Skip to main content

Distinguished Professor of Computer Science Pierre Baldi has co-authored a new paper, “Deep Learning for Drug Discovery and Cancer Research: Automated Analysis of Vascularization Images,” with a group of researchers from UCI. Appearing in the May/June 2019 issue of the journal IEEE/ACM Transactions on Computational Biology and Bioinformatics, the paper presents a machine learning model that takes a first step toward automating data analysis for high-throughput drug screening.

As the paper notes, likely drug candidates identified in traditional pre-clinical drug screens often fail in patient trials, in part because traditional in vitro models of drug response don’t accurately mimic the complex properties of human biology. The authors introduce a new microphysiological system for growing vascularized, perfused microtissues. The system, suitable for large drug screens, more accurately models human physiology. The machine learning model that the authors developed can “quickly and accurately flag compounds that effectively disrupt vascular networks from images taken before and after drug application in vitro.” The system achieves “near perfect accuracy while committing potentially no expensive false negatives.”

— Shani Murray