Rewriting Rare Disease R&D with Foundation Models

Drug development has long been a costly, trial-and-error effort, with nine out of ten clinical programs failing despite major scientific advances. One reason is that biological information remains fragmented in silos, and traditional R&D approaches often rely on narrow, task-specific datasets. Bioptimus aims to change this by using AI to build a foundation model that integrates multimodal, multiscale biological data into a single body of knowledge. The approach has particular promise for rare diseases, where patient numbers and data are scarce, preclinical models are poor, and development economics are challenging. We spoke with Jean-Philippe Vert, co-founder and CEO of Bioptimus, about the inherent messiness of biology, the potential to transform rare disease drug development with a foundation model, and how uncovering similarities between conditions could enable repurposing of existing drugs.