In this poster, I outline
how I chose to be on only some of the bandwagons (Rex Bernardo 2016) to rebuild
Purdue’s soft red winter wheat breeding program since 2015. Wheat breeding
programs in KY, OH, IL, MI, NY, and IN created a genomic selection (GS) consortium
in 2020. In this project, we increased the capacity of stage-1 testing outside
of the state of origin, referred to as sparse testing. We also use genomic
prediction besides phenotypic selection to increase the confidence of selection
from the single-replicated stage-1 trials. Genotypic data were produced after
planting of 600 stage-1 Purdue lines. After harvest, genomic estimated breeding
values (GEBVs) based on current season “raw yield” and “relative yield” data
were predicted. Across 5-fold cross-validation with 20 iterations, the averages
and ranges of accuracies, in terms of correlation of true phenotypes and GEBVs,
were 0.53 (0.37-0.63) for “raw yield” and 0.57 (0.38-0.69) for “relative
yield”. The stage-1 lines used in 2020-21 were progeny of 2015 crosses and
breeding populations that were developed in large field plots to allow natural
and breeder selection for multiple years. I avoided speed breeding because it is
associated with the risk of small size breeding populations, lack of natural in-field,
and lack of breeder selection during the early generations. In our line
production pipeline, lines are extracted from F4 bulk generation plots, head
rows are planted as F5, and stage-1 testing are planted as F6 generation. For incorporating
genomic prediction, I used a combination of phenotype and GEBVs as selection
criteria to advance from stage-1 testing. Those that showed highest phenotypes
and highest GEBVs received highest priority for advancing. We are now able to
produce much more lines than our capacity to place them in stage-1 testing.
Maybe genomic prediction of untested lines can be used as a strategy to
pre-select what should go for testing. We collaboratively continued NorGrains
under a new brand of BIG6. The key change is massively increasing the number of
advanced lines testing in multiple locations.