GA4GH Clinical Genomics Laboratory Community Meetings

Meeting Schedule and Topics

Background

The Global Alliance for Genomics and Health (GA4GH) and the Clinical Genome Resource (ClinGen) have launched the Clinical Genomics Laboratory Community (CGLC) to unite clinical genomics laboratories worldwide in advancing genomic medicine. This initiative fosters international collaboration, consensus-building, and knowledge exchange to enhance genetic testing and analysis for improved patient care. By addressing critical challenges such as variant classification, clinical reporting, data sharing, and technical hurdles, the CGLC aims to develop standards, best practices, and innovative methodologies. Experts will convene to create a dynamic forum, driving progress and enabling laboratories to contribute effectively to cutting-edge research.

Key Resources

Agenda and Minutes

Topic: The Matched Annotation from NCBI and EMBL-EBI transcript set

2026-11-02 at 8 EDT/ 12 UTC

Recording

Passcode

Slides

Session Description: This presentation will cover a description of MANE dataset, MANE release cycle, rare updates to MANE data, scenarios resulting in assignment of new MANE Plus Clinical transcripts and plans for a new 'MANE Plus' set.

Presenters:

Shashikant Pujar - NCBI/NLM/NIH

Jane Loveland - EMBL-EBI

Terence Murphy - NCBI/NLM/NIH

Adam Frankish - EMBL-EBI

Summary

The Matched Annotation from the NCBI and EMBL-EBI (MANE) is a genome-wide set of representative transcripts and corresponding proteins. Each MANE transcript is an exact match in exon coordinates, coding region, and sequence between a Refseq transcript and its Ensembl/GENCODE counterpart. Chosen based on biological criteria such as transcript expression levels and conservation of coding regions, MANE transcripts match the GRCh38 reference perfectly. Recently, clinically relevant non-coding genes have been added. The following are provided in MANE dataset:

• MANE Select - One transcript at each locus across the genome that is representative of biology at that locus; Useful as a standard for clinical reporting, as a default display on browsers/key genomic resources, and as a starting point for comparative or evolutionary genomics.
• MANE Plus Clinical - The set includes transcripts for genes where MANE Select is not sufficient on its own to report all "pathogenic" or "likely pathogenic" clinical variants in public resources.

Topic: SVCv4 Data Model for Software Platform Network

2026-09-21 at 21:00 UTC

Recording

Passcode: 3K@W5yky

Slides: SVCv4 Data Model - September CGLC

Session description: This session will provide an update on the development of the GA4GH SVCv4 Data Model by ClinGen and GA4GH, which is being co-developed alongside the ACMG/AMP/CAP/ClinGen SVCv4 standards to provide an accurate, common technical foundation for their consistent and computable implementation. As software platforms begin planning how they will support SVCv4, we particularly encourage developers, engineers, and platform teams to attend and become familiar with the Data Model early. Building from shared structures can reduce duplicated implementation effort and the risk of unnecessary divergence, while supporting compatibility with the broader SVCv4 ecosystem and the exchange and reuse of data across tools, resources, and organisations.

Summary

The meeting focused on the Sequence Variant Classification version 4 (SVCv4) data model that was developed by ClinGen in collaboration with the GA4GH Genomic Knowledge Standards (GKS) workstream. Heidi highlighted that although the data model is not mandatory, it ensures consistency in variant classification and helps software platforms implement standards. Additionally, once SVCv4 is formally approved, the team will be launching a discussion forum similar to that of gnomAD where users can ask questions and query topics such as the data model.

The call featured three speakers, Alicia Byrne, Larry Babb, and Christon Preston. Alicia provided an overview of the SVCv4 standards and data model, including the shift towards more prescriptive workflows and structured evidence representation while Larry explained the technical aspects of the model (e.g., use of codes, shared namespaces). Larry also provided a demonstration of the provisional documentation site. Finally, Christine covered the ClinGen Criteria Specification (CSpec) Registry and Variant Curation Interface (VCI) tools as example implementations.

SVCv4 Standards and Data Model (Alicia Byrne)

• The data model was co-developed with the v4 standards working group to ensure accuracy and fidelity with the intended use of the framework; The framework enters formal review in the coming weeks through ACMG, AMP, CAP, and ClinGen, with publication targeted for spring 2027
• The data model consists of two complimentary models

1. A classification model that records the variant classification conclusions and evidence
2. A method model that documents the structured methods used for classification

• The models illustrate where and why classifications differ between groups; however, Alicia stressed that differences in classification aren't inherently bad. The problem is differences whose origin can't be understood.
• Potential benefits of the shared infrastructure include interoperable evidence resources, integration with clinical and laboratory workflows, improved communication of classification outcomes and data-driven learning and standards refinement.

Overview of the draft SVCv4 Data Model (Larry Babb)

• The model follows the GA4GH Variant Annotation (VA) Specification.
• Larry walked through the model’s basic structure:

• A statement (classification) is tied to a separately defined proposition, which makes it easier to compare assertions from different groups. It is supported by nested evidence lines, each linked to the method (rule) applied and to evidence items with provenance.

• Larry called for "applied evidence," meaning the minimal data points attached to the specific rule they satisfy, rather than free-text summaries, so others can verify outcomes without re-curating.
• Lary also explained that because computational use needs a code at every decision point, the data modeling group will add roll-up codes and "subcodes" where the official v4 table has gaps, such as for top-level categories, combined caps, and the workflow steps beneath each code. He used a missense prediction code to illustrate this.
• He encouraged groups to start thinking about how their specializations will adapt to future versions such as v4.1.
• Early adopters were encouraged to reach out to give feedback and help shape the SVCv4 profile standard

Example Implementations (Christine Preston)

• The ClinGen Criteria Specification (C-SPEC) registry will provide computable methods and specifications for variant curation expert panels; It implements the method side of the VA Spec model and will include:

• Gene, disease, inheritance scope
• Precise code definitions and application criteria
• Workflow logic with task dependencies and points
• Permitted refinements to default methods
• Rationale and provenance for each choice

• The ClinGen Variant Curation Interface (VCI) implements the classification side of the VA Spec model; It will integrate C-SPEC through the data model to allow automation and standardization

Draft SVCv4 Classification Model

Topic: SVCv4

2026-07-20 at 20:00 UTC

Recording

Passcode: wC@YHb6f

ACMGv4_CGLC_July2026 

Session Description: The group will review the details of the final SVCv4 pilot. Participants will have the chance to hear from Steven Harrison, PhD FACMG (Ambry Genetics) who will provide an overview of the SVCv4 criteria, as well as Leslie Biesecker (National Human Genome Research Institute (NHGRI)) who will discuss predictive and functional data and the impact on benign classifications and variants of uncertain significance (VUS). 

Summary

The meeting focused on the launch of the final pilot for the Sequence Variant Classification Version 4 (SVCv4) standards. The pilot officially begins on July 21st with draft documents and an updated pilot calculator being released to participants. The speakers reviewed the rationale for the fourth version, major methodological changes, results from previous pilot testing, and demonstrated the new web-based calculator.

