MIDAS COVID-19 Special Webinar Series: The perils and promise of machine learning for life-threatening infection – J. Scott Van Epps
Wednesday, June 24, 2020
1:00pm - 2:00pm
The ongoing COVID-19 pandemic has brought the chasm between the speed of viral spread and the speed of scientific discovery in to stark reality. Our fear of the devastating outcomes and desire to change the course of this pandemic disease has led to undisciplined interpretation of the limited data available. This particular event is not the first of its kind, nor will it be the last. Emerging infections, whether from novel viruses or known bacteria that have developed resistance to the only therapies we have against them, will require continued vigilance, scientific discovery, and rigorous validation of new treatments for the foreseeable future. Machine learning holds promise to accelerate the discovery and development of novel diagnostics, prognostics, and therapies for emerging infectious disease. Indeed, the combination of the electronic health record and ‘omics’ biomolecular information provides a data-rich environment to feed machine learning algorithms. However, the heterogeneity of the data and paucity of well-defined labels makes this problem quite unwieldy. As a result, specific applications may lack generalizability, foster false interpretations, and perpetuate a slow adoption of machine learning to clinical medicine and biomedical research. With this background in mind, I will provide an overview of some ongoing projects that show the breadth of topics in the field of infectious diseases that are amenable to machine learning.