2025 · Models
PlantCAD2 released for the plant-genomics community
A 676M-parameter model trained on 65 angiosperm genomes that outperforms models more than 10× its size on cross-species tasks.
We mine the natural diversity of plant genomes to find the DNA changes behind complex traits — and we teach machines to read that diversity, building foundation models that turn raw sequence into biological understanding for maize and thousands of other species.
Four connected threads run from the diversity in wild and cultivated genomes to the sustainable crops agriculture will need under a hotter, more variable climate.
Any two maize varieties differ more than humans do from chimpanzees. We map that variation across thousands of landraces and wild relatives — even across species — using genome–environment association to connect DNA to the climates plants are adapted to.
/02PlantCAD2 and GeneCAD read DNA the way language models read text — capturing evolutionary conservation and predicting gene structure directly from sequence across the flowering-plant tree of life.
/03Most nitrogen applied to maize is lost to the environment. Through CERCA, we're designing nitrogen-efficient, cold-tolerant corn that starts earlier, stays green longer, and slashes fertilizer loss, emissions, and cost.
/04Ideas become software the community can use. The Practical Haplotype Graph, TASSEL, and imputation tools bring high-tech breeding within reach of programs from Cornell to sub-Saharan Africa.
Recent releases, preprints, and milestones from the group.
A 676M-parameter model trained on 65 angiosperm genomes that outperforms models more than 10× its size on cross-species tasks.
Sequence-only genome annotation that removes the need for costly transcriptomic evidence — and extends to animals.
The CERCA / Nitrogen 2.0 vision for redesigning maize to slash fertilizer loss while lifting yield.