BOTANIC-1: a series of long-context plant genomic foundation models in the agentic era

Developing climate-resilient crops requires models that can reason directly over plant genomic sequences and pinpoint trait-associated loci. Predicting the impact of DNA base changes remains difficult, and understanding regulatory mechanisms is still an active research area. Genomic language models (gLMs) trained on unannotated data can learn DNA syntax and grammar that go beyond current annotations, complementing standard bioinformatics analyses.

Here the authors describe their agent-powered Model Factory and its first outputs: the Botanic1 family of gLMs designed for plant research. These models reliably operate on sequences from hundreds of base pairs up to 128 kbp, and are reported to outperform all generalist and plant-specific gLMs—as well as specialized baselines—on one of the largest sets of plant genomics evaluation tasks to date, all at a much smaller budget than concurrent models.

Mechanistic interpretability analysis identified features associated with biologically meaningful sequence properties, including coding region boundaries and splice site motifs, indicating that the models provide biological insight beyond their benchmark performance. To make the gLM practically useful, Botanic1 is integrated as a specialized genomic layer callable by a generalist large language model (LLM) agent, illustrating how such hybrid systems could accelerate plant biology research.

The authors will release Botanic1-S, Botanic1-M, and Botanic1-L for research use via Hugging Face.