Inference and analysis of population structure using genetic data and network theory

Abstract

Clustering individuals to subpopulations based on genetic data has become commonplace in many genetic studies. Inference of population structure is most often done by applying model-based approaches, aided by visualization using distance-based approaches such as multidimensional scaling. While existing distance-based approaches suffer from lack of statistical rigor, model-based approaches entail assumptions of prior conditions such as that the subpopulations are at Hardy-Weinberg equilibria. Here we present a distance-based approach for inference of population structure using genetic data by defining population structure using network theory terminology and methods. A network is constructed from a pairwise genetic-similarity matrix of all sampled individuals. The community partition, a partition of a network to dense subgraphs, is equated with population structure, a partition of the population to genetically related groups. Community detection algorithms are used to partition the network into communities, interpreted as a partition of the population to subpopulations. The statistical significance of the structure can be estimated by using permutation tests to evaluate the significance of the partitions modularity, a network theory measure indicating the quality of community partitions. In order to further characterize population structure, a new measure of the Strength of Association (SA) for an individual to its assigned community is presented. The Strength of Association Distribution (SAD) of the communities is analyzed to provide additional population structure characteristics, such as the relative amount of gene flow experienced by the different subpopulations and identification of hybrid individuals. Human genetic data and simulations are used to demonstrate the applicability of the analyses. The approach presented here provides a novel, computationally efficient, model-free method for inference of population structure which does not entail assumption of prior conditions. The method is implemented in the software NetStruct, available at https://github.com/GiliG/NetStruct.