CAREER: Reducing barriers for comparative methods

进化生物学中的一个主要问题是物种如何以及为什么会随着时间的推移而发生变化。例如，从对目前物种的观察中，我们知道竞争可以推动物种变得不同。这些相互竞争的物种可能会发生变化，其中一个会变得更大或更小，而不是保持两个体型相似的物种。类似的变化过程可能在过去也发生过。关于过去的过程还有其他问题：非鸟类恐龙的灭绝如何影响哺乳动物的进化，随着人类更多地在城市定居，人类疾病是如何进化的，等等。进化生物学家通过编写一个描述这一过程的模型，然后评估该模型的各个方面来解决这些问题：物种相互排斥的强度有多大，哺乳动物的排斥率增加了多少，毒力随着时间的推移发生了怎样的变化。然而，对于这些问题中的许多问题，开发模型一直是一件棘手的事情，使得这些问题基本上无法回答。这个项目将开发工具，使创建这些模型变得更加容易。它通过使用计算机模拟特定模型下的进化，并尝试调整模型参数以最好地预测观测数据，这种方法被称为近似贝叶斯计算。黑客马拉松将加强这一点，这是科学家和开发人员聚集在一起创建或扩展软件的紧张会议，也有助于增加生物学家的编程技能和工具。该项目的另一个成果是创建了一门公开课，教授使用现有和灭绝物种的谱系图了解过去的尖端方法。该项目结合了用于比较方法的近似贝叶斯计算(ABC)的新方法，作为实施新模型的一种方式，通过会议让新的和有经验的程序员共同参与项目(黑客马拉松)，以及一门新的开放系统发育课程。考虑到R在比较方法用户社区中的广泛使用，ABC方法将用R编写，并使用离散时间模型，该模型允许生物学家为一代人的运动编码模型，然后在系统发生树上使用该模型。原型代码已经可用(包TreEvo)，并将在开放源码模型下开发。总之，研究和教学将加快方法开发，扩大能够创造新方法回答问题的人才库(在数量和多样性上)，并加强总体上的系统发育培训。

A major question in evolutionary biology is how and why species have changed over time. From observing species in the present, for example, we know that competition can drive species to become different. Rather than remaining two species with similar sizes, those competing species can change, with one becoming larger or smaller than the other. Similar processes of change likely occurred in the past. There are other questions about past processes: how did the extinction of non-avian dinosaurs affect mammal evolution, how have human diseases evolved as humans settled more in cities, and so forth. Evolutionary biologists address these questions by writing a model to describe the process and then estimating aspects of the model: how strongly do species repel each other, what was the rate increase in mammals, how has virulence changed through time. However, for many of these questions, developing the models has been intractable, leaving these questions largely unanswerable. This project will develop tools to make creating these models much easier. It does so by using computers to simulate evolution under a particular model and tries adjusting the model parameters to best predict observed data, an approach known as approximate Bayesian computation. This will be enhanced by hackathons, intense meetings where scientists and developers gather to create or extend software, and which also serve to increase the programming skills and tools of biologists. An additional outcome of this project is the creation of an open course to teach cutting edge methods for understanding the past using family trees of current and extinct species.This project combines new approaches for approximate Bayesian computation (ABC) for comparative methods as a way to implement new models with meetings to have new and experienced coders work together on projects (hackathons) as well as a new open phylogenetics course. The ABC approach will be written in R, given R's widespread usage in the comparative methods user community, and uses a discrete time model that allows biologists to code a model for movement in one generation and then uses this model on a phylogenetic tree. Prototype code is already available (package TreEvo) and will be developed under an open source model. Together, the research and teaching will speed up methods development, broaden the pool (in number and diversity) of people who can create new approaches to answer their questions, and strengthen phylogenetic training in general.