Abstract 1096: Harmonization standards from the Variant Interpretation for Cancer Consortium
Bibliographic record
Abstract
Abstract The use of clinical gene sequencing is now commonplace, and genome analysts and molecular pathologists are often tasked with the labor-intensive process of interpreting the clinical significance of large numbers of tumor variants. Numerous independent knowledgebases have been constructed to alleviate this manual burden, however these knowledgebases are non-interoperable. As a result, the analyst is left with a difficult tradeoff: for each knowledgebase used the analyst must understand the nuances particular to that resource and integrate its evidence accordingly when generating the clinical report, but for each knowledgebase omitted there is increased potential for missed findings of clinical significance.The Variant Interpretation for Cancer Consortium (VICC; cancervariants.org) was formed as a driver project of the Global Alliance for Genomics and Health (GA4GH; ga4gh.org) to address this concern. VICC members include representatives from several major somatic interpretation knowledgebases including CIViC, OncoKB, Jax-CKB, the Weill Cornell PMKB, the IRB-Barcelona Cancer Biomarkers Database, and others. Previously, the VICC built and reported on a harmonized meta-knowledgebase of 19,551 biomarker associations of harmonized variants, diseases, drugs, and evidence across the constituent resources.In that study, we analyzed the frequency with which the tumor samples from the AACR Project GENIE cohort would match to harmonized associations. Variant matches increased dramatically from 57% to 86% when broader matching to regions describing categorical variants were allowed. Unlike precise sequence variants with specified alternate alleles, categorical variants describe a collection of potential variants with a common feature, such as “V600” (non-valine alleles at the 600 residue), “Exon 20 mutations” (all non-silent mutations in exon 20), or “Gain-of-function” (hypermorphic alterations that activate or amplify gene activity). However, matching observed sequence variants to categorical variants is challenging, as the latter are typically only described as unstructured text. Here we describe the expressive and computational GA4GH Variation Representation specification (vr-spec.readthedocs.io), which we co-developed as members of the GA4GH Genomic Knowledge Standards work stream. This specification provides a schema for common, precise forms of variation (e.g. SNVs and Indels) and the method for computing identifiers from these objects. We highlight key aspects of the specification and our work to apply it to the characterization of categorical variation, showcasing the variant terminology and classification tools developed by the VICC to support this effort. These standards and tools are free, open-source, and extensible, overcoming barriers to standardized variant knowledge sharing and search. Citation Format: Alex H. Wagner, Reece K. Hart, Larry Babb, Robert R. Freimuth, Adam Coffman, Yonghao Liang, Beth Pitel, Angshumoy Roy, Matthew Brush, Jennifer Lee, Anna Lu, Thomas Coard, Shruti Rao, Deborah Ritter, Brian Walsh, Susan Mockus, Peter Horak, Ian King, Dmitriy Sonkin, Subha Madhavan, Gordana Raca, Debyani Chakravarty, Malachi Griffith, Obi L. Griffith. Harmonization standards from the Variant Interpretation for Cancer Consortium [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 1096.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.296 | 0.284 |
| Meta-epidemiology (narrow) | 0.001 | 0.002 |
| Meta-epidemiology (broad) | 0.002 | 0.003 |
| Bibliometrics | 0.011 | 0.013 |
| Science and technology studies | 0.004 | 0.003 |
| Scholarly communication | 0.014 | 0.005 |
| Open science | 0.013 | 0.013 |
| Research integrity | 0.006 | 0.007 |
| Insufficient payload (model declined to judge) | 0.008 | 0.009 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; the direct Gemma label and the distilled Codex classifier agree on what is shown here.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".