Mining peptides against SARS-CoV-2 entry from constructed RBMs-hACE2 isomer libraries using machine learning and molecular docking

Abstract

Cellular entry of SARS-CoV-2 initiates from the protein-protein interactions (PPIs) between viral surface protein S and human angiotensin converting enzyme 2 (hACE2). Peptide-based drugs have the advantage of small molecule compounds to block such viral-host PPIs. Thus the viral targetregions on hACE2 have been believed as promising templates for designing specific inhibitory peptides against SARS-CoV-2 infection. However, starting from a few potential templates, in silico design and prediction between binding affinity and bioactivities in vivo are very challenging, herein a novel design strategy was implemented by mining constructed template isomer libraries using feature filters, supervised classifier and peptide protein docking.Applying these methods and the isomer libraries, 4 peptides were identified from 12 millions candidates owing to their distinct stability, interaction activity, inhibitory specificity, binding affinity, transmembrane potentials and effective conformation. These results have supplied a panel of specific anti-COVID19 leads for further drug development, supporting a new feasible antiviral strategy for targeting both intracellular and extracellular SARS-CoV-2 S proteins simultaneously. The methods have provided a useful tool for mining antiviral-peptides against viral diseases.