en.Wedoany.com Reported - In the keynote speeches on the second day of the 2026 Bio-IT World conference, multiple experts discussed the application of foundation models and federated learning in drug discovery. Panelists included Dr. Woody Sherman, Founder and Chief Innovation Officer of PsiThera; Dr. Mohammed AlQuraishi, Assistant Professor of Systems Biology at Columbia University; Dr. Jonathan Gilbert, Senior Director of Ecosystem Growth and Contributor Collaboration at Eli Lilly; Dr. José-Tomás Prieto, Director of AI Projects at Apheris; Dr. Arman Zaribafiyan, Head of AI Simulation Strategic Alliances at SandboxAQ; and Dr. Christina Taylor, Senior Research Scientist and Head of Computational Molecular Design at Bayer.

Sherman emphasized that open platforms help drive progress in the field of drug discovery, and the industry should jointly build foundation models, then leverage federated learning for further optimization.

AlQuraishi noted that AlphaFold has significantly advanced protein structure prediction in the short term, but its modeling of non-protein components remains limited. He introduced OpenFold as an open-source alternative to AlphaFold, comprising over 40 institutions, aiming to provide a reproducible public platform. OpenFold is both a toolkit and an academic-industrial consortium, offering a steerable mechanism for the industry.

Gilbert introduced that Eli Lilly launched the TuneLab platform in 2025, offering AI models free of charge to biotechnology companies, but requiring them to contribute training data. The value of data invested by Lilly in model training exceeds one billion dollars. The platform currently has over 75 partners and is used in dozens of countries across three continents. Lilly has also reached an agreement with Collaborative Drug Discovery (CDD) to integrate TuneLab into the CDD Vault platform.

Prieto discussed the complexities in implementing federated learning, pointing out that the new paradigm is that data preparation does not need to be centralized, and each company must coordinate standards and computing environments. The AI Structural Biology Network supported by Apheris brings together multiple top-20 biopharmaceutical companies, allowing AI models to train on proprietary protein data while ensuring data security.

Zaribafiyan emphasized that fine-tuning can significantly narrow the gap between general models and real-world drug discovery use cases. SandboxAQ collaborated with OpenFold and announced the launch of a new platform that connects quantitative models with large language models, enabling scientists to initiate simulations and workflows using natural language.

Taylor noted that community-driven foundation models have accelerated innovation in the field of biomolecular AI. Bayer uses these models to resolve crystal structures, improve drug development and crop science traits, and study protein interactions and simulate enzyme catalysis.

This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com