Using Protein-Protein Interactions to Understand Drug Phenotypes

Wednesday, February 19, 2020 -
2:00pm to 3:00pm
FUNG Auditorium
Jennifer Wilson


Postdoctoral Scholar

Chemical and Systems Biology Department

Stanford University

Using Protein-Protein Interactions to Understand Drug Phenotypes


Adverse drug reactions (ARs) are among the top-10 reasons for death in the US (Lazarou, JAMA, 1998) and are estimated to cost up to $30.1 billion annually (Sultana, J Pharmcol Pharmacother, 2013).  While clinical trials test novel therapeutics in patients before marketing, the average drug exposure is only 1500 individuals (Friedman, JAMA, 1999) and clinical trials are unable to fully characterize a drug’s safety profile. In silico techniques, specifically protein interaction network methods, have uncovered associations of drug targets to AR phenotypes including severe ARs that could remove drugs from the market. Towards understanding potential safety risks, we developed a novel network interaction algorithm, PathFX, and a complementary web-server (PathFX-web) in collaboration with FDA scientists. We demonstrated that PathFX identified associations to efficacy and safety phenotypes for marketed drugs, and further that safety and efficacy associations shared interaction pathways. Yet, it is unknown if these pathway associations represent any mechanistic insight about the sources of drug-induced ARs. Through meta-analysis, we investigated patterns across networks of drugs with the same DME. After discovering shared interactions across these drugs, we considered that non-DME associated drugs could influence DMEs if they bound proteins within these shared interaction pathways and would represent novel drug combination effects. For instance, this analysis predicted that an anti-thrombotic agent would combine with a subset of anti-psychotics to exacerbate myocardial infarction, which is listed on the antipsychotic drug labels as a warning or precaution. Using a natural language processing method and published drug combination datasets, we’ve identified preliminary support for this drug combination effect. These results suggest that shared interaction paths contain mechanistic-like insights into sources of drug-induced adverse events, and that these paths represent a novel type of “long-ranging” drug combination effect.


Jennifer Wilson, PhD, is currently a fellow with the SPARK program at Stanford where she is exploring how network methods can enhance selection of druggable protein targets for improved therapeutic efficacy. Previously, she was a UCSF-Stanford Center for Excellence in Regulatory Science and Innovation (CERSI) postdoctoral fellow. With Dr. Russ Altman (Stanford University, 2016), she developed the PathFX algorithm to identify potential safety signals for drug targets in development.  She completed an ORISE fellowship at the US FDA with the Genomics and Targeted Therapy Group working with Dr. Michael Pacanowski (fall, 2017) where she tested PathFX as an aid for regulatory review of novel products. She developed the PathFXweb application for non-computational scientists and the web application is currently still in use by FDA reviewers. She completed a one-year rotation at Genentech where she supported the Tecentriq program in Product Development-Regulatory (fall, 2018), and developed a QSP approach to explore the translation of novel in vitro hematopoietic toxicity assays to clinical rates of cytopenias (spring, 2019). She has a bachelor’s degree in Biomedical Engineering from the University of Virginia and a PhD in Biological Engineering from MIT, where she worked with Dr. Doug Lauffenburger. In her PhD, she demonstrated that network algorithms uncovered hidden gene candidates from RNAi screens and validated these predictions in two cancer contexts: acute lymphoblastic leukemia and growth-factor-driven cancer systems.