Genomic prediction for soybean segregating populations: selection strategies and training set establishment

Abstract

Genomic prediction for soybean segregating populations: selection strategies and training set establishment New soybean cultivars are generated from bi-parental crosses, followed by selection and homozygosis increasement stages, which the order of number of generations can vary according to the breeding method adopted. In the initial steps, the low quantities of seeds per progeny and the large number of individuals to be tested, makes it impossible to obtain a high-quality evaluation on field. In this context, genomic selection comes as an alternative predictive method, instead of simple random sampling. Therefore, the objective of this research is to explore relevant aspects related to the application of genomic prediction in the initial stages of a soybean breeding program. The results show good prediction ability (above 0.4) for traits tested evaluated (yield, plant height and maturity), showing that it is possible to apply genomic selection already in ealy steps of breding and obtain selection gains. In addition, it has been shown that it is possible to obtain predictive abilities equivalent to a full-sibs training set, establishing it only with pure lines, allowing the generation high predictive training populations without prior evaluation of within-family progenies, which allows the creation of stable training sets over the years and applicable in different families.