Response to selection to different breeding methods for soybean flood tolerance

Selection for flood-tolerance in soybean [Glycine max (L.) Merr.] is mainly phenotypic. With the development of new molecular breeding tools, our research objective was to assess the effect of different selection methods at the F-4:5 and F-4:6 stages on the response to flood tolerance and yield at F-4:6 stage. Four breeding populations were subjected to six selection treatments: (a) flood tolerance screening using hill-plots (VIShill); (b) flood tolerance screening using long rows (VISrow); (c) genomic selection using population-specific training (GS(hill)); (d) genomic selection using broad-based training population (GS(row)); (e) marker-assisted selection (MAS); (f) advanced based on agronomic adaptation under nonflooded conditions (random selection, RND). The top 15% lines within were tagged for selection, except for MAS that was adjusted based on recovery of desired haplotype. The complete base populations (BP) were advanced into flood and yield trials to determine probability of discard (POD), tolerance index (TOL), and seed yield. Analysis of variance was conducted across populations, and means were separated via Dunnett to the BP. Results indicated significantly different responses for flood tolerance (POD and TOL) across selection methods (p < .0001), with VISrow and GS(row) consistently having better tolerance selections than the BP (p < .0001). In addition, lines selected by RND had lower tolerance than BP (p = .0053 and .0618 for POD and TOL, respectively). Moreover, no significant differences were observed among selection treatments (p = .6797) for yield. In conclusion, when breeding for flood tolerance, selections under standard agronomic practices are inadequate, and genomic selection (GS(row)) or field screening using long rows (VISrow) are favored selection methods.