ABSTRACT
SARS-CoV-2 has been deregulating society for over two years. Whereas targeted interactions allow the virus to recruit signaling pathways in order to facilitate replication, fortuitous host-pathogen interactions lead to the emergence of covid-19 at various levels of severity. The CAPRI community has been mobilizing its resources and expertise to model the structures and interaction interfaces of SARS-CoV-2 to human protein complexes. A comprehensive list of putatively interacting protein pairs has been prioritized from a Y2H-determined protein-protein interaction map, out of which five targets were offered for prediction in a uniquely open initiative, sharing data, analyses and results as they were being produced. Close to 30 predictor groups produced close to 30 GB of data, representing more than 50,000 three-dimensional interaction models. These have been processed and filtered by approximately the same amount of scorer groups using a plethora of scoring functions to produce an enriched subset of some 1000 structures. We have further reduced this set through clustering and contact overlap scoring to produce a high-resolution community consensus prediction for the host-pathogen interactions offered. The results show a varying degree of reliability for the targets, with generally a better consensus for the interaction surface on the human protein than on the SARS protein. The results also show the feasibility of such large-scale approaches and the added value of using several, distinctly different, prediction methodologies to reach a consensus prediction.