Gaps between medical biology and AI drug discovery
Yueming Yina b✉, Afu Fuc, Lipo Wangc d✉
- aInstitute for Digital Molecular Analytics and Science, Nanyang Technological University, 636921, Singapore. Electronic address: yueming.yin@ntu.edu.sg
- bInstitute for Digital Molecular Analytics and Science, Nanyang Technological University, 636921, Singapore. Electronic address: yueming.yin@ntu.edu
- cInstitute for Digital Molecular Analytics and Science, Nanyang Technological University, 636921, Singapore
- dSchool of Electrical and Electronic Engineering, Nanyang Technological University, 639798, Singapore. Electronic address: elpwang@ntu.edu.sg
- ✉Corresponding author
Papers open inside SciRah from the original publisher or repository; SciRah never hosts them.
Abstract
Drug discovery is a complex, iterative process spanning biology, chemistry, pharmacology, and computational sciences. Artificial intelligence (AI) can accelerate this process but often misaligns with biological realities. Here, we highlight three crucial gaps in AI-driven drug discovery. First, conflating binding affinity with bioactivity ignores distinct experimental setups. Second, reliance on simplified bioactivity metrics limits the ability of AI models to capture richer biological contexts. Third, AI applications remain fragmented, addressing isolated tasks rather than integrating insights across pipeline stages. To bridge these gaps, we propose a biologically contextualized AI framework and provide guidelines for researchers in both medical biology and AI drug discovery.
TL;DRThis work proposes a biologically contextualized AI framework and provides guidelines for researchers in both medical biology and AI drug discovery, highlighting three crucial gaps in AI-driven drug discovery. (Semantic Scholar)