The HLA typing system we used covers the class I genes in their full sequence. With exception of DPB1, the class II loci are covered in their full sequence or a few exons are missing. The sequence fragments put in phase at least 300 bases. Given the number of SNPs, the sequences of almost all alleles were put in phase.

There were no ambiguities in the class I or class II loci with exception of DPB1. We used the IMGT sequence database version 3.25 for the HLA assignment. We had virtually no ambiguities in DPB1 because more than 65 percent of the subjects were homozygous in DPB1 resulting from the high frequency of DPB1*04:02 alleles (gf for DPB1*04:02:01:01 and DPB1*04:02:01:02 > 0.8). Among the few ambiguities observed, all of them resulted from genotypes that include alleles with identical sequences in exon-2 that differ in exon-3 and cannot be placed in phase because of lack of informative SNPs. Almost all ambiguous DPB1 genotypes (< 2%) included DPB1*04:02; we assigned the likely genotype as the one including this allele). The rate of DPB1 ambiguities has increased significantly with the addition of novel alleles after version 3.30 of the HLA nomenclature. The ambiguities result from the inability of shotgun sequencing to place in phase exons 2 and 3 from DPB1 because of the lack of SNPs in a long segment of intron2.

In summary, there were no ambiguities for all loci with exception of DPB1; the rate of ambiguity at DPB1 was low because of low allelic diversity at this locus in this population; the most likely genotype was assigned on the basis of allele distributions in unambiguous genotypes.