隔离迁移(Isolation with Migration)模型数值计算的并行实现

A Parallel Implementation of the Isolation with Migration Model

在群体遗传学和分子生态学研究中,种群的分化分析是一个重要的内容。IM(Isolation with Migration)模型以DNA序列数据为基础同时评估两个种群分化过程中分化时间和迁移概率。IM模型能够模拟许多真实世界中一个种群分行为两个种群的现象。IM模型应用MCMC(Markov Chain Monte Carlo)方法进行参数的推断。然而马尔科夫链需要经过漫长的时间才能达到一个稳定的概率分布,并且参数推断所需要的空间可能超过一个计算机的内存。本文利用MPI(Message Passing Interface)实现了基于数据并行的IM模型,在减少单个计算机内存使用和总的运行时间方面都有很好的效果。

The analysis of population subdivision is a major focus in population genetics and molecular ecology. The aim of Isolation with Migration model （IM） is jointly estimating divergence times and migration rates between two populations from DNA sequence data. IM can capture many phenomena that can occur when one population splits into two. The parameters inferences for IM are based on Markov Chain Monte Carlo method （MCMC）. However it is a time-consuming process for equilibrium distribution and the memory required for parameters inferences may beyond the capacity of a single computer. In this article, we proposed a data parallel implementation of IM model based on MPI （Message Passing Interface）. The memory and time required for parameters inferences are reduced effectively.