Precision Medicine and Machine Learning to predict critical disease and death due to Coronavirus Disease 2019 (COVID-19)

Abstract

Abstract The early stage of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection, may be accompanied by high activity of the nucleotide-binding domain, leucine-rich repeat and pyrin domain-containing protein 3 (NLRP3) inflammasome and a cytokine storm. The aim of the study was to construct Machine Learning (ML) models that predict critical disease, severity of Coronavirus Disease 2019 (COVID-19), and death due to COVID-19. This cross-sectional study recruited 528 COVID-19 patients divided into those with critical (n = 308) and non-critical (n = 220) disease. The ML models included baseline imaging, demographic, and inflammatory data as well as NLRP3 (rs10754558 and rs10157379) and IL18 (rs360717 and rs187238) genetic variants. Partial least squares analysis showed that 49.5% of the variance in severity of critical COVID-19 can be explained by SpO2 and the sickness symptom complex (SSC) (inversely associated), chest computed tomography alterations (CCTA), inflammatory biomarkers, severe acute respiratory syndrome (SARS), body mass index (BMI), type 2 diabetes mellitus (T2DM), and age (all 7 positively associated). In this model, the four NLRP3/IL18 gene variants showed significant indirect effects on critical COVID-19 which were completely mediated by inflammatory biomarkers, SARS, and SSC. Neural network models, which entered SSC, SARS, CCTA, SpO2, age, T2DM, hypertension, inflammatory biomarkers and gene variants, yielded a significant prediction of critical disease and death due to COVID-19 with an area under the receiving operating characteristic curve of 0.930 and 0.927, respectively. Our ML methods increase the accuracy of predicting the severity, critical illness, and mortality caused by COVID-19 and show that the genetic variants contribute to the predictive power of the ML models.