Context-Specific Genomic Selection Strategies Outperform Phenotypic Selection for Soybean Quantitative Traits in the Progeny Row Stage
Smallwood, C., A. M. Saxton, J. D. Gillman, H. Bhandari, P. A. Wadl, B. D. Fallen, D. L. Hyten, Q. Song, and V. R. Pantalone.  2019.  Crop Science, 59:54-67.

Evaluating different selection methods for relative utility is necessary in order to choose those which maximize breeding results. Soybean [Glycine max (L.) Merrill] yield, fatty acids, protein, and oil are all commercially important traits that display quantitative inheritance. Thus, it is of interest to evaluate breeding methods for these traits that can account for the entire genome. In addition to phenotypic selection (PS), the molecular breeding methods chosen for this study were BayesB, G-BLUP, and Epistacy. These methods were evaluated in a soybean population consisting of 860 F5 derived recombinant inbred lines (RILs), which were genotyped with 11,633 polymorphic SNPs. In order to simulate progeny rows, each RIL was grown in a single plot in 2010 in Knoxville, TN and phenotype was recorded. A subset of 276 RILs was then grown in multi-location, replicated field trials in 2013 to evaluate the relative utility of each selection method. Notably, for each trait the preferred method was a molecular selection strategy. Epistacy was the best method for yield, and BayesB and/or G-BLUP were the preferred methods for each of the other traits. Yield was the only trait for which the predictions had a large change when the number of SNPs and the number of RILs were randomly reduced for the G-BLUP model, with the best predictions occurring when RILs not grown in 2013 were removed. These findings provide important implications for how soybean breeders can maximize selections from the progeny row stage for yield, fatty acids, protein, and oil.