Relate example for quantifying positive selection

This repository contains a fully contained example of running Relate on CEU samples of the 1000 Genomes Project dataset. We include data a few Mb around the LCT gene and construct Relate trees and subsequently compute selection p-values using the module RelateSelection. 

The output is a p-value for every SNP in the dataset. These p-values are computed by first extracting the number of lineages remaining in the tree when the mutation was carried by two lineages. We then compute the probability of a mutation spreading to at least its present-day observed frequency. The p-value therefore quantifies to what extend the mutation has outcompeted other lineages in the tree. We can also condition on other time points in the tree, such as when around half the lineages remaining carry the focal mutation.

Running the scripts

Please download Relate binaries from https://myersgroup.github.io/relate/ and then run the script run.sh included in this repository.

本仓库提供了在1000基因组计划（1000 Genomes Project）数据集的CEU样本上运行Relate的完整可复现示例。我们纳入了LCT基因上下游约数Mb的区域数据，构建Relate谱系树，并通过RelateSelection模块计算选择检验p值。

本数据集的输出结果为每一个单核苷酸多态性（SNP, single nucleotide polymorphism）对应的p值。此类p值的计算流程为：首先提取当突变由两个谱系携带时，谱系树中剩余的谱系数目；随后计算该突变扩散至至少现有观测频率的概率。因此，该p值可量化该突变在谱系树中相较于其他谱系的竞争优势程度。我们还可基于谱系树中的其他时间点设置条件，例如当约半数剩余谱系携带目标突变时的场景。

运行脚本
请从https://myersgroup.github.io/relate/ 下载Relate二进制文件，随后运行本仓库中包含的run.sh脚本。