colony. We will focus on how honey bee colonies solve a central problem in looking for food that humans also face. That is, how to allocate resources to exploiting a known resource versus exploring for new resources. Failure to efficiently perform both tasks by the several thousand foraging honey bees risks failure of the colony. We focus on a gene locus that has been repeatedly affiliated with one or another foraging specialty. We propose to investigate how different alleles at this locus influence the behavioral choices of individuals, and then investigate how those individuals are integrated into a colony's strategy for solving this foraging problem. We will use a novel multiscale modeling approach that integrates three biological scales using standard agent-based modeling, mean field approximations, decision making models of the brain, and gene regulatory models. Through a back-and-forth interplay between modeling and experiments, our approach will identify critical parameters that allow groups to face environmental challenges.