When does a prediction earn your trust? A fireside chat with Essam Metwally from Merck

A prediction looks convincing. Is that actually a reason to trust it? For most discovery teams working with structure prediction tools like AlphaFold and Boltz-2, that’s becoming a regular judgement call, and getting it wrong can mean chasing a lead that never had a chance. Essam Metwally, Scientific Director at Merck Research Laboratories, has spent more than 20 years on both sides of that question, first solving structures experimentally, now leading modelling, AI/ML and data science teams. That combination gives him an expert, grounded view of when these tools earn a team’s trust and when they don’t. In this fireside chat, Essam joins Nathan Brown for a candid conversation about the state of the latest AI methods in drug discovery. When does a prediction earn the right to change what a team does next, rather than simply looking convincing? Why do promising models so often disappoint once they meet real data and real objectives? And as automation advances, what should scientists never hand over? Join us to learn where leaders should focus their effort to capture real value rather than chase the hype. Timestamps: 00:00 Introductions 01:44 Essam Metwally's Journey: From Bench to Computation to Merck 07:12 Where AI Really Stands in Drug Discovery Today 14:03 When Does a Prediction Earn Your Trust? Data, Models & Human Judgment 35:34 Audience Q&A & Closing Thoughts