SVCv4 Near-Final Draft Available (Heidi Rehm)

• SVCv4 represents the next evolution of the ACMG/AMP sequence variant classification framework, now led by ACMG/AMP/CAP/ClinGen.
• Previous standards are now retrospectively referred to as SVCv1–v3.
• Future updates will be versioned (e.g., SVCv4.1, SVCv4.2) through a planned standing committee rather than waiting many years between revisions.
• Development is also underway to align CNV classification standards with the same Bayesian framework.

Major Changes (Les Biesecker)

1. Bayesian Points-Based Framework - The largest conceptual change is replacing qualitative evidence strengths with a quantitative point system

• Evidence is assigned positive points for pathogenic evidence and negative points for benign evidence.
• Final classifications are determined by the accumulated score.
• The new system improves consistency, simplifies combining evidence, makes reasoning more transparent and, better reflects uncertainty

1. VUS Subclassification - Variants of Uncertain Significance (VUS) are now subdivided into:

1. VUS-low
2. VUS-mid
3. VUS-high

• This allows laboratories to distinguish variants that are close to likely pathogenic from those with minimal supporting evidence.
• The presenters noted that ACMG recommendations regarding reporting VUS, including addressing subclasses, have recently been published.

• Evidence Framework - Evidence is broadly broken down into human observational data and predictive and functional data. Scoring includes assessment of variant type to drive decision tree selection and consideration of disease, disease mechanism, exon relevance and informative variants.

SVCv4 Data Model (Alicia Byrne)

• A ClinGen working group has developed a common, computable foundation for accurate, interoperable, and scalable implementation of SVCv4
• A call will be scheduled in August to share the data model

Phase 2 Pilot (Steven Harrison)

• The phase 2 pilot involved approximately 25 laboratories, 30 curated variants, and 26 independent classifications per variant
• All participating laboratories received identical evidence to remove any variation due to evidence identification

• The goal was to evaluate whether laboratories interpreted the workflows consistently.
• Strong overall concordance was reported and participant feedback was used to refine the evidence framework, decision trees, and calculator functionality.

Pilot Calculator Demonstration (Steven Harrison)

• The calculator is designed to minimize manual arithmetic and facilitate application of the framework and decision trees.

Materials

Links to documents and the calculator are below. Please note that all materials released during pilot phases and during the professional society reviews of the framework are considered draft. We ask that any work testing or commenting on the new framework be limited to abstracts and be held for publication until the final version is published, and that any interim analyses be updated with the final version of SVCv4. This will avoid confusion over studies that involve draft versions versus final versions of SVCv4.

CGLC Slide Presentation ACMGv4_CGLC_July2026

PILOT CALCULATOR

Access requires a login. You must create an account when using the Pilot Calculator for the first time.

• ClinGen SVCv4 Pilot Calculator: https://calculator.clinicalgenome.org/v4/pilot/ui/classification
• Video tutorial introducing using the Pilot Calculator for variant classification: https://drive.google.com/file/d/14HPyORkQG23K0h6aKSxudNCoGcISpUQe/view?usp=drive_link

SVCv4 DOCUMENTATION

These documents will be available as supplemental files with the SVCv4 publication.

• DRAFT Code-specific workflow usage descriptions: https://drive.google.com/drive/folders/19RF57OC1mR-954nC9iuiOX0_cVR7suOf?usp=drive_link 
• DRAFT High-res images of variant-type-specific workflows: https://drive.google.com/drive/folders/1dMjLBmXo4dtIZA2PItWNvnAHiRyfvapG?usp=drive_link

FEEDBACK

• General feedback on the SVCv4 framework: https://forms.gle/s6y6E8U8SFW9DyQi6
• Feedback on the classification of any variant: https://forms.gle/LR2Vi3msdH6jo7dP8
• Feedback on the classification of a practice variant (please select the appropriate form): https://docs.google.com/document/d/1SE908O-MjdnKdjukTxZM3ilEpCzX5qIyNHdN0nqG7fg/edit?usp=sharing

PRACTICE VARIANTS WITH EVIDENCE

These variants and their accompanying evidence are provided as examples to help users familiarise themselves with the SVCv4 framework and the Pilot Calculator tool.

The Pilot Calculator can be used to classify any variant.

• https://docs.google.com/spreadsheets/d/1cxgrH3EZKFLBkUvCRbYdm7evP_vSXApB9_UWgeDT6SI/edit?usp=sharing

SUPPORT CONTACTS

• For questions related to the use of the SVCv4 framework or questions about the Pilot process please contact: acmg_svc_pilot@clinicalgenome.org 
• For ClinGen Pilot Calculator tool support or to report issues with the tool, please contact: brl-cspec@bcm.edu 

Topic: Best practice recommendations for bioinformatics approaches applied to high-throughput sequencing for rare disease and cancer diagnosis within the UK National Health Service

2026-06-22 at 12:00 UTC

Recording LINK Passcode: KA41?H#r

Session Description: Establishing best practice recommendations helps to increase consistency, equity and innovation in clinical genomics services for rare disease and cancer. A broad range of international recommendations exist for genomic diagnostic testing and genetic variant classification, yet the current UK-specific best practice recommendations for bioinformatics approaches applied in this context are outdated. In this session, we will introduce updated best practice recommendations, developed through structured discussion, polls and surveys with experts across geographically distributed genomics laboratories in the UK National Health Service. The recommendations cover the spectrum of activities within a clinical genomics bioinformatics pipeline, including quality control, primary, secondary and tertiary analysis approaches and shared knowledge bases. Overall, we will present the agreed national framework for adoption and innovation of bioinformatics approaches across diverse clinical genomic testing strategies in the UK National Health Service

Publication is available here: https://pubmed.ncbi.nlm.nih.gov/42055801/ 

Context & purpose

• Jamie Ellingford from Genomics England presented updated UK consensus best practice recommendations for clinical bioinformatics in high-throughput sequencing diagnostics, focused on rare disease and cancer testing.
• The recommendations update the UK’s 2016 clinical bioinformatics guidance and aim to support greater consistency, equity, reproducibility, and innovation across NHS genomic services.
• Jamie emphasized that the guidance is not intended to replace existing literature, but to translate a rapidly evolving clinical bioinformatics landscape into clear, auditable expectations for accredited diagnostic services.

Development process

• The recommendations were developed over approximately 18-24 months through a broad UK working group under the Association for Clinical Genomic Science, with input from experts across geographically distributed NHS genomic laboratories.
• The final guidance includes 40 essential recommendations, including many new or expanded areas beyond prior published guidance.
• The recommendations cover the full clinical genomics bioinformatics pipeline, including quality control, primary analysis, secondary analysis, tertiary analysis, annotation resources, shared knowledge bases, software development, accreditation, and data/code sharing.

Clinical bioinformatics and accreditation expectations

• A central theme was that clinical bioinformatics pipelines must be validated before use, monitored during service, and auditable after results are reported.
• The guidance aligns with accreditation expectations such as ISO 15189, including documented scope, accountability, versioning, quality monitoring, performance evidence, change control, and auditability.
• Jamie highlighted the importance of reproducible and deterministic pipelines, where the same inputs should produce the same outputs or where any differences are clearly understood and controlled.

