Michael Keiser, Ph.D. investigates how small molecules perturb protein target networks to biological and therapeutic effect. In a forward polypharmacology campaign, we are using machine learning and computational methods such as the Similarity Ensemble Approach (SEA) to infer new combinations of targets underlying compound-induced phenotypes in cells and broader model systems. Thinking of each target as a musical note, we predict and test entire chords at a time via chemical-genetic epistasis experiments in models of complex diseases such as neurodegeneration.
These systems pharmacology inquiries tell us where current understanding of drug action is weak, and because we use human pharmacology to mine for mechanisms, we can apply the therapeutic chords we find directly to the prediction of patient drug responses, adverse drug reactions, and personalized medicine.