A Bayesian Approach to Inferring the Strength of Coevolution

物种之间的相互作用在生物群落的功能中起着重要的作用，对人类健康和农业具有重要的影响。例如，病原体及其宿主之间的相互作用影响自然群落的稳定性，并影响传染病在人类群体中的传播。同样，植物和传粉者之间的相互作用对于自然群落的正常运作以及农业系统的高效和经济运行也是必不可少的。出于这些原因，预测物种间的相互作用如何随时间变化对于了解生物群落、管理农业系统和最大限度地减少传染病非常重要。不幸的是，我们预测物种相互作用如何随时间变化的能力很差，部分原因是我们没有很好地了解推动相互作用物种进化的力量。这项研究将开发新的数学和统计方法来估计共同进化的强度，这种力量长期以来一直被假设在物种相互作用的进化中发挥关键作用。这项研究还将为数学、统计、计算和生物学领域的研究生培训提供重要机会，从而为STEM的人力资源开发做出贡献。最终，这项研究开发的数学和统计工具将通过开发免费计算机软件提供给科学界；该软件将使人们能够在广泛的生物系统中容易地估计共同进化的强度。目前有两大类技术可用于推断自然种群中共同进化选择的强度：(1)直接技术，它们做出可靠的推断，但仅限于特定类型的系统和少量的种群；(2)间接技术，它可以应用于广泛的系统和大量的种群，但提供的推理的稳健性有问题。即使在最好的情况下，这两种技术通常也只能产生定性的结果，因此不能提供关于自然种群中共同进化选择强度的关键定量信息。这项研究将利用贝叶斯统计学的最新进展，开发新的方法来估计建立良好的共同进化模型的参数，使用常规收集的数据，作为物种相互作用中特征匹配的广泛研究的一部分。一旦这些新的统计方法使用模拟数据进行了彻底的测试，它们将被用来在一个共同进化的教科书例子中估计共同进化选择的强度--有毒蝾螈和它们的吊袜蛇捕食者之间的相互作用。

Interactions between species play an important role in the function of biological communities and have important implications for human health and agriculture. For instance, interactions between pathogens and their hosts influence the stability of natural communities and shape the spread of infectious disease within human populations. Similarly, interactions between plants and pollinators are essential for the proper functioning of natural communities and also for the efficient and economical operation of agricultural systems. For these reasons, predicting how species interactions change over time is important for understanding biological communities, managing agricultural systems, and minimizing infectious disease. Unfortunately, our ability to predict how species interactions change over time is poor, in part because the forces driving evolution of interacting species is not well understood. This research will develop new mathematical and statistical methods for estimating the strength of coevolution, a force long hypothesized to play a key role in the evolution of species interactions. The research will also provide significant opportunities for graduate student training at the interface of mathematics, statistics, computation, and biology and thus contribute to human resource development in STEM. Ultimately, the mathematical and statistical tools developed by the research will be made available to the scientific community through development of free computer software; this software will allow the strength of coevolution to be easily estimated in a broad range biological systems. Two broad classes of techniques currently exist for inferring the strength of coevolutionary selection in natural populations: (1) Direct techniques that make robust inferences but are limited to specific types of systems and small numbers of populations, and (2) indirect techniques that can be applied to a broad range of systems and large numbers of populations but which provide inferences of questionable robustness. Even under the best case scenario, both techniques generally yield only qualitative results and thus cannot provide crucial quantitative information on the intensity of coevolutionary selection in natural populations. This research will capitalize on recent advances in Bayesian statistics to develop novel methods for estimating parameters of well-established coevolutionary models using data routinely collected as part of broad scale studies of trait matching in species interactions. Once these new statistical methods have been thoroughly tested using simulated data, they will be used to estimate the strength of coevolutionary selection in a textbook example of coevolution - the interaction between toxic newts and their garter snake predators.