Computational Prediction of the Comprehensive SARS-CoV-2 vs. Human Interactome to Guide the Design of Therapeutics

Kevin Dick; Kyle K Biggar; James R Green

The rapidity of our response is thanks, in part, to having produced an analogous study during the emergence of the Zika Virus outbreak of 2015, where our sequence-based PPI prediction method (PIPE) was used to identify putative human-Zika inter-species PPIs and inform possible synthetic biology approaches for novel interventions and therapeutics [9] . In the present study, of the ∼285,000 predicted pairs, we selected only the highly conservative set of predicted interactions for each of the 14 SARS-CoV-2 proteins resulting in the identification of 225 putative human protein targets. We publicly released these predictions and related meta-data for use by the broader scientific community in the following DataVerse repository: 10.5683/SP2/JZ77XA, [10] . The dataset of experimentally elucidated human-virus PPIs was obtained from the VirusMentha database [11] . These 10,693 known PPIs are used to train the PPI predictors and infer new putative interactions between human proteins and the SARS-CoV-2 proteome. For the all schema, the proteomes of the 43 viral families were collected from Uniprot and are summarized in the appendix Table 4 . In anticipation of the generation of the predicted interactome using the proximal schema, we tabulate the 689 training PPI and the Group IV viral families over which they are distributed (Table 1) . Finally, the human reference proteome (UP000005640) was obtained from Uniprot, retaining only the high-quality "Reviewed" Swiss-Prot proteins.