Cell Foundation Models in Drug Discovery

=== ENTRY DETAILS ===

- QR code with entry information coming soon. It'll be available on this Event details site.
- Gate closes at 18:15 - no late entries.

======================

Cell foundation models are being hailed as the "GPTs of biology," promising to solve diverse tasks with a single model. It feels like we've just had the GPT-2 moment for cell biology.

But is the hype justified?
In this talk, we'll take a critical look at the current landscape of cell foundation models and the drug discovery related benchmarks used to evaluate them. Our findings suggest that:

1. Many popular benchmarks are poor proxies for the real biological tasks we care about.
2. Surprisingly simple baseline models often outperform these sophisticated foundation models.

We'll walk you through our analysis, explore why these models may not be working as expected, and share concrete ideas for building better, more meaningful benchmarks.

If you're interested in machine learning, single-cell biology, or the future of AI-driven drug discovery, this talk is for you.

We'll start with a quick overview of the field to provide context for all who are coming from a non biology related domain.