ClinGen Curation of ClinVar: Improving a Critical Community Resource
Laboratory Genetics and Genomics
-
Primary Categories:
- Laboratory Genetics
-
Secondary Categories:
- Laboratory Genetics
Introduction:
ClinVar, a public NCBI-maintained database, has aggregated over 4.7 million classified variants, aiding clinical genomic interpretation and research. However, ClinVar has >28,000 variants with medically significant conflicts (pathogenic/likely pathogenic (P/LP) vs uncertain significance/likely benign/benign), many due to outdated or inaccurate submissions. The Clinical Genome Resource (ClinGen) develops authoritative resources on variant pathogenicity and gene-disease relationships. Working with ClinVar, ClinGen launched the ClinGen Curation of ClinVar (CvC) project to review conflicting variant submissions in ClinVar as well as other records deemed to be inaccurate or outdated.
Methods:
Records are prioritized for review based on medically significant conflicts, community nominations, submitter self-conflicts, and P/LP claims with insufficient evidence for a gene-disease relationship. Submissions are curated using an in-house developed Google Chrome extension and may be marked a ‘flagging candidate’ if believed to be incorrect, or marked as ‘no change’ if expert review is needed. One of 11 structured reasons for flagging, with optional comments, are included in the curation to tell submitters why their record was flagged. Flagging candidates are sent in monthly batches to ClinVar and then ClinVar sends these to submitters. Submitters have 60 days to update their submission to prevent the record from being flagged in ClinVar or they can update their submission at any time afterward, which will remove the flag. Flagged records remain viewable in ClinVar, enabling access to historic evidence, but do not contribute to the aggregate classification.
Results:
Following a successful pilot, the CvC project has been active for one year with >11,000 records reviewed on >9000 unique variants. 4020 records from 320 submitters were flagged on 3285 variants leading to resolution of 82% of conflicting variants. The most common flagging reasons were one or more attributes of outlier, outdated, and/or lacking evidence (84%) with most records having at least two of these attributes. Another 13% were flagged due to P/LP claims on variants in ClinGen-curated gene-disease relationships with insufficient evidence (Limited/Disputed/Refuted), 1.7% were flagged due to submission errors (e.g. an older record not being replaced by a new submission), and 1.3% were flagged due to unnecessary conflicting claims on a less relevant disease (e.g. GJB2 variant submitted as benign for dominant hearing loss when it is pathogenic for recessive hearing loss). We received questions from 18 submitters related to 22 (0.5%) flagged records leading to removal of 6 flags. To date, 29% of flagged records have been updated by submitters following notification.
The CvC project is also being deployed in collaboration with ClinGen’s Variant Curation Expert Panels (VCEPs) to address easily resolvable conflicts for VCEP-related genes to avoid labor-intensive resources being allocated to these variants. In a pilot study involving 3 VCEPs, we found that for roughly half of conflicting variants, the source of conflict can be addressed through flagging in a few minutes per variant compared to many hours of expert review time per variant.
Conclusion:
The CvC project improves the accuracy of classified variants in ClinVar by removing outdated or inaccurate records and enables ClinGen VCEPs to focus on tackling challenging variants. We plan to continue our work across all ClinVar conflicts, prioritizing variants in genes from over 75 ClinGen VCEPs. We welcome community input via nominations of variants to review through our portal (https://clinicalgenome.org/curation-activities/clingen-curation-of-clinvar/),or by training of interested community members to review variants on a volunteer basis. In summary, we are excited by the progress achieved in just one year in improving and sustaining the valuable resource that ClinVar has become for the broader genomics community.
ClinVar, a public NCBI-maintained database, has aggregated over 4.7 million classified variants, aiding clinical genomic interpretation and research. However, ClinVar has >28,000 variants with medically significant conflicts (pathogenic/likely pathogenic (P/LP) vs uncertain significance/likely benign/benign), many due to outdated or inaccurate submissions. The Clinical Genome Resource (ClinGen) develops authoritative resources on variant pathogenicity and gene-disease relationships. Working with ClinVar, ClinGen launched the ClinGen Curation of ClinVar (CvC) project to review conflicting variant submissions in ClinVar as well as other records deemed to be inaccurate or outdated.
Methods:
Records are prioritized for review based on medically significant conflicts, community nominations, submitter self-conflicts, and P/LP claims with insufficient evidence for a gene-disease relationship. Submissions are curated using an in-house developed Google Chrome extension and may be marked a ‘flagging candidate’ if believed to be incorrect, or marked as ‘no change’ if expert review is needed. One of 11 structured reasons for flagging, with optional comments, are included in the curation to tell submitters why their record was flagged. Flagging candidates are sent in monthly batches to ClinVar and then ClinVar sends these to submitters. Submitters have 60 days to update their submission to prevent the record from being flagged in ClinVar or they can update their submission at any time afterward, which will remove the flag. Flagged records remain viewable in ClinVar, enabling access to historic evidence, but do not contribute to the aggregate classification.
Results:
Following a successful pilot, the CvC project has been active for one year with >11,000 records reviewed on >9000 unique variants. 4020 records from 320 submitters were flagged on 3285 variants leading to resolution of 82% of conflicting variants. The most common flagging reasons were one or more attributes of outlier, outdated, and/or lacking evidence (84%) with most records having at least two of these attributes. Another 13% were flagged due to P/LP claims on variants in ClinGen-curated gene-disease relationships with insufficient evidence (Limited/Disputed/Refuted), 1.7% were flagged due to submission errors (e.g. an older record not being replaced by a new submission), and 1.3% were flagged due to unnecessary conflicting claims on a less relevant disease (e.g. GJB2 variant submitted as benign for dominant hearing loss when it is pathogenic for recessive hearing loss). We received questions from 18 submitters related to 22 (0.5%) flagged records leading to removal of 6 flags. To date, 29% of flagged records have been updated by submitters following notification.
The CvC project is also being deployed in collaboration with ClinGen’s Variant Curation Expert Panels (VCEPs) to address easily resolvable conflicts for VCEP-related genes to avoid labor-intensive resources being allocated to these variants. In a pilot study involving 3 VCEPs, we found that for roughly half of conflicting variants, the source of conflict can be addressed through flagging in a few minutes per variant compared to many hours of expert review time per variant.
Conclusion:
The CvC project improves the accuracy of classified variants in ClinVar by removing outdated or inaccurate records and enables ClinGen VCEPs to focus on tackling challenging variants. We plan to continue our work across all ClinVar conflicts, prioritizing variants in genes from over 75 ClinGen VCEPs. We welcome community input via nominations of variants to review through our portal (https://clinicalgenome.org/curation-activities/clingen-curation-of-clinvar/),or by training of interested community members to review variants on a volunteer basis. In summary, we are excited by the progress achieved in just one year in improving and sustaining the valuable resource that ClinVar has become for the broader genomics community.