Columbia informatics expert reports on integrative data techniques for screening drug interactions


Published March 8, 2018 This content is archived.

The February 1 Clinical and Translational Science Institute (CTSI) Seminar Series addressed the topic of “Observational data for biomedical discovery.” 

Electronic Health Records contain a wealth of data for selecting prospective clinical studies

“Integrative data science produces verifiable drug safety. ”
Nicholas Tatonetti, PhD, Herbert Irving Assistant Professor of Biomedical Informatics and director of Clinical Informatics
Herbert Irving Comprehensive Cancer Center at Columbia University

Nicholas Tatonetti, PhD, the Herbert Irving Assistant Professor of Biomedical Informatics and director of Clinical Informatics at the Herbert Irving Comprehensive Cancer Center at Columbia University, spoke to an audience of about 100 scholars, researchers and students in UB’s Pharmacy Building about an exciting biomedical informatics approach for detecting dangerous drug-drug interaction (DDIs) using available Electronic Health Records (EHRs).

Tatonetti was the invited guest of the CTSI and the School of Pharmacy and Pharmaceutical Sciences, sponsors of the eventHis group is well-known for its expertise in developing algorithms, mathematical techniques and computational methods that integrate hospital data with high-dimensional biological data to correct for bias and confounding in observational analyses. 

He said that 10 to 30 percent of adverse drug effects reported to the FDA can be attributed to DDIs. However, those reports fail to account for the basic fact that “drugs are biased toward side effects caused by their indication.” A heart medication that is frequently co-prescribed with other common heart medications is associated with heart attack. A diabetes drug regimen is similarly linked to high incidences of hyperglycemia, a symptom of the very disease the medication is used to treat. These are among the kinds of serious confounding errors that often go undetected.

Tatonetti’s method can correct for bias by using traditional biomedical techniques to test effects predicted by their computations. 

“Integrative data science,” concluded Tatonetti, “produces verifiable drug safety.” 

A lively question-and-answer session followed the presentation, including a discussion of practical considerations for accessing and interpreting EHR data and the feasibility of using that data for testing heritability of DDIs without genetic testing.

CTSI Seminars feature important topics in clinical and translational science presented by outstanding speakers. Each seminar is co-hosted by one of the five UB health sciences schools.