Simplicial Flow Matching for Drug Discovery

Canada CIFAR AI Chairs Guy Wolf (Mila, Université de Montréal) and Renjie Liao (Vector Institute, UBC) are developing a novel end-to-end approach to retrosynthesis using simplicial flow matching. Retrosynthesis — the process of breaking down target molecules into simpler precursors — is a fundamental bottleneck in drug discovery that traditionally requires expert chemists and expensive trial-and-error laboratory work.

This research matters because automated retrosynthesis could dramatically accelerate the drug development pipeline. By applying techniques from topological data analysis and generative modeling, the approach captures molecular structure relationships that simpler graph-based methods miss. Funded through CIFAR's high-risk, high-reward AI Catalyst Grants, the project exemplifies Canada's strength in fundamental AI research with clear applied potential.

The strategic context is that AI-driven drug discovery is a multi-billion-dollar global race, and Canada has structural advantages through the density of AI talent in close proximity to major pharmaceutical and biotech research. This work also bridges the gap between pure mathematics and applied chemistry — a cross-disciplinary approach that Canadian institutes are uniquely positioned to foster.