Dependence of variant rankings on other variants?

[cbohrson]

Hi, and thanks for making this tool. Suppose exomiser is evaluating variation in a gene and considering a compound heterozygous, autosomal recessive model of inheritance, and suppose the final report includes two variants from the same gene. Is the ranking of these two variants elevated by the fact that 2 variants are observed in the same gene (given compound het/ autosomal recessive model is being evaluated)?

This is relevant to me because I would like to compare existing exomiser results against a new set of variant calls for which running exomiser is not possible.

[cbohrson]

Would really like an answer on this if you have time and apologies if its somewhere in the docs and I missed it.

More context through an example scenario: suppose a patient's disease is an AR mendelian disorder caused by an intronic retrotransposon (RT) insertion on one allele and a nonsense mutation on the other. Exomiser is run on the results of best practices short variant calling (e.g., GATK) which in this case is not sensitive to the RT insertion.

The nonsense mutation is seen by exomiser, and the gene-disease linkage is robust. What is the expected behavior of exomiser for reporting the nonsense mutation given only one mutation of 2 necessary for the AR model is observed?

[julesjacobsen]

Hi @cbohrson, sorry not to have responded sooner and thanks for asking a great question.

Ordinarily, under an AR model Exomiser expects either a homozygous alt (1/1) or two heterozygous (0/1) variants to have passed all the filtering steps for the full phenotype score to be applied. In this case, if the disease-gene association is AR and only one heterozygous variant is present, the phenotype score will be halved. This will lead to a low overall combined score, even if the nonsense variant gives a variant score of 1. Lastly, in the compound het case the gene variant score will be the mean of the two variants.

If you are able to spike in the RT variant (I guess it's an ) in a test VCF with the other variant of interest, Exomiser will take that into account and hopefully retain it in the final results. Assuming you're running Exomiser 13.3.0+, SV insertions in the splice regions are scored highly, but this also depends on how rare the variant is and whether or not is has been classified in ClinVar. Previous versions of Exomiser over-score variant scores, so use version 13.3.0+ for these.

If you want to follow up over email, feel free to contact me at my QMUL account.