Download or read book Comparison of Full-Sib Hybrids and Testcross Progeny Tests for Genomic Prediction and Line X Tester Analysis of Hybrids of Two Multiparental Maize Populations written by Brett Burdo and published by . This book was released on 2018 with total page 0 pages. Available in PDF, EPUB and Kindle. Book excerpt: In this experiment, we compare two progeny tests for unevaluated maize doubled haploids (DH) from two six-line synthetic populations with founders representing the Stiff Stalk and Iodent heterotic patterns. Either testcrossing, where lines are crossed to a common tester of the opposite group, versus information from hybrids produced by randomly crossing pairs of untested inbred lines, deemed Full-Sib (FS) hybrids, in the context of a genomic prediction enabled breeding program. Hybrids were evaluated for grain yield and 6 ancillary traits at two locations in 2014 and 2015, with two replications per location. Marker-based variance component estimates that account for deviations from Hardy-Weinberg equilibrium, suggest that the variance present in this population is predominantly additive. Coincidence of selection of the top 25 lines based on traditional General Combining Ability (GCA) for grain yield within the testcross and random pairs was 48% for the Stiff Stalk and 11% for the Iodent. Coincidence of selection using an additive model for genomic GCA estimation and testcross traditional GCA estimation was higher, at 56% for parental DH lines from either heterotic group. Partitioning dominance variation in the model increased coincidence of selection to 62% for the Stiff Stalk parents but did not increase coincidence beyond 56% for the Iodent. General Combining Ability (GCA)-based prediction accuracy of FS hybrid performance based on midparent of testcross performance was 0.73 and could not be improved with genomic prediction of the hybrid values. 10-fold cross validated prediction of the grain yield of hybrids of untested DH lines was 0.45 and 0.48 under an additive and additive + dominance G-BLUP model trained on the randomly paired lines, respectively, and 0.49 and 0.48 when trained on the testcrosses, none significantly different. Prediction of grain moisture, test weight, plant height and ear height were worse when training on the intercrossed population, but better for prediction of flowering traits. Incorporating marker information does allow for greatly improved GCA estimation from FS hybrids, but this is mostly due to accurate partitioning of additive effects, rather than a significantly enhanced capacity to capture non-additive effects with genomic estimation in the FS hybrids.