https://orcid.org/0000-0003-1914-2749
Abstract
ClinVar (www.ncbi.nlm.nih.gov/clinvar/) is a free, public database of human genetic variants and their relationships to disease, with >3 million variants submitted by >2800 organizations across the world. The database was recently updated to have three types of classifications: germline, oncogenicity and clinical impact for somatic variants. As for germline variants, classifications for somatic variants can be submitted in batches in a file submission or through the submission API; variants can also be submitted and updated one at a time in online submission forms. The ClinVar XML files were redesigned to allow multiple classification types. Both old and new formats of the XML are supported through the end of 2024. Data for somatic classifications were also added to the ClinVar VCF files and to several tab-delimited files. The ClinVar VCV pages were updated to display the three types of classifications, both as it was submitted and as it was aggregated by ClinVar. Clinical testing laboratories and others in the cancer community are invited to share their classifications of somatic variant classifications through ClinVar to provide transparency in genomic testing and improve patient care.
Introduction
ClinVar (www.ncbi.nlm.nih.gov/clinvar/) is a free, public database of human genetic variants and their relationships to disease, maintained by the National Center for Biotechnology Information within the National Library of Medicine at the National Institutes of Health. Since its production release in 2013, the database has grown to hold >3 million variants submitted by >2800 organizations across the world. ClinVar is used routinely by clinical genetics testing laboratories, research laboratories, expert panels and others who classify variants for their clinical significance. Importantly, the database provides a way for clinical genetics testing laboratories to share data that would otherwise never be publicly available. Laboratories use ClinVar to learn about variants they have never seen before, as well as to resolve differences in classifications (1–4).
While the vast majority of data in ClinVar are classifications of germline variants for inherited disease, the flexibility of ClinVar’s data model technically allowed submitters to provide somatic variants as well. However, the data model did not represent somatic variant classifications in a robust way; nor did it did include classification terms that are specific for somatic variants evaluated for their relevance for cancer. The result was that very few somatic variants were submitted to ClinVar, and those that were submitted were difficult to evaluate. It also meant that ClinVar was not well-positioned to help address the growing number of variants identified in cancer samples. According to the National Cancer Institute, ∼2 million new cases of cancer are anticipated to be diagnosed in the United States by the end of 2024; worldwide, new cases are expected to reach almost 30 million in 2040 (www.cancer.gov/about-cancer/understanding/statistics). The burden of cancer has prompted a long, rich history of research on genomic alterations in cancer. In recent years, genomic profiling of tumor DNA has become common in the clinical setting, particularly for late-stage cancers (5–8). Sequencing of DNA from patients’ tumors provides the potential to identify many new genomic variants that may be clinically relevant for different types of cancers. However, that potential depends on the ability of clinical testing laboratories to publicly share their variants and the evidence for their classifications (9).
Significant efforts to advance data sharing for somatic variants have been made, including establishment of data standards. In 2016, The Somatic Working Group of the Clinical Genome Resource (ClinGen; www.clinicalgenome.org/) published a consensus set of minimal variant level data for the curation of cancer variants for their clinical utility (10). In 2017, a standard for interpreting the clinical impact of somatic variants for cancer was published by the Association for Molecular Pathology (AMP), the American Society of Clinical Oncology (ASCO) and the College of American Pathologists (CAP) (11). A second standard to classify somatic variants for their oncogenicity, specifically their potential to confer growth and survival advantages to tumor cells, was published by ClinGen, the Cancer Genomics Consortium (CGC) and the Variant Interpretation for Cancer Consortium (VICC) in 2022 (12). Databases focusing on somatic variants and cancer have also been developed. The Clinical Interpretation of Variants in Cancer (CIViC) database is a free, expert-curated database that crowd-sources curation of the cancer literature (13). The Catalog of Somatic Mutations in Cancer provides expert curation of the cancer literature, supplemented with data from other sources; however, the data are not freely available as it must be licensed with a fee for commercial use (14). OncoKB™ is an expert-curated database based on the literature, other public datasets and treatment guidelines (15); non-academic use also requires a license with a fee. However, these databases have two limitations: (i) they do not provide a mechanism for clinical genomic testing laboratories to share unpublished variant classifications and (ii) they do not show germline classifications and evidence alongside classifications of somatic variants for cancer. Building upon the existing standards for somatic variants, ClinVar is uniquely positioned to fill these data sharing needs in the cancer community.
