This can happen only if the difference in resolution is not considered sufficiently relevant in the given context by Hap-E Search and ATLAS. Below are some possible (not mutually exclusive) reasons: - The allele designations encode for the identical ARD (antigen recognition domainAntigen Recognition Domain) and Hap-E Search and ATLAS only matches for the ARD part of the HLA protein complex. For example B*07:ANVB stands for B*07:02/07:61 which both encode for the same ARD (here: exon 2 and 3). Hence Hap-E Search and ATLAS is regarding the codes B*07:02, B*07:61 and B*07:ANVB all as allele matches for each other but might give different matching probabilities for phenotypes containing each of those three codes since haplotypes involving B*07:02 and B*07:61 are having individual frequencies.
- The resolution only looks higher but, in fact, is lower (e.g. B*15:12 currently is only a single allele while B*15:12:01G covers three), identical (e.g. A*80:01, A*80:01:01 and A*80:01:01G cover exactly the same alleles) or not directly comparable since each code is containing some alleles not covered by the other (e.g. comparing B*07:TDVB and B*07:02:01G the former also covers a lot more variants of B*07:02 like B*07:02:02 to B*07:02:05 while it is excluding other alleles like B*07:44 and B*07:49N).
- The broader code only contains additional alleles which have never been observed in the population whose haplotype frequencies are used - at least not in a relevant haplotypical context. Their frequency could also be so low that it is disregarded due to rounding or the 10% probability grouping explained elsewhere. This applies, in particular, to all null variants of expressed alleles with identical first two field designations.
|