AI Tool Interfaces Cut Tumor Digital Twin Build Time from Months to Minutes

The npj Systems Biology and Applications paper details open MCP servers that translate natural-language queries into calls on established modeling platforms. Researchers tested iterative prompting on three language models and showed convergence on core parameters such as proliferation rates and oxygen diffusion after repeated refinement rounds. Absolute time dropped from an estimated 120–180 person-hours to under 10 minutes for an initial draft, though final calibration still requires domain expertise and wet-lab validation.

This advance aligns with wider efforts to embed mechanistic modeling inside oncology workflows, including FDA guidance on digital twins for trial simulation issued in 2024. Earlier single-lab attempts at automated PhysiCell scripting remained unpublished or limited to narrow use cases; the BSC approach generalizes across tools and releases the servers publicly, addressing reproducibility concerns through documented prompt histories.

Remaining questions center on whether models generated this way will match the predictive accuracy of manually curated twins when tested against independent patient cohorts. Prospective studies comparing AI-drafted versus expert-drafted models on progression-free survival endpoints are needed within two years.