Multifactor Analysis of a Genome-Wide Selection System in Brassica napus L

Brassica napus is one of the most important oil crops. Rapid breeding of excellent genotypes is an important aspect of breeding. As a cutting-edge and reliable technique, genome-wide selection (GS) has been widely used and is influenced by many factors. In this study, ten phenotypic traits of two populations were studied, including oleic acid (C18:1), linoleic acid (C18:2), linolenic acid (C18:3), glucosinolate (GSL), seed oil content (SOC), and seed protein content (SPC), silique length (SL), days to initial flowering (DIF), days to final flowering (DFF), and the flowering period (FP). The effects of different GS models, marker densities, population designs, and the inclusion of nonadditive effects, trait-specific SNPs, and deleterious mutations on GS were evaluated. The results highlight the superior prediction accuracy (PA) under the RF model. Among the ten traits, the PA of glucosinolate was the highest, and that of linolenic acid was the lowest. At the same time, 5000 markers and a population of 400 samples, or a training population three times the size of an applied breeding population, can achieve optimal performance for most traits. The application of nonadditive effects and deleterious mutations had a weak effect on the improvement of traits with high PA but was effective for traits with low PA. The use of trait-specific SNPs can effectively improve the PA, especially when using markers with p-values less than 0.1. In addition, we found that the PA of traits was significantly and positively correlated with the number of markers, according to p-values less than 0.01. In general, based on the associated population, a GS system suitable for B. napus was established in this study, which can provide a reference for the improvement of GS and is conducive to the subsequent application of GS in B. napus, especially in the aspects of model selection of GS, the application of markers, and the setting of population sizes.