In silico drug profiling of the human kinome based on a molecular marker for cross reactivity

Protein kinases are paradigmatic targets in molecular cancer therapy. Affinity profiles of kinase drug inhibitors are of considerable interest to assess and modulate clinical drug impact. In the initial stages of discovery, a thorough experimental screening of lead compound libraries becomes critically limited by the size of the kinase sample. This work introduces a computational screening approach which provides a tool for extensive screening that uses experimentally obtained small-scale profiles as input data and makes predictions for a larger kinase set. These predictions result from a propagation of the reduced profile, exploiting a structural comparison of kinases based on a feature-similarity matrix. The comparison focuses on a molecular marker for specificity and promiscuity of kinase inhibitors. Our approach enables the computational high-throughput screening of entire libraries of compounds to search for suitable leads, mapping their inhibitory impact on a sizable sample of the human kinome. Our in silico tool is validated by contrasting predictions against reported high-throughput screening experiments.