Michael Raymer, Ph.D. (Advisor); David Cool, Ph.D. (Committee Member); Michael Leffak, Ph.D. (Committee Member); Jeffery Gearhart, Ph.D. (Committee Member); Courtney Sulentic, Ph.D. (Committee Member)
Doctor of Philosophy (PhD)
G protein-coupled receptors (GPCRs) initiate intracellular signaling pathways via interaction with external stimuli. [1-5] Despite sharing similar structure and cellular mechanism, GPCRs participate in a uniquely broad range of physiological functions.  Due to the size and functional diversity of the GPCR family, these receptors are a major focus for pharmacological applications. [1,7] Current state-of-the-art pharmacology and toxicology research strategies rely on computational methods to efficiently design highly selective, low toxicity compounds. ,  GPCR-targeting therapeutics are associated with low selectivity resulting in increased risk of adverse effects and toxicity. Psychoactive drugs that are active at Class A GPCRs used in the treatment of schizophrenia and other psychiatric disorders display promiscuous binding behavior linked to chronic toxicity and high-risk adverse effects. [16-18] We hypothesized that using a combination of physiochemical feature engineering with a feedforward neural network, predictive models can be trained for these specific GPCR subgroups that are more efficient and accurate than current state-of-the-art methods.. We combined normal mode analysis with deep learning to create a novel framework for the prediction of Class A GPCR/psychoactive drug interaction activities. Our deep learning classifier results in high classification accuracy (5-HT F1-score = 0.78; DRD F1-score = 0.93) and achieves a 45% reduction in model training time when structure-based feature selection is applied via guidance from an anisotropic network model (ANM). Additionally, we demonstrate the interpretability and application potential of our framework via evaluation of highly clinically relevant Class A GPCR/psychoactive drug interactions guided by our ANM results and deep learning predictions. Our model offers an increased range of applicability as compared to other methods due to accessible data compatibility requirements and low model complexity. While this model can be applied to a multitude of clinical applications, we have presented strong evidence for the impact of machine learning in the development of novel psychiatric therapeutics with improved safety and tolerability.
Department or Program
Year Degree Awarded
Copyright 2021, all rights reserved. My ETD will be available under the "Fair Use" terms of copyright law.