In classical pharmacology, drugs struck single notes, where one drug would hit one target to treat one disease. But drugs frequently modulate entire “chords” of targets at once, and this can be essential to their action. Our lab is decoding this molecular music, both in terms of new and useful chords for the treatment of complex diseases, and also to identify the jarring notes that existing drugs unintentionally hit when they induce side effects.
Much remains unknown about multi-target drug activity. Is it best to block two molecular targets strongly, or to triangulate weaker inhibition at many targets, such that activity only arises by their overlap? We’re developing new systems pharmacology methods to investigate the molecular target binding profiles of drugs and their downstream biological consequences. For this, we are adapting the Similarity Ensemble Approach (SEA) and new deep learning methods to predict molecular targets and drug effects in model systems.
Decoding phenotypic screens
How might the application of systems pharmacology methods to phenotypic screens extend or even contradict known biology? We’ve found that no existing dataset or approach comprehensively links cellular, tissue, or organismal drug-induced phenotypes with their mechanistic molecular targets. So how might we anticipate when a drug’s effect arises from action at multiple targets simultaneously, and when one target is sufficient?
To investigate this, we are using phenotypic screens to deconvolute the mechanism of action targets of drugs and novel small molecules. With polypharmacology predictions as a guide, we can focus over 2.6×1014 possible target combinations down to a more manageable number, for testing in zebrafish (with the Kokel lab), in cell models of cancer, and in neurodegenerative diseases. One key goal is to actively mine for overlooked and discrepant biological mechanisms.
Precision drug response
How does our understanding of pharmacology vary across patients, and treatments over time? Precompetitive drug safety data, FDA adverse event reports, and electronic medical records (EMRs) can provide us with broad population baselines. But only by linking actual patient drug responses with genomic markers and pharmacological predictions can we identify patient-by-patient variations. And what about time? Is pharmacology more properly a path function? Drug response is a function of time, yet researchers often treat a drug’s efficacy as a timeless (equilibrium) property instead of a process. To what extent does a drug’s efficacy depend on the drug regimen preceding it?