Large language models complement established workflows for mechanistic model construction and refinement in systems biology

Octomod, a multi-agent LLM workflow, matched a published mechanistic model without user guidance and enabled interactive refinement and data-driven reduction across four systems biology problems.

• Why it matters: Mechanistic model building is interdisciplinary and hard to reproduce, and existing tools lack transparent, user-guided refinement that still leverages biological knowledge and hypothesis testing.
• What they did: Octomod integrates biological knowledge retrieval, model construction, parameter estimation, and hypothesis testing using AMICI and pyPESTO with SBML and PEtab, while recording every input, check, and user decision.
• The result: Across three published-data applications and one lipidomics case, it matched a published model, supported interactive gene-regulatory refinement, explained responses of five immune cell types to stimuli, and reduced a supplied lipidomics model to measurable processes.

AI summary of the abstract · may contain errors · not medical advice · read the paper.