Is AI Capable of Real-World Drug Discovery?
TL;DR
This article examines how the close integration of computational and medicinal chemistry coupled with a deep understanding of chemical shape and protein interactions enabled targeted computational molecular design to produce disproportionate gains in efficacy and selectivity.
Abstract
Small-molecule drug discovery frequently operates in regimes where slight structural changes have large consequences. These are situations in which current artificial intelligence (AI) methods, trained on mass data, may perform poorly. For a substantial fraction of drug-discovery projects, limited biological understanding and sparse data further constrain progress and define regimes where AI approaches remain difficult to apply. Despite these limitations, the global investment in AI continues to grow rapidly. Using two case studies characterized by structural and biochemical data, this article examines how the close integration of computational and medicinal chemistry coupled with a deep understanding of chemical shape and protein interactions enabled targeted computational molecular design to produce disproportionate gains in efficacy and selectivity. These examples highlight a challenge for current AI methods: sensitivity to subtle, low-data perturbations. For AI to achieve a transformative impact in drug discovery, it must move beyond pattern recognition to understand, or explicitly simulate, the mechanistic “why” linking subtle structural changes to biological outcomes. Pending such advances, clear opportunities for AI to productively complement human efforts, rather than be used as a stand-alone solution, are delineated.