Identifying Key Long-COVID Biomarkers with Machine Learning Analysis

According to the CDC, 1-in-5 adult COVID-19 survivors suffer from widespread symptoms referred to as "Long-COVID". Long-COVID symptoms vary greatly between patients which results in increased disease complexity and diagnostic/therapeutic challenges. This project aims to identify the key Long-COVID biomarkers to improve clinical practice while also increasing our understanding of the disease. Although several general mechanisms have been proposed to explain the symptoms, they are limited by a lack of Long-COVID specific biomarkers and pathophysiological mechanisms. The varying symptoms make it challenging and inefficient to conduct research that targets specific physiological regions or mechanisms. As blood flows throughout the body, measuring blood biomarkers enables the investigation of multiple system changes and interactions that contribute to Long-COVID. A total of 2943 blood biomarkers will be measured for healthy controls, COVID-19 Ward inpatients, COVID-19 ICU inpatients and Long-COVID outpatients. I will analyze the large dataset with machine learning algorithms (AI) to identify the most important biomarkers that differentiate Long-COVID subjects from the other groups. Moreover, I will identify where these biomarkers are expressed to determine key organ systems and cell types affected by Long-COVID. Expression analysis in this manner will be a novel application of AI and will advance the field of blood biomarker analysis. As there is currently no test for Long-COVID, the identified biomarker model can serve as a diagnostic test. Individuals affected by COVID-19 can be proactively screened for their risk of Long-COVID and receive preventative measures. The blood biomarkers also enable targeted treatments and new streams of inquiry that investigate the identified biomarkers. The biomarker model developed will impact the decision-making of physicians and result in better patient outcomes.