QSP in the era of ML

Integrating AI/ML techniques with quantitative systems pharmacology (QSP) models presents promising opportunities and challenges. This presentation explores several families of approaches leveraging AI/ML for QSP at varying levels of increasing complexity. These include: 1) Using ML to enrich informational context and landscape analysis at the start of QSP modeling efforts. 2) Developing ML surrogate models trained on QSP model outputs to enable efficient exploration of complex scenarios. 3) Constructing hybrid QSP-ML models that fuse mechanistic biological understanding with data-driven models. 4) Applying ML methods to infer network topologies and automate the translation of data to dynamic QSP models. 5) Utilizing prompt engineering with large language models to streamline and augment various aspects of QSP model development. Key challenges are also discussed, including interpretability to ensure transparency, robust assessment frameworks, and the need for standardization - all critical for regulatory acceptance of AI/ML-driven QSP models. Promising opportunities are balanced against formidable hurdles in this emerging interdisciplinary field.