Explain

[MattWellie]

https://main-web.populationgenomics.org.au/ohmr3-mendelian/reanalysis/2022-08-29/summary_output.html Sample 13Y03696

ASPH variants which don't form a reciprocal comp-het, not exactly sure of the reason

[MattWellie]

Current behaviour opposite to expected, even in latest report version https://main-web.populationgenomics.org.au/ohmr3-mendelian/reanalysis/2023-01-16/summary_output.html

Variant coords ending *095 looks the more likely to be found as AD (completely absent from population data)

[MattWellie]

This is a non-issue. Investigated:

• The variants in question are both Cat. 1, so the MOI test is run under a partial penetrance model.
• The family contains 2 samples, both affected, but no relationship in the pedigree - no dummy parent/child, just a shared family ID.
• The variant 8:61548140 occurs in both family members, so it passes the dominant filter.
• The 8:61526095 variant is only present in one participant, the other is WT. As it does not explain the phenotype across all affected family members, it does not succeed under a dominant model.
• The participant with both single variants has 2 more entries in the report, representing the compound het with each variant as a 'primary' (the orientation is not important here) - even though the comp-het does not explain the whole family's phenotype (an affected member lacks the variant pair).

---

I thought this would be an error case, as the comp-het doesn't explain the whole family's phenotype, so it should be dismissed. That's not actually how the logic is written -

• for each primary and partner variant we only dismiss the variant when both variants are present in a single parent
• in this case the samples share a family ID, but not familial structure, so we can't construct an informative relationship
• No parent with both variants - test passes

TL; DR - I think this is a fair way to run the tests. We could also use a flat family structure to say 'a comp-het variant pair has to explain all affected family members or it is dismissed' rather than only testing the parents.

---

Bonus 'issue' - This gene probably explains both family members, in an irritating way which gets at a crack in the logic here:

1. The gene model for ASPH permits both Dominant and Recessive inheritance.
2. 8:61548140 is Hom in one participant and Het in the other.
3. It passes the Dominant filter as that can explain both affected family members (Hom variants are considered as both Het and Hom in parallel, so this passes on the basis of 'at least one alt allele in each member').
4. The Hom variant does not pass the recessive filter, as the Hom doesn't explain both affected participants. Even with partial penetrance this cannot be rescued in the current logic.
5. The Het sample has a second-hit elsewhere in the gene, which causes a compound-het. This also doesn't explain the whole family, but we only check inheritance from parents so it passes.
6. If the gene model here was only Recessive, rather than Dominant and Recessive, the Homozygous Clinvar Pathogenic call would not have reached the report.

This isn't easy to solve without undermining the tight MOI tests currently in place. This is a teeny tiny corner case that can't occur in Singletons or Trios (majority of our datasets), but in a duo or quad it's a possibly missed diagnosis.

[cassimons]

Great effort stepping through all of the logic here! My read is that it seems to be doing exactly what we want it to. Yes there will be corner cases like this one where the outcome is slightly unintuitive, but I think that is ok.

In terms of the family structure, I am hesitant to move to a flat family structure as if we do not have a pedigree then the chances of the phenotyping also being rubbish are much higher. Keeping it as is seems to be the way to go.