New Constraint to handle fusion / Rearrangement / translocation variants

[DanielPuthawala]

Discussion thread to host discussion on the development of a new constraint or constraints to handle Fusion variants, one of the largest and most clinically actionable (see Wagner lab manuscript in prep) variant categories left in the Cat-VRS variant test set. Similar coverage is needed also to handle rearrangement and translocation variants.

One very basic outstanding question is whether we think we need a constraint specifically to handle fusions, or if fusion variants are 'simply' a constrainted subset of rearrangment variants.

I will mock up something in this space soon based on one or more of the fusion variants in the test set. If you are an expert or know someone who is an expert on the fusions, I would love to get your input here or meet with you on a call so you can tell me everything we should know from a clinical perspective when modelling these concepts.

[brendanreardon]

Just adding some relevant papers and resources:

• Bruford, E. A. et al. HUGO Gene Nomenclature Committee (HGNC) recommendations for the designation of gene fusions. Leukemia vol. 35 3040–3043 (2021)
• Wagner, A. H. et al. Recommendations for future extensions to the HGNC gene fusion nomenclature. Leukemia vol. 35 3611–3612 (2021).
• HGVS general recommendations
• VICC gene fusion specification

[DanielPuthawala]

Question: in this figure here (from the 2021 HUGO paper), (large?) sequence deletions are counted as a type of unbalanced rearrangment in that two hitherto non-adjacent bits (the borders of the deletion) are now found adjacent to each other. Are all deletions therefore commonly considered to be a subset of rearrangements? Is this an equivalency that we need to make transparent in Cat-VRS?

[brendanreardon]

Are all deletions therefore commonly considered to be a subset of rearrangements? Is this an equivalency that we need to make transparent in Cat-VRS?

I'm interpreting "all" as rearrangement deletions, copy number deletions, and (short) nucleotide deletions. From my understanding, this is not true and thus is not an equivalency that we need to model. Instead I tend to think of it as a venn diagram with some overlap. The key distinction is probably scale, with a lot more overlap between rearrangement deletions and copy number deletions.

• Rearrangement deletions are when the flanking regions are joined. These also tend to be large, involving huge portions of chromosomes.
• Copy number deletions really are talking about if there is loss of genetic material at a given location, which can be focal (at scales like an exon or a few genes) or broad (generally refers to it being a large chunk of a chromosomal arm). These can happen with or without a rearrangement.
• Nucleotide deletions tend to be just a small number of bases and impact just a single gene or region. With NGS we can really only detect "short" insertion or deletions, which tend to be on the scale of tens or low hundreds of bases. One of the reasons that people are excited about long-read sequencing is that it should help us detect longer insertions and deletions and small structural variants, which will help us identify events that blur the lines between these different categories.

If it's...

• 1s - low 100s of bases: nucleotide deletion
• 100s - megabases: copy number deletion
• 1000s of bases or more: structural variant deletion

This is a nice review on aneuploidy / the copy number side of things, in cancer.