Topology based learning models for SARS-CoV-2 mutation analysis

Abstract

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) resulted in a global pandemic after its first appearance in December 2019 remains circulating in our society. Since then, many mutations had emerged such as Alpha, Beta, Gamma, Delta, Omicron and many other mutations even with the discovery of vaccinations. Therefore, the possibility of a mutation of higher infectivity appearing exists as long as the virus remains in circulation. The ability to predict infectivity of possible mutations remains crucial in protecting lives. Infectivity is measured by the interaction between the receptor-binding domain (RBD) on the S-protein of SARS-CoV-2 (antibody) and the Angiotensin-Converting Enzyme 2 (ACE2) on human cells(antigen). Both of which are proteins and can be generally classified as protein-protein interactions in an antigen-antibody complex. Topological descriptors generated by persistent homology captures the intrinsic biological information of protein-protein interactions upon mutation through the extraction of essential features from high-dimensional dataset. Machine learning model employed like the Gradient Boosting Tree (GBT) incorporates topological descriptors to predict changes in binding affinity upon mutations. The changes in binding affinity indicates if the mutation has strengthened in its infectivity.