Shunichiro Tomura 'Ensemble approach for genomic prediction in crop breeding'

Shunichiro Tomura PhD Student, The University of Queensland ABSTRACT The application of genomic prediction has increased genetic gain in crop breeding. However, further acceleration across varying environments is required to meet the increasing yield demand. One bottleneck has been a low selection accuracy in genomic prediction models. A variety of models have been developed and evaluated, revealing that their prediction performance remains low and none of them have consistently outperformed others. Other approaches need to be proposed to mitigate such issues, and our research investigates the power of ensembles of multiple prediction models. Conceptually, by combining information for prediction captured by each model, more comprehensive information can be formed as “a collection of wisdoms”, expected to raise prediction performance. We naïvely averaged prediction results from each model, and despite the simplicity, the ensemble approach outperformed other models demonstrating that there is a potential that ensemble modelling approaches can enhance prediction accuracy in crop breeding. This webinar was recorded as part of the ARC Centre of Excellence for Plant Success in Nature and Agriculture seminar series.