Artificial intelligence predicts the immunogenic landscape of SARS-CoV-2: toward universal blueprints for vaccine designs

Brandon Malone; Boris Simovski; Clement Moline; Jun Cheng; Marius Gheorghe; Hugues Fontenelle; Ioannis Vardaxis; Simen Tennoe; Jenny-Ann Malmberg; Richard Stratford; Trevor Clancy

Many of the previous SARS-CoV studies have found promising CD8 targets [9, 15, 18, 24] , including sustainable memory T cell responses [9, [15] [16] [17] [20] [21] [22] [23] that recogise epitopes in proteins across the entire spectrum of the virus, although the S-protein have been reported to be enriched for dominant CD8 T cell responses [24] . Taken together, this supports the approach taken in this study, which is to map computationally, a broad epitope landscape across the global viral SARS-CoV-2 proteome, which includes integrated CD8, CD4 and B cell targets in the modeling. There has been some preliminary efforts submitted into preprint servers recently that describe epitope maps generated [29] [30] [31] , however it appears that the emphasis in those approaches were based mostly on HLA binding. It is important to profile in whole viral proteome epitope screens, as carried out in this study using an extensive artificial intelligence platform, not only the candidates that may bind to the HLA molecule but also those CD8 epitopes that are naturally processed by the cell's antigen processing machinery, and presented on the surface of the infected host cells. Layered on top of the antigen presentation predictions in the host infected cells, we also make predictions across the entire viral proteome that measure the likelihood that the peptides presented on the host infected cells are capable of being recognized by T cells that are not yet tolerized or deleted from a patient's T cell repertoire. The subsequent immunogenic landscape of the SARS-CoV-2 that we present here is taken further to analyze the immunogenicity of all the non-synonymous variations across approximately 3400 different SARS-CoV-2 sequences, to map author/funder. All rights reserved. No reuse allowed without permission.