ERStruct: An Eigenvalue Ratio Approach to Inferring Population Structure from Sequencing Data

Inference of population structure from genetic data plays an important role in population and medical genetics studies. The traditional EIGENSTRAT method has been widely used for computing and selecting top principal components that capture population structure information (Price et al., 2006). With the advancement and decreasing cost of sequencing technology, whole-genome sequencing data provide much richer information about the underlying population structures. However, the EIGENSTRAT method was originally developed for analyzing array-based genotype data and thus may not perform well on sequencing data for two reasons. First, the number of genetic variants $p$ is much larger than the sample size $n$ in sequencing data such that the sample-to-marker ratio $n/p$ is nearly zero, violating the assumption of the Tracy-Widom test used in the EIGENSTRAT method. Second, the EIGENSTRAT method might not be able to handle the linkage disequilibrium (LD) well in sequencing data. To resolve those two critical issues, we propose a new statistical method called ERStruct to estimate the number of latent sub-populations based on sequencing data. We propose to use the ratio of successive eigenvalues as a more robust testing statistic, and then we approximate the null distribution of our proposed test statistic using modern random matrix theory. Simulation studies found that our proposed ERStruct method has outperformed the traditional Tracy-Widom test on sequencing data. We further use two public data sets from the HapMap 3 and the 1000 Genomes Projects to demonstrate the performance of our ERStruct method. We also implement our ERStruct in a MATLAB toolbox which is now publicly available on GitHub through https://github.com/bglvly/ERStruct.