Grant Number:	1U24CA305456-01 Interpret this number
Primary Investigator:	Griffith, Obi
Organization:	Washington University
Project Title:	Accelerating the Expert-Crowdsourcing of Cancer Variant Interpretation in Civic
Fiscal Year:	2025

Project Summary Precision oncology involves the use of prevention and treatment strategies tailored to the unique features of each individual cancer patient and their disease. The number of molecular alterations or “variants” identified as cancer drivers or linked to cancer prognosis, diagnosis, or drug response has exploded. As a result, cancer care-givers are faced with a deluge of patient-specific variants that must be interpreted in the context of a vast and growing biomedical literature describing their significance. Currently, these variant interpretations exist largely in private or encumbered databases resulting in extensive repetition of effort. Widespread adoption of precision medicine requires this knowledge to be centralized, standardized and expert-curated for application in the clinic. To address this need, we created CIViC, a community-driven web resource for Clinical Interpretation of Variants in Cancer, available online at civicdb.org. CIViC is uniquely distinguishable from other resources due to its fully open access, rich data model, and large community of volunteer expert curators. CIViC has been widely adopted by the community with many individual users, incorporated into numerous academic and commercial workflows, as the official curation platform for ClinGen Somatic variant curation, and as a Global Core Biodata Resource. Due to widespread adoption, CIViC has seen a dramatic increase in the numbers of users integrating CIViC into their workflows and submissions of new content, which require expert moderation and review. This proposal will sustain and build on the success of CIViC by developing and implementing new features to accelerate dissemination of high-quality knowledge, and automate and increase efficiency of biocuration and moderation. The curation interface will be improved for users including new variant matching modules, cancer gene classifications, pre-submission of evidence to support journal submissions, and automatic classification of pathogenicity or oncogenicity according to established guidelines. We will develop an editorial toolbox to increase efficiency of expert editor workflows. This will include dashboards for tracking progress on specific classes of moderation, more discrete editor level permissions to allow editorial sub-tasks to be delegated, new tools for automating variant coordinate curation based on curated allele registry identifiers, and others. We will develop sophisticated new natural language processing (NLP) tools to automatically “fact-check” key components of new submissions to the database, automatically reject problematic or unsalvageable submissions, automatically annotate key supporting statements from source publications, and score submitted evidence for estimated accuracy to prioritize moderation efforts. Finally, we will engage in significant outreach, education and collaborative activities to support editor recruiting, training and incentivization. This will include hackathons, workshops, and new online educational materials developed in collaboration with the ITCR training network.




Evaluating Language Models for Biomedical Fact-Checking: A Benchmark Dataset for Cancer Variant Interpretation Verification.
Authors: Reisle C. , Grisdale C.J. , Krysiak K. , Danos A.M. , Khanfar M. , Pleasance E. , Saliba J. , Hanos M. , Patel N.V. , Jain A. , et al. .
Source: Biorxiv : The Preprint Server For Biology, 2025-09-15 00:00:00.0; , .
EPub date: 2025-09-15 00:00:00.0.
PMID: 41000838
Related Citations