Quality control and analysis recommendations

• Quality control was described as a key clinical safety gate, with pre-defined thresholds needed for sequencing metrics, contamination, sample identity, sex karyotype checks, familial relationships, tumour-normal comparisons, and variant-specific QC.
• Secondary analysis recommendations focus on validated approaches for alignment, variant detection, and mutational signature detection, supported by performance evidence such as sensitivity, specificity, error rates, uncertainty, and limits of detection.
• Tertiary analysis recommendations address how services reduce large variant call sets into clinically manageable shortlists, using appropriate annotation resources and clear, documented filtering approaches.

Annotation, reporting, and limitations

• The guidance recommends using current and clinically appropriate resources such as MANE Select or MANE Plus Clinical transcripts, gnomAD, ClinVar, in-house frequency databases, and relevant in silico tools.
• Jamie cautioned that tools such as REVEL and SpliceAI can support annotation and interpretation, but should not generally be used to filter variants out of review unless specific guidance supports that use.
• Clinical reports should include audience-appropriate statements of limitations, including the test’s scope, accreditation status, technical boundaries, and variant types that may not be reliably detected.

Discussion

• Discussion focused on practical implementation issues, including contamination thresholds, how saliva samples may differ from blood samples, reproducibility of tools such as GATK, VUS reporting, and the use of variant inclusion lists for pathogenic variants above population frequency thresholds.
• Participants also discussed how often ClinVar should be updated, with monthly updates viewed as ideal but quarterly updates appearing more feasible for many clinical pipelines.
• Jamie noted that some UK services use small, context-specific inclusion lists for variants above frequency thresholds, but there is not a broad public list of all pathogenic or likely pathogenic variants above 1%.

→ Slido Results: LINK

Topic: gnomAD & population database use

2026-05-11 at 20:00 UTC

Recording link

Passcode: dzBc&?2t

Slides

Session Description: gnomAD is one of the most widely used population databases, but there are still many misconceptions regarding the content of the dataset. These misconceptions include how the types of studies included impact the phenotypes that may be in gnomAD and interpretation of gnomAD data, as well as what data can and can’t be shared and how data changes between versions. This will be a mix of presentations and discussion so we can address as many new and frequently asked questions as possible.

Summary

gnomAD (Genome Aggregation Database) is a population database with over 730,000 exome and 76,000 whole genome sequences. It aggregates genomic data from over 100 studies around the globe. With the exception of age, sex, ancestry, and case-control status, the experts explained that the resource does not contain phenotype data due to regulatory and patient consent limitations. The group spoke about types of studies in gnomAD and use of different data subsets (e.g., UK Biobank), emphasizing that gnomAD is not a dataset of healthy controls. Overall, the discussion highlighted that gnomAD is best used for frequency filtering, variant interpretation, and population-scale genomic references. The meeting concluded with a discussion on plans for regional-specific databases and the Federated gnomAD project.

Overview

• gnomAD is a large-scale population genomics reference database with over 730,000 exome sequences and over 76,000 whole genome sequences. The database aggregates genomic data from more than 100 studies and over 308 contributors worldwide.
• Population databases like gnomAD are built from specific cohorts and studies. Each has its own recruitment criteria, consent frameworks, biases, and metadata definitions. The experts stressed that “control” populations may differ across disease studies due to this heterogeneity (e..g, a control participant in a schizophrenia study may not be the same as a control in a cancer study; many have diseases unrelated to the cohort study).
• gnomAD does not recruit or sequence any participants. It relies on secondary use of previously collected genomic datasets. Note that only those who have consented to secondary research and data sharing are eligible.
• The project collects age, sex, ancestry and case-control status. Detailed phenotype and individual-level DNA is not collected or shared publicly due to strict regulatory and ethical requirements. Even high-level HPO terms are not allowed in the dataset
• gnomAD prioritizes open, freely accessible aggregate data without requiring user accounts or special permissions. This is different from resources like the All of Us Research Program or the UK Biobank; Only aggregate allele frequency data from the UK Biobank for example, can be incorporated into gnomAD.
• Metadata definitions are determined by the original contributing studies and not normalized by gnomAD. For example “age” may represent age at consent, diagnosis, or data collection based on the study design.
• Older datasets may no longer meet newer quality thresholds due to the evolution of sequencing technologies, quality standards, and pipelines over the years. For example, the TCGA cohort (the only cancer cohort that was in gnomAD) was removed for v4.
• The experts emphasized that gnomAD is intended to be used as a genomic reference resource, not a traditional disease cohort.

Types of Studies and data use

• Approximately 50% of exome samples in gnomAD V4 come from the UK Biobank; The UK Biobank recruited about 500,000 volunteers aged 40-69 from England, Scotland, and Wales. While the cohort demonstrates health volunteer bias, Chao emphasized that 2-5% of the population had rare diseases depending on the definition used.
• There is rich phenotype data in the UK biobank but the gnomAD team only has access to genotype data. The experts encourage researchers to use resources like the UK Biobank or All of Us Research Program when they require detailed phenotypes.
• Earlier versions of gnomAD relied heavily on case-control studies, creating phenotype enrichment concerns for specific diseases when the database was smaller. Subsets like “non-neuro” and “non-cancer” were created to address this concern.
• Subsets are less important in gnomAD v4 because the sample sizes are much larger.
• gnomAD is not, and has never been, a “healthy control" database. It is a population database without access to detailed phenotype data. Therefore, Heidi Rehm stressed that the ACMG code BS2 should not be applied using gnomAD data unless the disease is highly penetrant and lethal in the young pediatric population (confidently absent in minors and adults).
• Questions about clonal hematopoiesis (CHIP) highlighted that gnomAD includes flags for some CHIP-associated genes and that users can inspect allele balance, age distribution, genotype quality, and allele fraction metrics to evaluate variants
• Elevated frequency of the CFTR ΔF508 variant (Cystic Fibrosis) in gnomAD was linked to recruitment bias at a pulmonary clinic.

Next steps for gnomAD

• Ongoing efforts are being made to improve diversity by focusing on more globally diverse cohorts and federated data sharing
• The Federated gnomAD project enables international groups to process data locally using gnomAD-compatible pipelines; Aggregate allele frequencies are then shared and merged.
• Jordan Lerner-Ellis announced that the gnomAD-Canada V1.0 is now available for download — with a browser interface coming soon; Regional databases provide advantages for population-specific allele frequencies, carrier frequency estimates, genetic prevalence calculations, and potential future phenotype integration.
• Michael Talkowski and team are working with gnomAD to develop a long-read sequencing browser.

Topic: ClinGen Low Penetrance/Risk Allele Working Group

2026-04-06 at 20:00 UTC

Recording

Passcode: #d4y&*iB

Poll: https://app.sli.do/event/i3p145noX6wxtudK8dPVgc/live/polls 

Session Description: This session will focus on the challenges associated with interpreting and reporting low penetrance variants and risk alleles in clinical genomics. Matthew Lebo and Marcie Steeves, from the ClinGen Low Penetrance/Risk Allele Working Group, will provide an overview of the current landscape, highlighting the complexities related to terminology and reporting thresholds for these variants. The broader group will then be engaged in the discussion through live polling to explore current practices across the community and how such findings are communicated in clinical contexts.

