Incorrect handling when patient has null allele that maps to a small g group

[mmelchers]

Describe the bug When performing comparison between Hap-E and ATLAS the following issue was found. To Reproduce Steps to reproduce the behavior: create search with HLA:

perform 0 mismatch donor search in WMDA dev environment

Expected behaviour No donors returned with zero mismatch probability larger than 0%

Screenshots

Additional context Suggested info from the Hap-E developers: Of course, it has to do with the null allele in the patient typing. 04:09N is part of G code 04:01:01G. Below, I have listed the 5-locus genotypes for the first donor (in HF set 5). The difference between Hap-E and ATLAS lies in the most frequent genotype in the first line. Hap-E considers this genotype to have one mismatch to the patient on locus C: In the genotype, we have 04:01, while the patient has 04:09N. With high resolution null allele, we treat the patient as homozygous on locus C. So we would only consider a genotype with homozygous C08:02 or with C*08:02 + an explicit null allele to be a match.

ATLAS seems to classify the first genotype to be a match genotype to the patient – presumably because allele 04:09N is part of 04:01g. If I calculate f_genotype / f_overall = 3.80381E-07/4.29066E-07 = 88.6 I get the ATLAS p0 value.

[zabeen]

Hi @mmelchers Thanks for raising the ticket, I'll take a look ASAP

[zabeen]

Dev Testing Notes

• Regression test: run 10/10 search with match prediction on, and check it completes successfully and match prediction object is populated in results file.

• Bug fix test - create donor with HLA described in ticket, but homozygous for the C-expressing allele, and then run 10/10 search using C-null allele. Donor should come back as 10/10 match, and have predicted OverallMatchCount of 10.

[zabeen]

Bug fix testing

• created donor with HLA described in ticket, but homozygous for the C-expressing allele (08:02)
• ran 10/10 search using C-null allele (04:09N)
• Donor came back as 10/10 match, and had predicted MatchPredictionResult.OverallMatchCount of 10.

Test passed

[DmitriyShcherbina]

@zabeen Regression testing status: Ok