Applying High Density Markers To Crop Improvement: Initial Experiences From The CIMMYT Maize Program

Genotyping by sequencing (GBS)  has made it possible to track thousands of polymorphisms for less than $30 per line.  Over 9,000 lines from the CIMMYT maize program have been genotyped via GBS., including a subset of 23 full-sib families of F_2:3 lines in nine half-sib families, with an average of 187 lines per full-sib family. This population is being used to test the accuracy of genomic selection (GS) for predicting both genotypic value for commercialization and breeding value for parent selection. At moderate marker densities and repeatabilities typical of CIMMYT stage 1 testing, accuracies for yield are approximately half of phenotypic screening accuracy. This accuracy could result in accelerated rates of genetic gain from rapid-cycle GS relative to gains from phenotypic selection. However, the currently achievable accuracy for yield prediction is not adequate to justify advancement of lines based on genotype only. Modest increases in accuracy seem achievable relative to phenotypic selection if all entries phenotyped are also genotyped, and genomic estimated breeding values used to select lines for the commercialization pipeline. The key issue to be resolved is the usefulness of information from relatives more distant than full sibs. Strategies that use GS to shorten the breeding cycle will have a greater impact on gains than those that increase selection intensity, even if genotyping costs drop substantially, because the cost of generating a maize line is high. However, the beneficial effects of using GS for stage 1 selection are likely to warrant genotyping all lines entering replicated testing.