Summary:

The discussion covered terminology recommendations, classification, and challenges in distinguishing between risk alleles, low-penetrant variants, and high-penetrant variants. The group shared the results of the survey and while many labs are classifying these variants, specific terminology is not widely used, and there's ongoing debate about appropriate penetrance thresholds. Participants discussed the goals of labeling these variants to help clinicians understand and communicate different risk levels to patients, with Heidi suggesting that penetrance information should be reported at both gene and variant levels. The group also addressed challenges in classifying variants in complex genetic scenarios and received feedback from the community about terminology preferences and practical implementation concerns.

Risk Alleles Classification Framework

Matt Lebo presented work from the Clinical Pensions and Risk Alleles Working Group. The discussion covered challenges in classifying these variants, including their higher frequency in the general population and difficulties with traditional evidence types like segregation studies. Matt presented a risk allele framework with specific terminology and classification tiers, including categories like "established risk allele" and "likely pathogenic low penetrance." The group also discussed criteria for classifying risk alleles based on association studies, with different levels of evidence required depending on the size and replication of the studies. Sharon raised a question about the number of cases and controls needed for strong evidence, to which Matt responded that the criteria were based on studies available at the time of framework development, which was several years ago.

Genetic Variant Classification Framework

Matthew presented on the classification of genetic variants, focusing on risk alleles and low penetrance variants. He discussed the challenges in distinguishing between common variants for GWAS studies versus those with less power, noting that larger datasets now available could improve case control studies. The presentation covered the group's work on establishing a framework for classifying low penetrance variants, including challenges with genetic evidence, experimental data, and population frequency. Matthew shared survey results from the community regarding current practices in classifying and reporting risk alleles and low penetrance variants, finding that while many labs include these variants in reports, specific terminology usage is limited. The discussion highlighted ongoing work to develop more quantitative thresholds for distinguishing between different levels of variant classification and to address complex scenarios involving recessive inheritance patterns.

Genetic Penetrance Threshold Discussion

The group discussed challenges with defining penetrance thresholds for genetic variants, particularly around the "low penetrance" label. Heidi questioned whether the current 20% threshold was appropriate given evolving understanding of lower penetrance in many diseases, while Matthew explained that the goal of these labels is to help clinicians and patients understand actual disease risk, not just to make lab processes easier. The discussion highlighted the need to distinguish variants that require special attention from those with typical penetrance levels, with Heidi suggesting that separate penetrance information at both gene and variant levels would be more useful than the current pathogenicity categorization system.

Low Penetrance Variant Threshold Framework

The working group discussed challenges in establishing thresholds for low penetrance variants across different diseases. Matthew explained they are seeking to create a framework that can be applied generally while acknowledging that every gene-disease combination has unique characteristics. The group explored various approaches to defining and reporting low penetrance variants, including whether to categorize them at the disease, gene, or variant level. They also discussed the distinction between risk alleles and low penetrance pathogenic variants, with Matthew noting that risk alleles are currently reported based on odds ratios rather than penetrance levels.

Topic: Community Discussion on the ACMG Secondary Findings List and Future Evolution

2026-03-02 at 13:00 UTC

Recording: Listen to the call

Passcode: bH@0dn3E

Poll: https://app.sli.do/event/fFdFs1G2C4iXCBLwETK2XX 

Session Description: This session will focus on engaging the CGLC community on current practices and experiences in supporting the ACMG Secondary Findings policies. We will send a brief survey in advance of the March 2nd meeting that individuals can fill out in advance of, or upon joining, the call. We will review the survey results on the call and have discussion to gather community input that will be shared with members of the ACMG Secondary Findings Working Group and be shared in a session at the upcoming ACMG 2026 meeting entitled: Beyond the List – Scope and Implementation of ACMG Secondary Findings.

Summary:

The meeting focused on discussing ACMG recommendations for the return of secondary genetic findings, with Heidi presenting the laboratory perspective ahead of her upcoming ACMG presentation. The discussion covered key aspects including the definition and implementation of secondary findings versus incidental findings, current practices across labs, and challenges in variant interpretation. Participants shared their experiences with the 84-gene list, with most supporting the current practice of offering secondary findings at no additional charge. The group explored issues around variant penetrance, reporting guidelines, and the potential inclusion of pharmacogenomic results, while also discussing the burden of supporting return results and the need for VCEP criteria. The conversation ended with an overview of ClinGen website's reporting guidance resources and a preview of an upcoming presentation by Matt Lebo and Ryan Schmidt on low penetrance risk alleles.

ACMG Genetic Findings Recommendations

Heidi led a discussion on the ACMG recommendations for returning secondary genetic findings, providing background on terminology differences between secondary and incidental findings and reviewing the evolution of the 84-gene list. She presented data from the All of Us research program showing a 2.8% positive return rate after reviewing 40,000 variants, noting significant variability in yield across different genes. The discussion included questions about reporting findings outside the ACMG list and patient consent requirements, with Heidi explaining that policy implementation and consent practices vary between labs.

Community Survey Results

Heidi presented the survey results, focusing on secondary findings practices. 77% of respondents support the current practice of offering secondary findings as part of testing at no additional charge, while 14% favor offering them as an add-on test with separate billing. Regarding the list of 84 genes, 84% of respondents felt it was just right, with only 9% considering it too long. The survey also addressed pharmacogenomic results, with 41% supporting their inclusion in secondary findings even if every patient had a result. Challenges in returning variants were discussed, with 59% of respondents unsure about returning variants based on penetrance.

With regards to variant reporting challenges, 41% were unsure about returning variants based on disease or phenotype associations, while 10% experienced difficulties confirming results due to analytical validity issues. The committee has removed one or two genes from the existing list of 84 and is considering further removals. Heidi announced a new resource launched on the ClinGen website to help guide reporting of results in the ACMG secondary findings list, which can be accessed through the curated genes menu and downloaded as a structured spreadsheet.

ClinGen Secondary Findings Resource Overview

Heidi presented an overview of the ClinGen secondary findings resource, which was developed in collaboration with the ACMG Secondary Findings Committee. The resource includes a list of genes with reporting guidance. She discussed various questions and concerns raised during the presentation, including the need for VCEP criteria, differences in approaches between US and non-US labs, and considerations around experimental therapies and pharmacogenomic findings. The ACMG committee meets monthly to review requests and concerns, and typically publishes updated lists on an annual basis.

