Dysregulation of kinase function has been a mechanism for small molecule inhibitors entering clinical trials for many diseases, including cancer, arthritis, and parasitic infections. Using phenotypic screening approaches, ATP-competitive inhibitors of a kinase of interest were identified. Our team combined 3D-shape analysis, inspired by proposed binding modes for our phenotypically discovered inhibitors, with a multitask, two step machine learning prediction method, pQSAR, for cellular activity predictions. Compounds with a similar shape to our scaffold of reference, in addition to cellular pQSAR pIC₅₀>6, were active in biochemical and cellular assays. Combining 3D descriptors and pQSAR was an effective virtual screening approach for selecting a small set of compounds (<1,000) for further evaluation, when compared to more “hands-on” approaches, e.g., SAR-by-archive. Compounds with cellular pQSAR pIC₅₀<6 showed a general trend towards weak biochemical inhibition. However, a few compounds coming from an internal program targeting a different kinase, which had pQSAR predictions 5<pIC₅₀<6, turned out to be highly active in biochemical and cellular assays. These outliers have a conserved amide near a bicyclic aromatic core, and X-ray structures were available for this series in the other kinase. In this off-target kinase, the amide accepts a critical H-bond from the catalytic lysine, whereas our binding model for the kinase of interest predicted that the amide would interact with the hinge region. We synthesized analogs of these outlier inhibitors with replacements for the amide. These modifications were designed with docking to hinder H-bond interactions with the off-target catalytic lysine. As predicted, they proved active against the kinase of interest, while dialing out inhibition of the off-target. We will highlight a strategy for using 3D-shape and pQSAR for an effective identification of kinase inhibitors, as well as a structurally guided pathway to overcome kinase off-target activity

Combining 3D and ML methods in drug design projects

Wilian Augusto Cortopassi

Novartis

Appearances