Julien Delile, Srayanta Mukherjee, Judith Mueller, Iya Khalil, Leonid Zhukov, Christoph A. Meier
We also discuss what foundation-model-powered drug discovery could look like in the future.
During the past decade, AI has evolved rapidly. Within the field, there has been particular innovation in the area of foundation models -- general-purpose AI algorithms that can be adapted to a broad range of tasks. Recently, researchers have started exploring how these foundation models can be applied to pharmaceutical R&D. In this review, we survey the landscape of foundation models in drug discovery research. We show that, starting in 2022, the number of foundation models has been growing extremely rapidly, with >200 such models published to date. We demonstrate that these cover a broad range of applications, including target discovery, molecular property optimization, preclinical applications and others. We also discuss what foundation-model-powered drug discovery could look like in the future.