Topic: ClinVar: Submission Guidance and Community Input

2026-01-12 at 13:00 UTC

Recording link; Passcode: t22AC8=*

Poll: https://app.sli.do/event/4Mq2kZQNEphu8V55gVdjJT 

Session Description: This session will offer a practical overview of ClinVar submissions and a look behind the scenes at how submissions are reviewed, structured, and displayed. An update on the ClinGen “Curation of ClinVar” project will be provided, along with an overview of how it aims to improve data quality and consistency. Following the presentations, we will host an open discussion to gather community input on the most important features and functions to best support data sharing in variant knowledge. Discussion topics will include: distinguishing variant-disease associations from patient phenotypes, locating and interpreting functional data submissions, potential refinements to single-star review criteria, clarifying key dates within ClinVar records, and opportunities for sharing structured and unstructured evidence more effectively.

Summary:

The meeting focused on improving ClinVar submission processes and addressing challenges faced by laboratories. Melissa Landrum, from NCBI, provided an overview of ClinVar submission workflows, including data validation and manual review steps, while Heidi discussed the ClinGen curation project for resolving conflicts in variant classifications. Participants discussed barriers to submission, such as preparing structured data and time constraints, and explored potential solutions like developing submission apps or integrating with existing software platforms. The group also addressed questions about assertion criteria, publication embargoes, and the incorporation of modified ACMG guidelines into submissions. The session concluded with an open discussion on how to enhance ClinVar submissions and improve data standards, with attendees encouraged to provide further feedback to Melissa and Heidi.

ClinVar Variant Classification Improvement Project

Melissa and Heidi presented on ClinVar's database and a project to address outdated or conflicting variant classifications. The project focuses on identifying obvious issues in 1% of ClinVar variants with medically significant conflicts, with over 25,000 records reviewed and 5,694 flagged. The process allows submitters 60 days to intervene if they disagree with the flagging, after which the flagged submission is marked as such but remains in the database. The project has successfully resolved conflicts in over 3,000 records, with no further challenges in the past year.

ClinVar Data Submission Challenges

The group discussed barriers to submitting data to ClinVar, with the top barriers being preparing structured data and time/staff capacity. Participants were surveyed on their submission processes, with a majority using software platforms and manual effort. The group discussed guidance needs for documenting assertion criteria, with examples being the most helpful. There was also discussion about ClinVar's embargo process for publications, where submissions can be held for up to 6 months if needed. The conversation ended with a question about structured evidence codes and support in ClinVar.

ClinVar Data Structure Development

The group discussed the development of structured evidence codes and scoring systems for ClinVar's Version 4 (SVC4). Melissa explained that while ClinVar is considering implementing structured data requirements, they are hesitant to add more burdens to laboratories already struggling with data submission. Heidi and Melissa agreed to work with software engineers and the ClinVar team to align on a common data structure that would enable laboratories to use consistent scoring and evidence codes. Larry raised questions about the potential ClinVar app, exploring different scenarios from on-premise solutions to cloud-based tools, with the primary goal of improving data submission workflows.

ClinVar App Development Collaboration

The group discussed the development of an app to support ClinVar submissions, with Melissa clarifying that NCBI does not have the resources to develop a new product and instead plans to collaborate with existing software developers. The group explored the potential functionality of the app, with Heidi suggesting it could serve as an intermediary tool for managing variant data and facilitating submissions. Melissa emphasized the need to understand current gaps in existing platforms and work with developers to improve ClinVar submission capabilities. The discussion also touched on assertion criteria for ClinVar submissions, with Melissa outlining current recommendations for handling gene-specific specifications and modifications to ACMG guidelines.

Topic: AI and LLMs in Variant Classification

2025-11-24 at 12:00 UTC

Recording - Passcode: ^Q.@C8j. 

Survey 

Session Description: Genome and exome sequencing generate millions of variants per patient, making classification under ACMG/AMP guidelines a major challenge that requires integrating diverse evidence and often hours of expert curation. While automation retrieves structured data, extracting insights from unstructured literature remains slow and inconsistent. Large language models (LLMs) offer a promising way to accelerate variant interpretation by processing studies, summarizing findings, and even suggesting classifications—while still relying on expert oversight. This session will showcase three AI tools—EvAgg, DeepSeek-R1, and AutoPM3—that demonstrate how LLMs can streamline evidence gathering and improve consistency in clinical genomics.

Summary:

The Clinical Genomics Laboratory Meeting focused on the use of AI in genomics, particularly large language models (LLMs) like DeepSeek and GPT-4. Dr. Brian Chung and Dr. Wei Ma from the Hong Kong Genomic Institute discussed their AI CURA tool, which uses LLMs to automate variant classification, achieving 96% concordance with human experts. Pam Ajuyah Robertson from the Broad Institute highlighted Microsoft's EvAgg tool, which reduced curation time by 34%. Shumin Li introduced AutoPM3, a system that automates PM3 criteria, achieving over 50% improvement in accuracy. The tools aim to enhance efficiency and scalability in genomics.

Introduction and Meeting Overview

• Heidi Rehm introduces the meeting, emphasizing the role of the Clinical Genomics Laboratory Community (CGLC) in addressing the needs of clinical laboratories.
• Jordan Lerner-Ellis outlines the session's focus on the use of AI in genomics, highlighting the challenges in variant interpretation due to the vast amount of data.
• Jordan explains how AI, particularly large language models (LLMs), are transforming the landscape by automating evidence extraction from unstructured literature.
• The session will showcase three AI tools: DeepSeek, OEAG, and AutoPM, and feature speakers from the Hong Kong Genomic Institute.

DeepSeek-R1: Brian Chung & Wei Ma

• Dr. Brian Chung provides an overview of the Hong Kong Genomic Project, funded by the Health Bureau, aiming to recruit 40,000-50,000 individuals for whole genome sequencing by 2026.
• Dr. Chung discusses the use of AI, particularly LLMs, in genomic sequencing, highlighting the benefits of tools like DeepSeek for transparency and decision-making.
• The presentation includes a graph showing the performance of LLMs compared to human experts in ACMG classification.

• Dr. Wei Ma explains the development of AI CURA, a semi-automated workflow for genomic curation, focusing on literature-dependent ACMG rules.
• The AI CURA tool uses prompt engineering and an augmented generation knowledgebase to assist in genomic curation.
• Dr. Ma provides examples of how AI CURA optimizes the interpretation of ACMG rules, reducing AI hallucinations and improving accuracy.
• The tool has achieved high concordance with human experts in variant classification, demonstrating its effectiveness in automating the curation process.

EvAgg: Pamela Robertson

• Dr. Pam Ajuyah Robertson from the Broad Institute discusses the use of Microsoft's EvAgg tool for efficient gene curation in ClinGen.
• Dr. Robertson explains the semi-quantitative evidence-based framework used by ClinGen for gene disease relationships, emphasizing the labor-intensive nature of gene curation.
• The EvAgg tool automates literature review, extracting relevant data from publications, and has shown significant time savings in gene curation.
• Dr. Robertson highlights the need for further development to address challenges in literature search and data extraction, aiming to create a comprehensive curator AI assistant.

AutoPM3: Shumin Li

• Shumin Li introduces AutoPM3, a large language model-based system designed to assist in rare disease variant interpretation.
• The system addresses the challenge of literature-based evidence in ACMG criteria, particularly PM3, by automating the extraction of genotype relationships from publications.
• AutoPM Three uses Mutalyzer to handle variant notation inconsistencies and a variant-specific retriever for text queries, achieving high accuracy in variant identification.
• The tool is fully open-sourced and provides a demo server for users to test its capabilities.