Materials and methods
User interviews
User interviews were conducted to assess interest in somatic variants in ClinVar, to define critical and optional data types to describe classifications of somatic variants, and to determine how data should be aggregated. Interview participants included individuals who support the CIViC database (13) to ensure alignment of data models. Participants also included individuals who contributed to the AMP/ASCO/CAP recommendations for assessing the clinical impact of somatic variants (11) and others who contributed to the ClinGen/CVC/VICC guidelines (12) for classifying variants for oncogenicity, to ensure alignment with community standards.
Prototypes
Initial prototypes for submission templates, public XML files and web displays for VCV pages were created manually and tested with additional user interviews. Feedback from the interviews was used to refine the prototypes. The refined prototypes were made available on ClinVar’s GitHub repository (github.com/ncbi/clinvar). ClinVar users were invited to review the prototypes and provide additional feedback.
Production development
ClinVar’s production codebase was updated to accommodate changes for somatic variants in submission processing, data aggregation, data files for download including XML files and public web pages. The initial release in January 2024 supported submission of somatic variants in files, either the Excel spreadsheet template or tab/comma-separated files based on the Excel template. Support for submission of somatic variants using the ClinVar Submission Application Programming Interface (API) was added in May 2024. Support for somatic variants in online forms for submission and update of single variants (the Single Submission Wizard and Single SCV Update) was added in August 2024. Throughout the development process, planned updates were announced on the ClinVar website, on GitHub, in release notes and on social media to ensure that users were aware of upcoming changes.
Results
Updates to submission
Classifications of somatic variants can now be submitted to ClinVar through all submission pathways. Batches of variants can be submitted in a file (Excel or tab/comma-separated files) or through the ClinVar Submission API (www.ncbi.nlm.nih.gov/clinvar/docs/api_http/). A new version of the ClinVar submission spreadsheet template (SubmissionTemplateSomatic.xlsx) was created for submission of somatic variants; it may be downloaded from the ClinVar FTP site (ftp.ncbi.nlm.nih.gov/pub/clinvar/submission_templates/). Variants can also be submitted and updated one at a time in online forms in the ClinVar Submission Portal (submit.ncbi.nlm.nih.gov/clinvar/). Three types of classifications can now be submitted to ClinVar.
Germline classifications
Most germline classifications in ClinVar are for variants that are classified for monogenic inherited diseases using the terms recommended by the American College of Medical Genetics (ACMG) for pathogenicity (16). At this time, this category in ClinVar also includes germline variants that are classified in other contexts, such as pharmacogenomic variants classified for a drug response. All classifications that were submitted to ClinVar prior to 29 January 2024 are represented as germline classifications. On some of these submitted records, the allele origin was submitted as ‘somatic’, so these records may actually represent somatic classifications. The submitters of these records have been contacted to request an update.
Somatic clinical impact
Somatic variants may be classified for their clinical impact on therapy, diagnosis or prognosis for a particular type of cancer. The terms for clinical impact are based on the recommendations from AMP/ASCO/CAP (11). Each submission of a somatic classification for clinical impact indicates whether the variant has an impact on therapy, diagnosis or prognosis. If the variant impacts therapy, the submission also indicates the relevant drug or class of drugs.
Oncogenicity
Somatic variants may be classified for oncogenicity, or their potential to give cancer cells advantages in growth and cell survival. The terms for oncogenicity are based on the recommendations from ClinGen/CGC/VICC (12).
Each submitted record, or SCV (Submission to ClinVar), has one and only one type of classification. However, a submitter may provide more than one classification for the same variant by submitting multiple records. For example, a laboratory may submit a germline classification for a variant in PTEN for the hereditary disease Cowden syndrome; the laboratory may submit the same variant with a classification of oncogenicity for colorectal cancer. Another example is a laboratory that submits multiple somatic classifications for the same variant and type of cancer; there may be one or more classifications for the variant and its effect on different therapies, as well as classifications for its impact on diagnosis and/or prognosis.
