In this talk, Jean-Philippe Vert, Co-Founder and Chief Executive Officer (CEO) of Bioptimus explores the transformative role of foundation models in accelerating drug discovery and biological research. He explains how self-supervised learning allows massive neural networks to identify patterns in genomics and histopathology images without human annotation. These advanced AI systems demonstrate a superior ability to predict molecular subtypes and treatment responses compared to traditional methods. Vert highlights that the current challenge lies in integrating different scales of biology, such as single-cell data and tissue organization, into a unified multimodal model. Ultimately, he envisions a "virtual cell" capable of simulating complex interactions to improve the success rate of clinical trials. This shift towards data-driven simulation represents a new era for computational medicine.

‍