Learn about groundbreaking research in AI-driven drug discovery through this technical talk that explores Meta Flow Matching (MFM), a novel approach for modeling dynamic biological and physical systems. Dive deep into how flow-based models can be enhanced to learn population-level dynamics, particularly crucial for personalized medicine applications. Discover how MFM integrates vector fields on the Wasserstein manifold by utilizing Graph Neural Networks to embed sample populations, enabling generalization across initial distributions. Explore the practical applications of this methodology in predicting individual treatment responses using large-scale multi-patient single-cell drug screen datasets. The presentation demonstrates how this innovative approach addresses the limitations of current flow-based models, especially when dealing with multiple initial populations and varying conditions that describe different dynamics in natural sciences.