Dimensionality reduction approach for genotypic data

Populations are genetically structured into distinct subpopulations. The analysis of population structure is the grouping of individuals into subpopulations based on their genetic data. One of the major challenges in population structure analysis is how to handle the high dimensionality of genetic datasets considering that hundreds of thousands of markers. In this paper, we propose unsupervised dimension reduction approach to reduce the dimension of genotype data. This approach is based on Weighted Network Analysis to identify the correlated markers. The approach is composed of three main steps: Network Construction, Modules Detection, and Modules´ Representative. The approach is implemented and tested in both simulated and real datasets. The experiments show robust and comparable results.