Guinevere Andrews
Research Mentor(s): Peter Toogood
Mentor Department: Medicinal Chemistry
Authors: Guinevere Andrews, Mark Newman, Austin Polanco, Peter Toogood
Session: Session 4 (1:00pm – 1:50pm)
Presentation Type: Poster 114
Abstract
Drug discovery is the process of finding and developing new drugs to treat disease. The
purpose of this research project is to use machine learning methods to accelerate drug discovery.
Part A focused on drug repurposing, finding new effective indications for already known drugs.
A set of predicted alternate uses for known drugs was derived from a network analysis of
available literature data. Interesting possible indications were found for the drugs pregnenolone,
prednisolone, tryptophan, butenafine, and griseofulvin. These drugs and their newly identified
indications went through extensive literature review which found that butenafine and
griseofulvin were the only compounds without current evidence to support their proposed new
indication. In part B, we identified a set of compounds of predicted utility in treating malaria,
where the majority of compounds were found to have no current research evidence that supports
their use in malaria and can be tested further. Part C focused on finding a completely new class
of drug for potential treatments of pancreatic cancer. A set of known malic enzyme (ME)
inhibitors was compiled as a training data. Next, several sets of machine learning generated
novel structures based on a representative known ME inhibitor were assembled for interrogation.
These compounds represent a test set. Experiments are now ongoing to identify the most likely
ME inhibitors from the test set using a model based on the training set.
