#29 · Genomics

Ensemble learning and multi-environment GWAS advance crop trait prediction and mapping

The common thread is moving association signals toward prediction and interpretation for complex agronomic traits in crops. A multi-environment GWAS of faba bean reports more than 100 yield and stability loci and pairs them with a random regression model that predicts performance in untested environments. GEG2P addresses prediction directly, using genetic-algorithm-optimized ensemble learning over 20 base learners and reporting a 4.02% accuracy gain across five crops. PanvaR, a preprint, addresses post-GWAS interpretation by integrating fine-mapping with LD, SNP effect predictions, and nearby candidate genes per peak in an R package. The new element is the combination of environment-aware modeling, ensemble optimization, and integrated peak-level reporting, though the three items differ substantially in scope: two are published studies with quantified results, while PanvaR is a software tool without reported benchmark performance in the given summary.