Discussion and Future Directions

• Jordan Lerner-Ellis opens the floor for discussion, addressing questions about the availability and open-source nature of the tools presented.
• Dr. Wei Ma confirms that the Hong Kong Genomic Institute's work is open-source and discusses the advantages of using open-source LLMs for data security.
• Dr. Pam Ajuyah Robertson mentions the ongoing efforts to develop a policy statement on the environmental impact of AI in variant curation.
• The session concludes with a call for suggestions for future meeting topics and a reminder to reach out for topic proposals.

Somatic Cancer Variant Classification and Data Sharing Platforms

2025-09-15 at 12:00 UTC

Zoom recording 

Passcode: #KQ.2LSk

Session Description: This session will focus on somatic cancer variant classification and data sharing platforms including VarCat and ClinVar including CIViC's system for submitting somatic cancer variant data to ClinVar.

1) Alex Wagner — ClinGen Somatic + GA4GH Standards → ClinVar

• ClinGen Somatic programme: aims to systematically classify clinical significance of somatic variants using shared guidance and expert panels across pediatric, hematologic, and solid tumor domains.

• Scale so far: ~1,300 papers curated; 536 evidence records accepted; ~250 active members contributing via the CIViC curation interface.

• Evidence model: CIViC captures variant, disease/tumor context, phenotypes, directionality, trust ratings, and therapy associations from PubMed/ASCO sources (no unpublished data).

• Classification frameworks used:

• Clinical significance (AMP/ASCO/CAP): tiers 1–4 with evidence levels A–D for therapeutic, diagnostic, prognostic assertions.

• Oncogenicity (ClinGen/CGC/VICC 2022): additive, ACMG-inspired scoring adapted for somatic drivers.

• Standardization stack (GA4GH):

• VA-Spec (variant annotation) for assertions/evidence lines + provenance;

• VRS (variant representation) for consistent, federated variant IDs/search;

• “Cat-first” (categorical variation): modeling canonical allele groupings and rule-based variant sets (e.g., EGFR L858R vs canonical allele contexts).

• CIViC → ClinVar flow: SCV-SEPs endorse curated assertions in CIViC; endorsed packets are exported as GA4GH-compliant objects and submitted to ClinVar (manual or automated). Goal: 3-star ClinVar records once SEP processes finalize

2) Melissa Landrum — ClinVar Somatic Support & Call for Submissions

• ClinVar today: >5.6M submissions representing ~3.7M variants from 300+ organizations; historically germline-heavy.

• Somatic support (since Jan 2024): variant pages now show separate classification types:

1. Germline (pathogenicity for inherited disease)

2. Somatic oncogenicity

3. Somatic clinical impact (tiered; specify diagnostic/prognostic/therapeutic and effect such as response/resistance, better/poor outcome).

• Gap: Only ~1,200 somatic variants and 15 submitting orgs to date. Clinical labs are critical to reach critical mass; expert panel work relies on lab submissions.

• What to submit (somatic): variant description, cancer type, oncogenicity or clinical impact classification, evaluation date, evidence summary/citations, assertion criteria, and observation metadata (method, allele origin, affected status; optional normal-tissue presence and VAF).

• How to submit:

1. Register org; 2) Prepare data (docs + email support available); 3) Choose method: wizard (one-offs), Excel templates (batch), or API (high-volume).

3) Katie Schieffer — VarCat in Clinical Practice (Nationwide Children’s)

• Motivation: Replace non-standardized Excel-based worksheets with a structured, reusable system.

• VarCat (“Variation Categorizer”): uses VICC normalization + GA4GH standards to structure evidence and compute both oncogenicity and AMP/ASCO/CAP classifications.

• Pulls external evidence (CIViC, PubMed, cBioPortal, MOAlmanac), provides manual curation with Hypothes.is annotations, and supports internal evidence reuse.

• Includes “Firefly” visualization for hotspots/domain context.

• Validation & go-live: 50 prospective variants double-classified (legacy Excel vs VarCat).

• Live since June 2025, with stepwise scaling as oncogenicity adds time/complexity.

• Local SOP refinements: clarified domain sources for OM1, guidance on downgrading OS2 (functional studies), broadened OP2 to capture second-hit scenarios (CNAs/second variants/in-trans evidence/literature); still supporting-level weight.

• Throughput toward ClinVar: >3,400 evidence lines supporting ~3,200 somatic clinical-significance classifications; ~904 diagnostic records currently in the ClinVar submission pipeline via GA4GH formats. Breadth covers common genes (e.g., TP53, PIK3CA, BRAF, CTNNB1, KIT) and diverse tumor types (CNS, sarcoma, neuroblastoma, rare tumors).

Discussion / Q&A

• “Can others adopt these pieces?”

• Open now: GA4GH standards (VRS, VA-Spec), and the ClinVar submission utility

• Planned: VarCat interface to be open-sourced; will be configurable so external labs can connect to their own back-ends/annotation stores.

• Non-oncogenic mosaic variants (e.g., UBA1, TLR8) and immune cohorts: These are not typically handled under somatic cancer frameworks; recommendation is to use germline ACMG/AMP frameworks and submit via germline pathways.

Key Takeaways for Labs

• Why submit to ClinVar? Enables cross-lab comparison, supports expert panels, and mirrors the impact ClinVar has had for germline.

• What you’ll need: A structured pipeline that maps your evidence and assertions to GA4GH models. Batch submission via templates/API is most efficient.

• Interoperability matters: Using VRS + VA-Spec future-proofs your assertions (reusability, deduplication, provenance, automated ClinVar packaging).

  ---

Genomic Analysis Platform Bake Off

2025-07-14 at 12:00 UTC

Meeting Recording

Presenter Slides

Session Description: Clinical laboratories are routinely burdened by the difficulty in evaluating new platforms being considered for use for genomic analysis. It can be labor intensive to evaluate many platforms simultaneously and difficult to get permission to share clinical cases with external platforms under consideration. We are proposing to join together as a clinical lab community to collaboratively evaluate platforms and share the results. To do this, we’ll ask labs to evaluate the platform(s) in use in their lab using a common set of cases. Labs may evaluate other platforms as they wish.

The core of this activity will be supported through participation in the CAGI-Rare Genomes Project (RGP) Challenges. In this challenge, variants in the form of a joint called vcf from short-read genome sequence data and phenotype data from ~30 solved and unsolved RGP families will be provided. Participants in the challenge will try to identify the causative SNV/indel variant(s) in each family. In a separate challenge, genomic sequence data in the form of CRAMs from 10-15 solved and unsolved probands will be provided. Participants in this challenge will try to identify causative SNV/indel, structural variants, mitochondrial genome variants, and tandem repeat expansions. The goal of the challenge is to evaluate genomic analysis platforms for their sensitivity and specificity. Teams can submit results from up to 5 models/platforms for each challenge. More details can be found here.