As for germline variants, submitters of somatic variants are asked to provide the evidence for their classifications. Two optional fields were added to describe observations of somatic variants. ‘Presence of somatic variant in normal tissue’ lets the submitter indicate whether the somatic variant was present in normal tissue or not, or if normal tissue was not tested. ‘Somatic variant allele fraction’ lets the submitter indicate the percentage of cells in the tumor sample with the variant.
Updates to files for download
XML files. The ClinVar XML files were updated to allow distinct classifications in the germline and somatic contexts. In the old XML format there was a single element, ClinicalSignificance, for the classification of the variant. In the new XML format, ClinicalSignificance is replaced by three elements: GermlineClassification, SomaticClinicalImpact and OncogenicityClassification. Aggregate records (with prefix VCV for Variation in ClinVar and RCV for Reference ClinVar record) may have one, two or all three types of classifications. Multiple types of classifications are expected for variants in genes that are relevant for both inherited disease and cancer, and when both germline and somatic classifications have been submitted to ClinVar. In the XML files, each aggregate record has an SCV section for each submitted record; each SCV section has one and only one type of classification as required by submission. Documentation on ClinVar’s GitHub repository outlines the changes between the old and new XML formats so that users can more easily transition to the new format.
It was anticipated that ClinVar users who download the XML files need time to update their code to parse the updated format. For that reason, the XML files are being generated in both the old and new formats through the end of December 2024. The XML directory on the FTP site (ftp.ncbi.nlm.nih.gov/pub/clinvar/) was re-organized to support both old and versions of the XML, and to make the new format of the VCV XML the default format.
The main directory (/clinvar/xml/) holds the VCV XML files in the new format. Each monthly release from the current year is in this directory. In 2025, the files from releases in 2024 will be moved to the archive directory (/clinvar/xml/archive/) and the monthly releases in 2025 will be found in the main directory. Weekly releases for the VCV XML in the new format are available in the weekly release directory (clinvar/xml/weekly_release); weekly releases are only kept until the next monthly release.
XML files in the other formats are available in sub-directories:
VCV XML files in the old format are in /clinvar/xml/VCV_xml_old_format.
RCV XML files in the new format are in /clinvar/xml/RCV_release.
RCV XML files in the old format are in /clinvar/xml/RCV_xml_old_format.
Each of these sub-directories has another level of sub-directories for weekly releases and archive files, as described for the VCV XML.
Updates to other FTP files
New INFO tags were added to the VCF files for somatic classifications (Table 1). The existing INFO tags CLNSIG, CLNDN, CLNDNINCL, CLNDISDB, CLNDISDBINCL, CLNREVSTAT, CLNSIGCONF and CLNSIGINCL were retained to report data for germline classifications. More details about INFO tags and the VCF file are available in a README file: ftp.ncbi.nlm.nih.gov/pub/clinvar/README_VCF.txt Data for somatic classifications were also added to three of ClinVar’s custom tab-delimited files: organization_summary.txt, submission_summary.txt and variant.summary.txt. Details about these files are available in a README file: ftp.ncbi.nlm.nih.gov/pub/clinvar/README.txt.
