Generative AI Proposes mTOR Inhibitor Leads That Still Need Laboratory Validation
Published in Artificial Intelligence, Drugs development & discovery, Journals.
A study published in Discover Chemistry on September 21, 2026 explores how generative AI can propose molecules targeting mTOR. The work received new media coverage on October 3; its original publication date is earlier.
Amisha Bisht and colleagues fine-tuned a recurrent neural network using 5,178 experimentally validated mTOR inhibitors from ChEMBL. They generated 200 candidate structures and screened them computationally using molecular docking, predicted drug properties, toxicity assessments and electronic-structure calculations.
Four molecules emerged as computational leads. S000031 was prioritized after the combined assessments. Docking scores were compared with the reference inhibitor Torin2, but these simulations do not establish superior anticancer activity, safety or efficacy.
The newly generated compounds were not experimentally validated in this study. Further computational work and laboratory testing are required before they can be considered validated inhibitors or therapeutic candidates. The findings illustrate an early discovery workflow rather than a clinical advance.
Source: Bisht et al., Discover Chemistry — original research, September 21, 2026. DOI: 10.1007/s44371-026-00994-x.