We expect these challenges to launch in July 2025 and close in September 2025, with results reported at the annual CAGI meeting in Boston December 6-8, 2025. If you may be interested in participating in this challenge, please fill out the brief survey here.

Meeting Minutes:

Survey

The following platforms are in use by groups that have indicated interest in participating:

• eMedGene
• Franklin
• VarSome
• Qiagen Clinical Insight
• Geneyx
• Golden Helix VarSeq
• Exomiser
• PierianDx
• Breakthrough Genomics Virtual Assistant
• GENEX (Japan)

Possible addition areas of platform evaluation:

• Variant Curation, Interpretation, and Reporting Capabilities - ACMG classification support, evidence handling, lit review, and variant database integration, report generation, variant knowledge database, reanalysis workflows
•  Filtering, Prioritization, and Decision Support - How easy or difficult is it to get to the answer? Variant ranking, predictive scores, AI, filtering tools, presets, and features that aid in decision-making
• Ease of Use and Workflow Efficiency - UI, UX, case setup, analyst navigation, in-platform communication, notifications, overall usability, reduction of overall clicks and out-of-platform internet/database searching
• Integration, Data Input Flexibility, and Test Type Compatibility - LIMS integration/API/connectivity capabilities, support for our preferred variant and file types, including visualization tools, panel/WGS/WES workflows supported            
• Compliance - Regulatory standards (ISO, HIPAA, GDPR)
• Customer Support - Quality of vendor support/new feature turnaround time
• Cost - Pricing model, licensing, and onboarding
• Suggestions from Oliver Hofmann: Some things to maybe add to the 'platform evaluation criteria' -- version management, ability to work on different versions of the platform/software (or at least to preview changes with upcoming releases); ability to take reported variants out of the platform (e.g., as part of a vendor migration), access to log files / audits to ensure submitted data has been received and annotated completely.

Zoom Chat Discussion Questions:

• Will the challenge provide a bed file for analysis regions?

• Anne: I don't think we've used any BED file in the analysis that we've done. So I'm going to say no
• Heidi: the goal is to not provide a bed file in the sense that we want you to utilize the entire genome to look for answers for these cases, and so we don't want to restrict that space to anything really

• What is your experience teasing apart the methods/databases used vs the experience of the curation team (or time invested in the review)?

• In the last challenge, it was hard to separate platform performance from the impact of expert curation. Some teams used only computational tools, while others had experienced geneticists involved. The upcoming challenge aims to better isolate these factors.

• Are there any criteria for participating in the challenge?

• Steven: CAGI is open to everyone.  Clinicians and labs, method companies, computer scientists
• Predrag: I am happy to give details on this. Individuals must register on the CAGI portal and in Synapse which we use to collect predictions and distribute data. CAGI will do our best to verify the participants have legitimate interest in downloading the data.
• Steven: There are a Data Use Agreement and Data Sharing Agreement that must be signed
• Anne: All participants do need to sign a data use agreement that agrees to the good community practices including not trying to re-identify individuals or search for secondary findings (as this is not part of this challenge).
• Predrag: Note that will require all participants to delete the genome data by a certain date post assessment. This is a part of our data use agreement.

• Can you please specify which type of SVs for the CAGI7? In particular del, dup, inversions, translocations, mob elements?

• Anne: We are still putting together the cases for the CRAM challenge part, we were trying to get some representative different types of SBS included. We will not probably have very complex SBS, just because that is why they are somewhat interesting. They might be a little bit harder for systems and we need to be careful that these are things that there will be agreement with these diagnoses. So if it gets too complicated, it gets a little harder to be confident. So it will definitely have deletions, probably duplications. Or our pipelines do look at C inversion, translocations and mobile elements. My guess is we're not going to have all of those represented, but you would want to look at all those types, because those are the things we have looked at in our analysis.

• Will you include any “negative controls” in the small variants challenge, e.g. where diagnosis is not in small variants, to see how well algorithms can tell when to stop looking?

• Yes, negative controls will be included. The test set of 30 cases contains a mix of solved and unsolved cases, and some may not have been previously analyzed. This allows for the possibility of new diagnoses and ensures algorithms can demonstrate when no diagnosis is found. The separate discovery set of 20 cases includes only unsolved or unreviewed cases to encourage novel findings.

• What will be the process of signing up?

• Steven: Sign up on CAGI website, sign data use agreement, then sign data sharing agreement., then access challenge site at Synapse.
• Predrag: There’s two places to sign up; First, participants register on the CAGI website, where their identity will be verified to ensure data is not shared with unknown entities. Second, participants must create an account on the Synapse platform, where the data will be distributed. After registering on both platforms and signing a Data Use Agreement, participants will receive encrypted data via Synapse. Decryption keys will be shared separately through secure channels (e.g., Signal or PGP email). All data must be stored in encrypted form, per the data transfer agreement with the Broad Institute.
• Anette: Is the sign up for the institution or just the individual signing up?

• Predrag: We don’t know yet but my latest update on that is that it may not be needed.

• What about results disclosure? Will all participating teams/software/labs be disclosed?

• Steven: Generally everything will be disclosed, with some exceptions…We have allowed deliberately anonymous predictions, which will be kept anonymous unless they are a top predictor. But groups that chose to be anonymous cannot be chosen to be identified later. The idea of anonymous submissions is for developers to try "wild and crazy" early-development ideas for internal development without risking getting a poor ranking.
• Susanna: So this time anonymous groups will have to register as anonymous groups since registration and this will not be modified during the challenge?
• Steven: It's the submission itself that had been flagged as anonymous at time of submission -- we internally do need to know who they are, as high-performing predictors have been revealed regardless.

Links:

• Here is the CGLC Survey to express interest in the CAGI-RGP challenge (not the official signup) and/or the working group to address needs of this challenge for the clinical labs https://docs.google.com/forms/d/e/1FAIpQLScatSBFcYOEY6iTQu4OGyA9siVxtJEi-sEqKPbeQxr1qNUxow/viewform 
• CAGI-6 RGP paper: https://pubmed.ncbi.nlm.nih.gov/38685113/
• Integrative paper: The Critical Assessment of Genome Interpretation Consortium. CAGI, the Critical Assessment of Genome Interpretation, establishes progress and prospects for computational genetic variant interpretation methods. Genome Biol 25, 53 (2024). https://doi.org/10.1186/s13059-023-03113-6 (Covers CAGI 1-5, Predates RGP in CAGI 6)

Sharing Patient Data in ClinVar

2025-06-16

Meeting Recording

Presenter Slides

This session will review best practices for sharing patient data in ClinVar. Danielle Azzariti will review ClinGen guidance on sharing patient data in ClinVar and Jordan Lerner-Ellis will review the current landscape in Canada for sharing variant data from clinical labs into ClinVar and associated privacy legal considerations as it pertains to the Ontario Privacy commissioner's guidance on sharing PHI related data. Diya Uberoi will review work of the GA4GH Regulatory and Ethics WG on developing guidance for sharing clinical lab data.

Meeting Minutes:

Actions:

• Develop a common template that labs could include in consent forms to facilitate data sharing.