New INFO tags for somatic classifications in ClinVar VCF files
| INFO tag | Description |
|---|---|
| SCI | Aggregate somatic clinical impact for this single variant |
| SCIDN | ClinVar’s preferred disease name for the concept specified by disease identifiers in SCIDISDB |
| SCIDISDB | Tag-value pairs of disease database name and identifier submitted for somatic clinical impact classifications, e.g. MedGen: NNNNNN |
| SCIREVSTAT | ClinVar review status of somatic clinical impact for the Variation ID |
| ONC | Aggregate oncogenicity classification for the variant |
| ONCDN | ClinVar’s preferred disease name for the concept specified by disease identifiers in ONCDISDB |
| ONCDISDB | Tag-value pairs of disease database name and identifier submitted for oncogenicity classifications, e.g. MedGen: NNNNNN |
| ONCREVSTAT | ClinVar review status of oncogenicity classification for the Variation ID |
| ONCCONF | Conflicting oncogenicity classifications for the variant |
| INFO tag | Description |
|---|---|
| SCI | Aggregate somatic clinical impact for this single variant |
| SCIDN | ClinVar’s preferred disease name for the concept specified by disease identifiers in SCIDISDB |
| SCIDISDB | Tag-value pairs of disease database name and identifier submitted for somatic clinical impact classifications, e.g. MedGen: NNNNNN |
| SCIREVSTAT | ClinVar review status of somatic clinical impact for the Variation ID |
| ONC | Aggregate oncogenicity classification for the variant |
| ONCDN | ClinVar’s preferred disease name for the concept specified by disease identifiers in ONCDISDB |
| ONCDISDB | Tag-value pairs of disease database name and identifier submitted for oncogenicity classifications, e.g. MedGen: NNNNNN |
| ONCREVSTAT | ClinVar review status of oncogenicity classification for the Variation ID |
| ONCCONF | Conflicting oncogenicity classifications for the variant |
New INFO tags for somatic classifications in ClinVar VCF files
| INFO tag | Description |
|---|---|
| SCI | Aggregate somatic clinical impact for this single variant |
| SCIDN | ClinVar’s preferred disease name for the concept specified by disease identifiers in SCIDISDB |
| SCIDISDB | Tag-value pairs of disease database name and identifier submitted for somatic clinical impact classifications, e.g. MedGen: NNNNNN |
| SCIREVSTAT | ClinVar review status of somatic clinical impact for the Variation ID |
| ONC | Aggregate oncogenicity classification for the variant |
| ONCDN | ClinVar’s preferred disease name for the concept specified by disease identifiers in ONCDISDB |
| ONCDISDB | Tag-value pairs of disease database name and identifier submitted for oncogenicity classifications, e.g. MedGen: NNNNNN |
| ONCREVSTAT | ClinVar review status of oncogenicity classification for the Variation ID |
| ONCCONF | Conflicting oncogenicity classifications for the variant |
| INFO tag | Description |
|---|---|
| SCI | Aggregate somatic clinical impact for this single variant |
| SCIDN | ClinVar’s preferred disease name for the concept specified by disease identifiers in SCIDISDB |
| SCIDISDB | Tag-value pairs of disease database name and identifier submitted for somatic clinical impact classifications, e.g. MedGen: NNNNNN |
| SCIREVSTAT | ClinVar review status of somatic clinical impact for the Variation ID |
| ONC | Aggregate oncogenicity classification for the variant |
| ONCDN | ClinVar’s preferred disease name for the concept specified by disease identifiers in ONCDISDB |
| ONCDISDB | Tag-value pairs of disease database name and identifier submitted for oncogenicity classifications, e.g. MedGen: NNNNNN |
| ONCREVSTAT | ClinVar review status of oncogenicity classification for the Variation ID |
| ONCCONF | Conflicting oncogenicity classifications for the variant |
Updates to aggregate data
ClinVar aggregates submitted data by the variant (VCV records) and by the variant-disease pair (RCV records). For each aggregate record, ClinVar now calculates an overall classification for each type of classification that has been submitted. VCV records may have only one type of classification; both types of somatic classifications; or both germline and somatic classifications for variants in genes that are associated with both genetic disease and cancer (Figure 1A). RCV records are expected to have only a germline classification or one or more somatic classifications, based on whether the disease for the RCV record is an inherited disease or a non-inherited form of cancer (Figure 1B). For example, Figure 2A shows a VCV record for a variant that has an aggregate germline classification of Pathogenic/Likely pathogenic. Its aggregate somatic classification of clinical impact is Tier 1 (Strong) and its aggregate oncogenicity classification is Oncogenic. Two of the RCV records for this variant are shown in Figure 2B and C. In Figure 2B, the RCV record represents germline classification where the specific disease was not provided by submitters; in Figure 2C, the RCV record represents both classifications of oncogenicity and clinical impact of the variant for colorectal cancer.
(A) An infographic depicting the data model for a variant, or VCV, record in ClinVar. A VCV record may have one or more types of classification for the variant. (B) An infographic depicting the data model for a variant-disease, or RCV, record in ClinVar. An RCV record is expected to have either a germline classification or one or more somatic classifications, depending on the disease for the record.
