Fig. 1.

Existence of exogenous mimics of pro-inflammatory host antimicrobial peptides (xenoAMPs) in SARS-CoV-2 proteins. (A) SARS-CoV-2 proteins are scanned with a machine learning AMP classifier. Each queried sequence is given a σ score that measures its AMP-ness. Three representative high-scoring sequences are studied: xenoAMP(ORF1ab), xenoAMP(S) and xenoAMP(M). The grey bars mark the location where the corresponding sequences are selected. (B) SARS-CoV-2 sequences are aligned and compared to their homologs in a common cold human coronavirus HCoV-OC43: Control (ORF1ab), Control(S) and Control(M). Asterisks, colons, and periods indicate positions that have fully conserved residues, those that have strongly similar properties, and those that have weakly similar properties, respectively. Color is assigned to each residue using the ClustalX scheme. (C) σ score heatmaps compare the distribution of high scoring sequences in three proteins from SARS-CoV-2 and HCoV-OC43. The first amino acid in each sequence is colored according to its average σ score; regions with negative average σ scores (non-AMPs) are colored white. “Hot spot” clusters of high-scoring sequences for SARS-CoV-2 (bright yellow regions bracketed in red boxes) have systematically higher scores and span wider regions of sequence space compared to HCoV-OC43. This trend suggests that hot spots in SARS-CoV-2 can generate higher scoring sequences for a greater diversity of enzymatic cleavage sites than those in HCoV-OC43.