Genomic Selection Breeding Plans for Maize Hybrid Development That Use the Haplotype as the Selection Unit.

Genomic selection (GS) is selection on marker or haplotype effects summed across the genome in a genomic estimated breeding value (GEBV) to predict breeding value or performance of individuals under selection.  GS differs from current approaches to marker-assisted selection in that it integrates information from all markers in GEBV estimation, rather than from a significant subset.  The main conceptual difference between GS and other breeding systems is that the haplotype, rather than the line, is the selection unit; lines are treated as experiment units in initial phenotyping.  Our objective is to exploit this concept in the design of more efficient maize breeding plans. Currently, the early phases of maize breeding programs are designed to estimate the GCA of lines within the target population of environments (TPE) of a breeding program.    GS-based programs could estimate both GCA and SCA for each haplotype in early testing by crossing subsets of lines to different testers; because each haplotype recurs across several lines, its effects with and across testers could be estimated by considering line effects to be random.  Similarly, haplotype effects across the TPE could be estimated in early testing. Large populations could be generated and evaluated without replication across testing sites, with each line evaluated in a single plot at only one location; haplotype effects across locations would be estimated considering line effects random.  These approaches could increase selection intensity and allow estimation of tester- and region-specific GEBVs in the initial testing phase.  Treating the haplotype as the selection unit will permit small breeding programs to collaborate in “open-source” breeding networks in which the local breeding program receives unique genotypes that have not yet been phenotyped, accompanied by GEBVs specific to their environment and testers.  Combined with the increased gains achievable via reduced cycle time for GS, these approaches could increase the effectiveness of small breeding programs in the developing world.