Genomic prediction – new tool in soybean breeding

Abstract

Yield and most of the agronomical important traits are quantitatively inherited, influenced by many loci with a small effect and affected by environmental conditions. When dealing with the improvement of quantitative traits, it is not particularly useful to perform the selection by using few major-effect loci as in traditional marker-assisted selection, but to simultaneously use genome-wide molecular markers able to capture all small effect loci influencing a trait. The training population consisted of 227 diverse soybean linesthat were used for genomic prediction model development. Training population was evaluated for yield at three consecutive years. DNA of each genotype was sequencing on Illumina HiSeq 2500, using GBS Discovery Pipeline for SNP calling. Prediction ability was evaluated using six mathematical models, including parametric and non-parametric and were validated on three different levels: self-prediction, cross-validation (5-fold) and external validation (historical data). Overall, genomic prediction ability for soybean yield was relatively high (0.60) and the results indicate a modest influence of mathematical model and marker number on the prediction ability using cross-validation and external validation. However, model had variable ability to predict phenotypic performance in separate environments, with especially high prediction ability in years not impacted by yield-limiting factors, when the genetic potential was fully achieved. Improvement of model performance in cross-validation and external validation was achieved by increasing the phenotyping intensity that must reflect the target environment variability. Obtained results indicate that genomic prediction can be integrating part of breeding process as useful tool that can increase breeding efficiency and decrees breeding time. Particular implementations are diverse, from germplasm screening and parental choice to the forward breeding and direct section based on genomic prediction.