The abstract is read at the publisher; the summary is JClub's.
PlantCAD2, a plant-specific DNA language model with 676 million parameters, captures evolutionary conservation and surpasses Evo2 in 10 of 12 zero-shot tasks across 65 angiosperm genomes.
- Why it matters: Angiosperms span ~200-fold genome-size variation with compact coding regions, creating a need for DNA models that can interpret plant genomes using long-range sequence context.
- What they did: PlantCAD2 was pre-trained at single-nucleotide resolution on 65 angiosperm genomes, then evaluated with zero-shot benchmarks and parameter-efficient fine-tuning across cross-species tasks.
- The result: With an 8,192-bp context window, PlantCAD2 improves maize accessible chromatin prediction (AUPRC 0.587 to 0.711) and outperforms AgroNT on seven cross-species tasks, supporting long-range distal regulation modeling.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.