Predicting agronomic performance of maize landraces in various and future environments by combining genomic prediction and ecogenetics

Maize traditional populations (landraces) hold valuable genetic diversity for addressing climate change and low-input agriculture but remain underutilized due to lack of evaluations. High-throughput pool genotyping (HPG) has been previously used to characterize diversity but its potential for implementing genomic prediction (GP) and genomic offset (GO) for maize landraces has not been tested yet. We developed HPG-based GP models, combined or not with GO and within-population gene diversity (Hs), calibrated using 397 European landraces evaluated across environments and use cases. GP alone showed high predictive ability for yield (0.75), plant height (0.92) and male flowering time (0.94). Including Hs and GO in the GP model improved by 13% the predictive abilities for grain yield of new landraces in new environments. Our model provided phenotypic adaptive landscapes for each landrace in future climatic scenarios and predicted that agronomic performance stability increases with Hs. Combining GP with eco-genetic predictions made it possible to identify promising landraces to improve adaptation to future or new cultivation conditions. TeaserIdentify promising landrace adapted to new and future environments by combining genomic selection and offset