The development of climate-resilient crops would be greatly accelerated by models able to reason directly over plant genomic sequences and to pinpoint trait-associated regions or loci. Anticipating the impact of DNA base changes (variants) remains challenging, and understanding regulatory mechanisms is still an active area of research. Through self-supervised training on unannotated genomic data, genomic language models (gLMs) can learn DNA syntax and grammar that go beyond current annotations, thus complementing standard bioinformatics analyses that rely on rules established by decades of genomics research. Here we present our agent-powered Model Factory and its first outputs: the Botanic1 family of gLMs designed for plant research, which operates reliably on sequences from hundreds of base pairs up to 128 kbp. These models outperform all generalist and plant-specific gLMs (as well as specialised baselines) on one of the largest sets of plant genomics evaluation tasks reported to date, at a much smaller budget than concurrent models. Mechanistic interpretability analysis identifies features associated with biologically meaningful sequence properties including coding region boundaries and splice site motifs, demonstrating that these models are a source of biological insight beyond their benchmark performance. Finally, because a gLM only becomes practically useful when embedded in a broader workflow, we integrate Botanic1 as a specialised genomic layer callable by a generalist large language model (LLM) agent, illustrating how such hybrid systems could accelerate plant biology research. To support the plant genomics research community, we release the four Botanic1 models, their pre-training corpus and the trained sparse autoencoder for research use at https://huggingface.co/spaces/living-models/botanic1-report.
A. Barozet, V. Cabeli, J. D. du Terrail et al.· bioRxiv· 0 citations
Gene regulation depends on coordinated interactions between promoters and distal cis-regulatory elements, yet understanding how these regulatory elements communicate remains a fundamental challenge in mammalian genomics. Chromatin interaction assays provide one approach for identifying potential regulatory relationships, but interpreting the biological significance of individual interactions remains difficult; chromatin interactions comprise multiple biologically distinct classes that are only partially captured by any single assay. Here, we integrate Hi-C, RNAPII ChIA-PET, and CTCF ChIA-PET with the ENCODE Registry of candidate cis-regulatory elements (cCREs) and complementary functional genomic datasets to develop an integrative framework for classifying and interpreting promoter-centric chromatin interactions. Using this framework, we identify a distinct class of candidate architectural promoter-enhancer interactions that are characterized by increased recurrence across cellular contexts, broader promoter connectivity, and reduced dependence on linear genomic proximity. We further show that many regulatory elements anchoring these interactions transition between enhancer and CTCF-only states while maintaining stable chromatin interactions. These dual-state regulatory elements also acquire context-specific transcription factor inputs within evolutionarily conserved architectural scaffolds, suggesting that stable chromatin architecture can be repeatedly repurposed for new regulatory functions. Genes connected to these dual-state regulatory elements are enriched for developmental and signaling pathways and exhibit increased expression specificity across cell types, consistent with specialized roles in context-dependent gene regulation. Together, our findings provide a biologically informed framework for classifying and interpreting chromatin interactions and support a model in which conserved chromatin architecture provides a stable foundation upon which new regulatory programs evolve.
