Fig. 1

Schematic flowchart of the data production. a A total of 12,503 different viral protein sequences representing 1432 different viruses known to infect primates were downloaded from Virus-host DB [41] b. The viral protein sequences were cut into small (nine amino acid) overlapping peptides by NetMHCpan, resulting in 5,788,712 peptide sequences. c We used NetMHCpan to predict the binding abilities of 219 different bonobo and chimpanzees MHC class I molecules for the loci A, B, C and A-like to all viral peptides in the dataset. Cartoons of different MHC molecules and their predicted peptide binding repertoire are shown at the bottom of the figure. d For each MHC molecule we removed all redundant peptides, e.g. for peptides which were bound more than once at a particular MHC molecule, we kept only one representative peptide. In total, our dataset contained 2,791,353 peptide sequences, each with a unique amino acid sequence