Regarding the peptides selection of the latest percentile data calculation for IEDB MHC I and MHC II binding prediction methods, we downloaded the "Reviewed (Swiss-Prot)" dataset from https://www.uniprot.org/downloads in FASTA format on 10/29/2018. The file contained 558,712 sequences. 555,970 of them were acceptable by tools of IEDB, and 544,147 of them had length of at least 50 aa. Then we randomly selected 10,000 of these protein sequences and further randomly selected peptides with specified lengths (8-15 for class I; 10-30 for class II) from each protein sequence.