Plant breeding prediction using genomic language models
Genomic prediction has transformed plant breeding, yet gains in prediction accuracy often plateau, despite continued increases in marker density and model sophistication.
These plateaus increasingly point to limitations in how genomic information is represented, rather than to an exhausted genetic signal or insufficient data.
Advances in functional genomics highlight the importance of regulatory, structural, and context-dependent sequence variation, which is only indirectly captured by marker-based models.
Genomic language models have recently emerged as a sequence-based framework capable of learning contextual genomic representations, prompting renewed discussions about how biological information is encoded for genomic prediction.
https://www.cell.com/trends/plant-science/fulltext/S1360-1385(26)00096-8