(A) The classification summary on VCV000016609.133 (www.ncbi.nlm.nih.gov/clinvar/variation/13961/) shows that the variant BRAF V600E has a germline classification as well as somatic classifications of clinical impact and oncogenicity. Each type of aggregate classification has its own review status, indicated by stars, and a total number of submissions with that type of classification. (B) The classification summary on RCV000080903.19 (www.ncbi.nlm.nih.gov/clinvar/RCV000080903.19/) shows that BRAF V600E has only a germline classification provided for records where the disease was not included in the submission (‘not provided’). (C) The classification summary on RCV001030023.13 (www.ncbi.nlm.nih.gov/clinvar/RCV001030023.13/) shows that BRAF V600E has only somatic classifications provided for colorectal cancer.
The aggregate classification for germline classifications is calculated using the same rules as for the old field of clinical significance (www.ncbi.nlm.nih.gov/clinvar/docs/clinsig/#clinsig_agg). The aggregate classification for oncogenicity is classified in a similar way. The classifications using the ClinGen/CGC/VICC recommendation are compared to calculate a consensus classification or a conflict. For oncogenicity, no terms other than those recommended by ClinGen/CGC/VICC are accepted in submissions, so no other terms are added to the end of the aggregate classification.
The aggregate somatic classification for clinical impact is calculated slightly differently. For this type of classification, it is expected that there will be multiple classifications, or tiers of evidence, reported for different types of assertions and different types of cancer. Therefore, conflicts in the tiers are not considered meaningful differences and the aggregate somatic classification for clinical impact is calculated as follows:
Submitted records with the highest review status are identified.
Only this subset of submitted records are used in subsequent steps.
The classifications using the AMP/ASCO/CAP recommendations are compared.
The highest classification, or tier, becomes the aggregate classification.
A review status is calculated for each aggregate classification, so that the user understands the level of review for each type of classification (17). For example, in Figure 2A, the germline classification has two stars indicating consensus from multiple submitters, whereas the oncogenicity classification has a single star because there is only a submission from one submitter so there is no consensus yet. The review status for germline and oncogenicity classifications are calculated using the same rules for the old clinical significance (www.ncbi.nlm.nih.gov/clinvar/docs/review_status/#revstat_agg). The review status for somatic classifications of clinical impact is calculated differently; in this case, consensus in the classification is not considered. For clinical impact, the review status is based on whether there is a classification from a practice guideline or an expert panel, and if not, whether any submitted record provided assertion criteria and evidence for the classification (or a public contact).
Dates are also calculated for each type of aggregate classification. These are the date that the variant first appeared in ClinVar with that type of classification; the date of the last (most recent) submission with that type of classification; and the most recent date that any submitter evaluated the variant for that classification type (Figure 3).
The variant details section on VCV000016609 includes three dates calculated for each type of classification: the date the variant was first in ClinVar, the date of the last submission and the date it was last evaluated by any submitter.
Updates to VCV web pages
As already shown in Figure 2A, the classification summary at the top of a VCV page is now separated into two sections, one for germline classification and one for the two types of somatic classifications. Each section shows the aggregate classification, total number of submissions and review status for the classification type. Each section also has a toggle button on the right that hides all the germline or somatic data on the page, when the user who wants to narrow their focus. The toggle can also be used to show the data again, when the user wants to broaden their focus to explore both germline and somatic data.
The VCV pages previously displayed information for the variant, conditions and the gene in a set of tabs. These data are now displayed directly on the page in a set of tables, so that the user does not have to click a tab to find the information. Additionally, the tables for conditions, submissions and citations are now split into separate tables for germline and somatic data (Figure 4). A page menu was added so that the user can quickly navigate to a specific section of interest (Figure 3).
The somatic data tables for VCV 000376363. The tables report the data aggregated by variant-condition pair (RCV-level data), data for each submission (SCV-level data), and a summary of citations submitted for somatic classifications of the variant. Analogous tables for germline data are also provided.
