Nicholas P. Tatonetti, Stanford University
January 24, 2012 @ 3:30 - 4:30 pmLocation: Blockley Hall - Room 701
Biostatistics
TITLE: Biomedical discovery through integration and analysis
of clinical and molecular data.
Large observational clinical databases represent a new opportunity to study disease states and drug effects in vivo. However, their utility is limited by selection biases -- a well recognized challenge of observational analysis. Further, analyses of these data is limited to discovering clinical risk factors for disease development and drug response. I hypothesize that by addressing the issues of bias in these large clinical databases and by integrating molecular measurements of metabolites, protein expression, genetic variants, and chemical features we will advance our understanding of basic disease biology and pharmacology. I have demonstrated the feasibility of this hypothesis by developing methods to (1) remove the bias in clinical data introduced by unmeasured confounding variables, (2) improve the detection of hidden or latent drug effects, (3) identify candidate genes involved in drug response, and (4) validate these findings using retrospective and prospective studies. I exemplify my approach by presenting the detection and validation a novel drug-drug interaction between paroxetine, a popular anti-depressant, and pravastatin, a popular cholesterol lowering drug. I will demonstrate how I discovered the interaction in a database of adverse drug event reports even though the adverse interaction itself was never actually reported. I will then present the methods for how I validated the finding retrospectively by using the electronic medical records (EMR) at three geographically distant institutions and prospectively in an insulin-resistant mouse model. Finally, I will present methods for identifying the molecular mechanism by which this idiosyncratic drug interaction occurs.
