Discordance calculation and minimum information standards for sharing cancer variant interpretations.

GracePendlebury commented 1 month ago

Submitter Name

Grace Pendlebury

Submitter Affiliation

Shariant / Australian Genomics

Submitter Github Handle

No response

Additional Submitter Details

Shariant is a web-based centralized platform designed to automate the sharing of variant interpretations and associated evidence between diagnostic laboratories in Australia and New Zealand. We notify labs of any discordances between classifications, providing a triaging and resolution page, and facilitate submission to ClinVar for the labs (via API). I work as Project Officer for Shariant, and look after the recently established somatic arm.

Project Details

Topic suggestion for the 2024 Cancer Variant Curation and Coding Unconference. I am new to the CGC and the GA4GH Cancer Community, so these topics may have been raised previously (apologies if so!), but it would be very interesting to discuss consensus and opinions or be pointed in the direction of existing information.

For germline variants in Shariant, classification against the ACMG/AMP 2015 guidelines is mandatory, and discordance/concordance is calculated between three categories (P/LP vs VUS vs LB/B). We want to provide the same core functionalities for somatic cancer submissions (against ClinGen/CGC/VICC 2022 SOP and AMP/ASCO/CAP 2017), which is fairly straightforward for oncogenicity classifications (O/LO vs VUS vs LB/B), however clinical significance is more complicated. The AMP VITAL working group used 3 methods of calculating interoperator concordance in their 2023 somatic challenge (PMID: 36503149). They didn’t include the AMP evidence level (A-D) in any of those methods, despite the levels being from the original guideline publication and used commonly by knowledge bases (incl. many DBs harmonized in metaKB).

The minimum information standards (particularly for clinical significance) discussion is related, depending on the level of granularity that would be best practice to calculate discordance on (i.e. tier only vs tier+evidence level vs tier+evidence level+assertion type etc). We have also encountered some barriers in submitting somatic variants with a clinical impact classification to ClinVar, as labs are limited in what information they can submit to Shariant by what they can actually export from local curation systems. Getting the level (A-D) at all in an accessible format is not possible for most labs so far, let alone the assertion type (P/T/D) or drug required by ClinVar if submitting a Tier I or II.

If there is some international agreement on the most accurate (and implementable) way to proceed, finalizing data standards would make sharing of somatic variant interpretations more interoperable.