Comparative study of different non-parametric genomic selection methods under diverse genetic architecture

Abstract

Genomic Selection (GS) is the most prevalent method in today’s scenario to access the genetic merit of individual under study. It selects the candidates for next breeding cycle on the basis of its genetic merit. GS has successfully been used in various plant and animal studies in last decade. Several parametric statistical models have been proposed and being used successfully in various GS studies. However, performance of parametric methods becomes very poor when we have non additive kind of genetic architecture. In such cases, generally performance of non-parametric methods are quite satisfactory as these methods do not require strict statistical assumptions. This article presents comparative performance of few most commonly used nonparametric methods for complex genetic architecture i.e. non-additive, using simulated dataset generated at different level of heritability and varying combination of population size. Among several non-parametric methods, SVM outperformed across a range of genetic architecture.