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

The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.04.21.052084 doi: bioRxiv preprint maps provide an overview of the most relevant regions of the SARS-CoV-2 virus that could be used to develop a vaccine. It is clear from Figure 1 that different HLA alleles have different Class I AP (and IP), and Class II binding properties. This strongly suggests, as one might anticipate, that the SARS-CoV-2 antigen presentation landscape (and IP) clusters into distinct population groups across the spectrum of different human HLA alleles. This trend is further illustrated in the hierarchichal-clustering map presented Figure 2 after the AP scores have been binarized. Figure 2 clearly demonstrates that some allelic clusters present many viral targets to the human immune system, while others only present a few targets, and some are unable to present any. This implies that different groups in the human population with different HLA's will respond differentially to a T cell driven vaccine composed of viral peptides. Therefore in order to design the optimal vaccine that leverages the benefits of T cell immunity across a broad human population we need to predict "epitope hotspots" in viral proteome. These hotspots are regions of the virus that are enriched for overlapping epitopes, and/or epitopes in close spatial proximity, that can be recognized by multiple HLA types across the human population.