Figure 1.

Exploring utility of machine learning in capturing the mutation signatures of SARS-CoV-2 for predicting severity outcomes. Mutation profile in a given SARS-CoV-2 genome is an important feature that may drive the course of infection. This can be engineered into derived features like epitope load created by the features. A numeric encoded (presence-absence) matrix of observed mutations along with patient age/gender/geography information for each genome can serve as an input data for machine learning (ML) methods. Supervised machine learning may therefore potentially enable prediction of infection severity by analyzing the patterns of important mutations in the large number of sequenced genomes and in the process, particularly through explainable machine learning techniques, enable identification of key features including mutations that drive the prediction.