Data-Driven Precision Pharmacology

We are making drugs safer through the analysis of data. Everyday millions of us or our loved ones take medications to manage our health. We trust in these prescriptions to improve our lives and give us hope for a healthier future. Often, however, these drugs have harmful side effects or dangerous interactions. Adverse drug reactions are experienced by millions of patients each year and cost the healthcare industry billions of dollars. In the Tatonetti Lab we use advanced data science methods, including artificial intelligence and machine learning, to investigate these medicines. Using emerging resources, such as electronic health records (EHR) and genomics databases, we are working to identify for whom these drugs will be safe and effective and for whom they will not. Browse our databases, contribute to our projects, and join us on this journey to make precision pharmacology a realty.

TLab dives into SARS-CoV-2 Research

May 4, 2020

As a (mostly) computational lab, we count ourselves lucky to continue our work without too much disruption during the pandemic. It is probably because of this situation and the fact that we all are part of a community hit hard by the virus, that we are feeling extra motivated to do what we can to help. These past two weeks have been big for the lab. We now have three manuscripts on SARS-CoV-2 research out for review. These include a paper that replicates a previous finding associating blood groups with infection status, a study that identifies new clinical and genetic risk factors for SARS-CoV-2 outcomes, and a paper that investigates the role of ACE inhibitor exposure on infection rates.

These works were made possible by a lot of sleepless nights from Michael, Undina, Vijendra, and Phyllis. ­čĺ¬

Congratulations to Dr. Thangaraj!

February 18, 2020

Phyllis Thangaraj, an MD/PhD student in the lab, successfully defended her dissertation entitled “Electronic health record-derived phenotyping models to improve genomic research in stroke.” Her work explores the use of machine learning methods to transform noisy binary indicators of disease status common in electronic health records to quantitative descriptors. Dr. Thangaraj’s central hypothesis is that these quantitative descriptors can provide more nuance to disease definitions that will enhance genetic studies that rely on these data to discovery new genomic variants. Her two papers are currently in submission and can be read as pre-prints here and here.

Tatonetti Lab pre-releases NSIDES

November 15, 2019

An updated version (NSIDES pre-release v0.1) of our OFFSIDES/TWOSIDES databases is now available. As with the previous versions, this database contains adverse drug effects for single drugs (OFFSIDES) and drugs used in combination with others (TWOSIDES). This version (v0.1) of OFFSIDES contains adverse event predictions for over 9 million drug-event pairs from 3,394 drugs and 17,552 adverse events. Approximately 125,000 of those drug-event pairs are statistically significant (~36 significant adverse events per drug). For TWOSIDES, over 222 million drug-drug-event triplets are evaluated with 5.7 million significant putative drug interactions. The full release of NSIDES will contain drug-drug-drug and higher order interactions of drugs as well and is currently being computed. The data are made available in csv format and as an MYSQL database dump.

Read all the release documentation on the nsides-release github repository.

Download the data.

Thanks to Dr. Rami Vanguri for leading the project and Michael Zietz for organizing the release data and information!