Discussion
As ClinVar has evolved, it has become clear that a single classification is not sufficient to capture the different contexts in which a variant may be classified. Most variants in ClinVar are germline variants that have been classified for monogenic disease, using either the classification guidelines from ACMG (16) or a submitters’ own classification rules that are often based on the ACMG guidelines. The publication of guidelines for the clinical impact (11) and oncogenicity (12) of somatic variants established standard terms for these types of classifications, and provided a framework for ClinVar to add support for somatic variants. It is expected that many submitters of somatic variants will indicate that they have classified variants based on either of the published guidelines; alternatively, submitters may use their own classification rules for somatic variants, presuming that they use the same terms as the published guidelines (e.g. Oncogenic or Tier I—Strong) or that their own terms for classification can be mapped to the standard terms. It is anticipated that at least one additional type of classification will be added to ClinVar in the future to support variants classified in the pharmacogenomic context.
This release supporting somatic variants in ClinVar has a few limitations. Cancer types can be submitted using names or with identifiers from Medical Subject Headings (www.nlm.nih.gov/mesh/meshhome.html) or MedGen (www.ncbi.nlm.nih.gov/medgen), which includes terms from the NCI Thesaurus (ncithesaurus.nci.nih.gov/ncitbrowser/); however, support for identifiers in databases that specifically focus on cancer have not been added. Similarly, drugs on therapeutic assertions are submitted only as the drug name and are not standardized using database identifiers. Standardization of cancer types and standardization of drug names may be added in the future; the ClinVar team welcomes suggestions for appropriate databases. Additionally, this work focused on classifications of somatic variants only for cancer. ClinVar also includes classifications of somatic variants for other types of conditions, including diseases of somatic mosaicism. No specific updates were made for these variants; currently they are submitted and displayed like germline variants. Improving representation of somatic variants for conditions other than cancer is another potential enhancement for ClinVar.
The addition of classifications of somatic variants for cancer in ClinVar has great potential to improve variant classification. Those who classify variants in the germline context will benefit from seeing the evidence for the variant in the somatic context alongside evidence in the germline context (18–22). Likewise, those who classify somatic variants from cancer will easily find evidence for the variants in the germline to inform their classification process (11,23–25). Laboratories that identify somatic variants will be able to pool their evidence and compare their approaches to classification, just as laboratories have done for germline variants. Sharing somatic variant classifications in ClinVar can help drive the improvement of standards for somatic variants, as it has done for germline variants. Introducing separate types of classifications for germline and somatic variants can be viewed as an initial step toward distinguishing other types of classifications in the future. For example, it is anticipated that a new classification type for pharmacogenomic variants will be added when standards for their classification are established.
As for many improvements to ClinVar, the utility of this feature depends heavily on its uptake by submitters. As of September 2024, ClinVar holds ∼700 variants with somatic classifications. This total includes an initial submission of a few variants submitted by CIViC, and further submissions from this curated resource are anticipated. As part of an approval process, the ClinGen Somatic Cancer Variant Curation Expert Panels are required to provide five examples of variants and associated data that will be submitted to ClinVar, via the curation tool hosted by CIViC (clinicalgenome.org/site/assets/files/3263/sc-vcep_4step_process_dec2020.pdf); thus somatic variants that are reviewed by ClinGen expert panels are anticipated as well. However, effective variant curation by groups like CIViC and ClinGen relies not only on variants published in the literature but on variants identified in clinical testing that do not get published. To fill this need, clinical genomic testing laboratories that classify somatic variants for cancer are encouraged to consider how they can share variant-level classifications through ClinVar. Research laboratories and curation groups are also encouraged to share their classifications of somatic variants for oncogenicity and/or clinical impact. Information to get started with submission is available online (www.ncbi.nlm.nih.gov/clinvar/docs/submit/). Specific instructions for submitting somatic variants are also available (www.ncbi.nlm.nih.gov/clinvar/docs/spreadsheet/#somatic). Contact the ClinVar team at [email protected] with any questions about submitting or using somatic variants in ClinVar.
Data availability
ClinVar is freely available at www.ncbi.nlm.nih.gov/clinvar/.
Acknowledgements
The authors gratefully acknowledge Obi Griffith, Alex Wagner, Marilyn Li, Catherine Cottrell, Arpad Danos, Heidi Rehm and Larry Babb for early user feedback that shaped this project. The authors also gratefully acknowledge all the organizations that share data through ClinVar (www.ncbi.nlm.nih.gov/clinvar/docs/submitter_list/).
Funding
National Library of Medicine. Funding for open access charge: National Library of Medicine.
Conflict of interest statement. None declared.
Comments