Location: Plant Science Research
Project Number: 6070-21220-016-06-R
Project Type: Reimbursable Cooperative Agreement
Start Date: Feb 1, 2019
End Date: Feb 14, 2022
1. Test the hypothesis that single-plant based training is effective for tropical-to-temperate adaptational genomic selection in maize. 2. Test the hypothesis that a prediction model trained on multiple genomic features can be used to enhance adaptational genomic selection.
We will resume artificial selection for early flowering time on G2 of the DE and NC lineages of the TropicS parallel selected populations. At each location, 10,000 plants of the corresponding location-specific lineage will be grown,  1000 random individuals will be marked and used as a [location-specific] training set (these plants will be genotyped and phenotyped), and phenotypic mass selection for early flowering time will be performed. Simultaneously, at both locations, the inbred lines will be grown in randomized complete block trials with two replications; genotype and phenotype data collected on these lines will serve as separate location-specific training sets for each location. At a winter nursery [in FL], the separate genomic prediction models generated per location in Stage 1 will be independently used for genomic selection of the corresponding location-specific lineage. For each of the four populations, 768 plants will be genotyped and 50 plants will be selected and intermated based on genomic predictions. In addition, using the same population size (n=768) and selection intensity (i=5%), a third sample of each seed stock (DE-G2:3 and NC-G2:3) will be randomly selected and intermated to serve as a control during Stage 3 evaluation. A replicated field trial at both the DE and NC locations will be used to evaluate the performance of genomic selection per se (relative to the randomly mated population that was generated in the off-season). The genotype-phenotype dataset used for genomic prediction in Aim 1 will be repurposed in Aim 2 for analysis by genome-wide association mapping. This dataset includes 768 individuals phenotyped in DE, 768 individuals phenotyped in NC, and 384 lines phenotyped in both DE and NC. Ancestral haplotype reconstruction on GBS data will be used to obtain genome-wide breakpoint maps. For regions detected in g3, we can combine large-scale data from extreme mapping in 11k individuals. In addition, a single year field trial will be used to compare phenotypically selected generations 0 and 5 of the TropicS-DE lineage in terms of changes in genomic features. Experimental units will be overplotted with individuals relative the number of individuals sampled for analysis in order to sample across the range of flowering time variation in each generation. Using 6-row plots (each having a total of 96 plants), a randomized complete block design with 6 replications will be executed as follows. The split plot design will include three main plot treatment levels (G0, G5 and a [B73/Mo17] hybrid control. Flowering time will be scored on each plant in order to test for differences in flowering time means between G0 and G5. Six individuals per replication will be sampled for genomic feature analysis: two individuals will be sampled at random and two individuals will be sampled from each extreme (early and late) in flowering time.