An approach to improve the selection of Single Nucleotide Polymorphism Associated with Trypanotolerant in cattle

Abstract

Genome-Wide Association Study (gwas) is a common approach for exploring relationship between observed phenotypes and genomics variations. The aim of gwas is to discover the genomic regions controlling a given trait. Gwas methods are facing analysis of high-dimensional data in which the number of variables is too large than the number individuals, known as curse of dimensionality. Research is turning to machine-learning approaches as they are well suited for this king of data. In this study, we proposed an approach for selecting significant snps. This approach is a two-stage process, in which, we remove non-informative snps using Kruskal-Wallis test and then select the significant snps based on a predictive model variable importance measure. In the second stage, we used XGBoost algorithm to train and assess a model. We applied this approach to cattle resistance to trypanosomiasis. The results show that our model based on XGBoost is more stable and robust selection of significant snps than previous work.