Search papers, labs, and topics across Lattice.
This paper introduces APOD, a language-model agent that leverages biological knowledge and fit diagnostics for open-ended ODE discovery to build population digital twins (PDTs). APOD addresses limitations of existing methods by explicitly modeling inter-individual variability and searching a broader model space, achieving rapid and accurate recovery of ground-truth ODE structures on synthetic benchmarks. The approach successfully identified a PDT for radioligand-therapy-induced platelet dynamics, enabling prediction of thrombocytopenia and simulation of toxicity-reducing dosing strategies from early clinical data.
Language models can autonomously discover complex pharmacological ODEs, outperforming existing methods by 12x and enabling predictive digital twins from sparse, noisy clinical data.
Establishing ordinary differential equations (ODEs) describing population data is a fundamental part of mathematical modeling in pharmacology, crucial to developing digital twins. However, doing so from sparse, noisy data is a slow, expert-driven task. Existing automated methods either search a restricted model space or ignore population inter-individual variability. Here we introduce APOD (Agentic Population ODE Discovery), a language-model agent that iteratively reasons over biological knowledge and fit diagnostics in an open-ended search space to discover a population digital twin (PDT), i.e., a shared ODE system with between-subject variability. On synthetic pharmacokinetic and tumor-dynamics benchmarks, APOD recovered ground-truth structures in 94-100\% of runs, 12-fold faster in median than an established library-based search. On real cohorts it converged to valid structures, and proposed a PDT of radioligand-therapy-induced platelet dynamics that predicts thrombocytopenia from first-cycle data and simulates alternative dosing schedules that lower the predicted risk of toxicity.