• Publish a policy brief on the privacy review process and risk assessment conducted at Mount Sinai Hospital.

Outline:

Sharing Patient Data in ClinVar: Introduction and Goals

• The meeting aims to address topics useful for the clinical genomics laboratory community, focusing on engagement, work groups, task teams, and useful resources.

• Heidi Rehm introduced the topic of sharing patient data in ClinVar, highlighting the lack of clarity on what can be shared and the challenges faced by labs.

• The goal of the session is to enhance the provision of clinical data in submissions and through additional mechanisms.

Review of ClinGen guidance on sharing patient data in ClinVar - Danielle Azzariti:

• Danielle presented a framework developed by ClinGen, NCBI, ClinVar, and NHGRI to guide clinical laboratories on sharing genomic data collected in a clinical setting.

• The framework outlines data elements that can be shared in a de-identified submission of variant-level data to ClinVar, focusing on minimizing the risk of re-identification.

• Danielle explained the importance of sharing clinical details and cases to support variant classifications and the challenges of relying on emailing ClinVar submitters for clinical data.

• The framework encourages clinical laboratories to share variant-level data without explicit consent, emphasizing the need for conservative data sharing to prevent re-identification.

Sharing clinical genomic data in Canada - Jordan Lerner-Ellis:

• Jordan discussed the Canadian Genomic Data Commons (CGDC) project, which aims to create core facilities for the generation and sharing of genomic data.

• The Canadian Open Genetics Repository (COGR) database shares data between over 20 clinical labs across Canada, with the goal of sharing consensus variant data and issuing variant discrepancy reports to public databases like ClinVar.

• Jordan highlighted the resource-intensive nature of ClinVar submissions, including the need for variant validation, multiple steps in the submission process, and the importance of text-based interpretations.

• Jordan focused on the privacy risks of sharing clinical genomic data, using Mount Sinai Hospital as a test case for patient privacy frameworks and submitting patient variant data to ClinVar.

• Jordan explained the privacy risks associated with sharing clinical genomic data, emphasizing the importance of de-identification to prevent re-identification.

• The Information and Privacy Commissioner of Ontario's cell size rule of five is discussed, which states that the risk of re-identification increases if the number of individuals sharing the same quasi-identifiers is less than five.

• Jordan outlines strategies to mitigate privacy risks, including generalizing phenotypes, grouping similar phenotypes, and suppressing exact case counts.

• The importance of documenting privacy risks appropriately and understanding the legal and regulatory context is emphasized.

Informing Clinical Data Sharing Standards - Diya Uberoi:

• Diya discussed the work of the GA4GH Regulatory & Ethics Work Stream (REWS) clinical data sharing group, which aims to clarify the extent to which genomic and patient-level data can be shared in the context of clinical care without explicit consent.

• The group conducted a survey and analyzed institutional policies from 33 clinical genomic institutions to understand the clinical context in which genomic data can be shared.

• Diya highlighted the variance in institutional policies, with some institutions permitting data sharing without consent and others requiring explicit consent.

• The group aims to develop best practice standards to guide the uptake of clinical genomic data sharing, with a focus on harmonizing frameworks and addressing privacy concerns.

Discussion and Open Questions

• Aqsa asked about the minimum information required to submit clinical data to ClinVar, and Heidi clarified that while detailed patient-level data is not required, it can be valuable to include summary-level clinical details.

• Jordan added that there is often uncertainty among clinical labs about what data can be shared and the need for education and clear policies.

• Heidi encouraged participants to reach out with any specific questions or challenges related to sharing clinical data in ClinVar, and to submit their clinical data to support evidence sharing.

Development of gene-level guidance and a variant exception list for classification of variants in the ACMG Secondary Findings genes:

2025-05-12

Meeting Recording

Presenter Slides

Meeting Minutes:

David Miller's Presentation - Overview of ACMG SF Policies and Lists

• David Miller is one of the co-chairs of the ACMG Secondary Findings Maintenance Working Group.
• The ACMG secondary findings list started as incidental findings in 2013 and was renamed secondary findings to indicate intentional reporting.
• The goal of the list is to identify and manage risks for highly penetrant, high morbidity, and high mortality conditions.
• The list is updated approximately every year to keep it current, with a major update in 2021 adding 14 new genes.

Challenges and Updates in ACMG Secondary Findings:

• The original list was created through an iterative consensus process, but a semi-quantitative framework was adopted in 2017 to assess actionability.
• The list is open to community nominations for gene-disease pairs, with criteria including penetrance, types of variants, and confirmatory diagnostic testing.
• Technical issues, such as low penetrance and clinical management concerns, are considered when deciding whether to include a gene-disease pair.
• The list aims to be a minimum list to avoid overwhelming labs with too much to report, while ensuring the reported conditions are actionable and meaningful.

Heidi Rehm’s Presentation - New website to support ACMG SF Reporting (Slides)

• Heidi introduced a new resource developed by ClinGen to support ACMG secondary findings reportability.
• The resource includes specific guidance for each gene, such as which diseases are reportable and which variants should not be reported.
• The resource is available on the ClinGen website and includes a downloadable spreadsheet for easier access.
• The goal is to provide comprehensive guidance for every gene, regardless of whether it is a secondary finding gene.

Jennifer Johnston's Presentation - Development of AF thresholds for all ACMG SF genes (Slides)

• Jennifer discussed the challenges of classifying variants in secondary finding genes, especially for smaller laboratories.
• The project aims to provide minimum classification criteria for secondary finding genes, starting with setting allele frequency thresholds for BA1 and BS1.
• The project includes 46 genes, with 34 already having VCEP guidance and 15 having no guidance.
• The allele frequency thresholds are calculated using disease allele frequency thresholds (DAFT) and compared to pathogenic variants in gnomAD.

Alexis Carere’s Presentation - Commercial laboratory strategies to support SF interpretation and return (Slides)

• Alexis from GeneDx discusses their strategies for supporting secondary finding interpretation and return.
• The project aimed to simplify the process, reduce human error, and minimize touchpoints for clinical experts.
• Gene disease curation was reassessed, and a comprehensive pre-launch clinical expert review was conducted.
• The new workflow includes annotating variants with secondary finding reportability and standardizing the evaluation process.

Open Discussion and Closing Remarks

• The HFE gene was not curated by ClinGen, but it will be added to the website soon.
• Specific guidance for TTN truncating variants outside the A band is available, and the goal is to provide exon-specific guidance.
• The Clinical Actionability Working Group's work is similar to ACMG, focusing on actionability but considering other factors as well.
• The next meeting in June will focus on clinical policies for clinical data sharing, with the ClinVar team participating.

Sequence Variant Classification v4.0 Standards:

2025-03-10

This session will give an overview of the draft version of the ACMG/AMP/CAP/ClinGen Sequence Variant Classification v4.0 Standards that are forthcoming and will describe how to participate in the community pilot of the standards. The ACMG/AMP/CAP/ClinGen committee is seeking feedback on the performance of SVCv4.0. Members of the community that have already signed up, as well as those still interested in participating are welcome to join the video conference session.

Meeting Recording

Presenter Slides