Design of Prediction Guided Plant Breeding Programs.

Plant breeding is enhanced by integrating different scientific innovations and enabling tools. One major challenge that comes with the wide adoption of genomics and biotechnologies is to rethink and redesign the breeding programs at different stages and different scales. The essence of this new wave of breeding methodology research is to effectively identify and exploit genotype to phenotype relationship so that desirable cultivars are continuously and efficiently developed. Data mining, successful in many other areas, may provide solutions to address this question, particularly when findings are integrated into the design of new plant breeding pipelines. In this study, three methods of representative subset selection from clustering, graphic network analysis, and genetic design are developed for training set design. With representative subset selection, we demonstrated that effective genomic prediction models can be established with a training set 2~13% of the size of the whole set in maize, rice, and wheat. Enhanced by design concept, genomic selection may reshape the plant breeding pipeline by enabling the efficient exploration of the enormous inference space of hybrid combinations. We propose three essential components to streamline the breeding in the post-genomic era: better product creation (BPC), knowledge discovery from data (KDD), and optimal program design (OPD).