DelineateQSP: Accelerating Quantitative Systems Pharmacology Model Building and Repurposing

Delineate Inc, Medford MA

Investigators

Abstract

Delineate Inc.’s proposed STTR Phase I project aims to develop an innovative platform that uses Large Language Models (LLMs) to accelerate the building and repurposing of Quantitative Systems Pharmacology (QSP) models from scientific literature. QSP involves the creation of mechanistic models that can de-risk the drug development process, improve clinical trial outcomes, and aid in the smarter design of novel treatments but the process of constructing models is heavily bottlenecked which currently limits their scientific impact. Through this project, Delineate Inc. in partnership with the Julia lab at MIT, aim to validate the use of LLMs and other AI methods to overcome the challenges modelers face when building models, such as efficiently extracting literature data and repurposing models for new use cases. The project has three main objectives: (1) develop an LLM-powered platform for efficient QSP model construction from literature, (2) validate the platform by building test QSP models and conducting time studies to demonstrate a significant reduction in model building time, and (3) enable model provenance tracking and prediction of invalid assumptions for accurate model repurposing. The successful completion of this project will lay the foundation for a commercially viable solution that can greatly improve the efficiency of the QSP model building process and accelerate the development of new life-saving therapeutics.