Predicting interactions between the SARS-CoV-2 spike glycoprotein and the human proteome using AphaFold and molecular dynamics simulations

Abstract Since the COVID-19 pandemic began, the SARS-CoV-2 virus has caused over 775 million cases and more than 7 million deaths worldwide. Despite progress in treatments and vaccines, we still need better ways to prevent and treat the disease. To do this, a clearer understanding of how SARS-CoV-2 interacts with human proteins is needed. We developed a new computational tool to predict and study these protein-protein interactions. Our method uses two stages of AlphaFold, an AI tool, to predict how SARS-CoV-2 proteins bind to human proteins, followed by molecular dynamics simulation to refine these predictions. We tested this method in a small study focusing on the S1 subunit of the SARS-CoV-2 spike protein interacting with human cell junction and synaptic proteins, and in a larger study screening the spike S1 N- terminal domain against the entire human proteome. We validated the method using experimental virus–human protein interactions. The results provide valuable insights into how SARS-CoV-2 interacts with human proteins, guiding future experiments to better understand COVID-19’s short- and long-term effects. This developed method could be utilized to study protein-protein interactions in other biological systems.