Why does the MHC class II binding prediction operate only on 15-mers?

Originally from ticket #2212.

Question:
I have been trying to use the MHC II binding algorithms, and there seems to be just one length (15 AA) of peptides in the output. I am curious, is there a specific reason for this? From my understanding the length of class II peptides can vary between 13-20, or was it a technical decision?

Answer:

Yes, class II alleles can bind with lengths other than 15 in real. However, the length is fixed at 15 for IEDB prediction class II binding prediction tool. Although class II alleles bind to other lengths, 15 seems to be an optimal and preferred length. 15mer peptides will reasonably capture 9-mer binding core and flanking residues.
The IEDB is planning to incorporate other lengths in the tool in near future. Meanwhile, you can use NetMHCIIpan server to do predictions for other lengths (http://www.cbs.dtu.dk/services/NetMHCIIpan/)