Application of machine learning to study opioid addiction
Date
2022
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
Historically, drug abuse and addiction has been thought of as a failing of the moral character. But recent studies have highlighted that contrary to the prevailing thought it is a disease of the brain. Hence, like any other disease it must be studied and those suffering from it must be given access to proper care and treatment. In recent years, opioid addiction in the United States has reached an epidemic proportion and continues to accelerate despite significant interventions by several federal and non-federal agencies. The factors driving opioid abuse and addiction are varied and complex. They include a wide array of socio-contextual and individual factors that interact in complex manner across both time and space. Hence, in this thesis I use machine learning technologies coupled with data collected from a wide range of databases to study opioid addiction at both the societal and individual level. ☐ In chapter 1, I provide an overview of the literature on addiction that characterizes it as disease that needs to studied like any other disease. I also provide an overview of the history of the opioid addiction including the events that lead to the current crisis of opioid addiction in the United States. In chapter 2, I lay the foundation for studying opioid addiction at a phenotypic level, by developing a machine learning model to learn from biomedical knowledge graphs. I then validate it's utility to predict kinase-substrate interactions by comparing it with existing knowledge graph learning methods. In chapter 3, I use this machine learning model to predict kinases that are involved in desensitization of the opioid receptors. I then investigate the utility of these kinases to discover novel drugs that can be used to reverse tolerance to opioids. When treating any health conditions, it is always advisable to follow a preventive approach rather than a reactive approach. Hence, in chapter 4, I develop a machine learning model to help proactively identify the population that could benefit the most from these novel therapeutic drugs. I also use the latest advancements in machine learning interpretability techniques to analyze the factors that the model considers important in an effort to gain a further understanding of the relation between these factors and the opioid epidemic. Finally, in chapter 5, I summarize the results of my thesis and offer suggestions for future directions of this research.
Description
Keywords
Opioids, Machine learning, Opioid addiction