ClinGen API platform for classification of human genetic variants.

The Clinical Genome Resource (ClinGen): Advancing Genomic Knowledge through Global Curation

Introduction: The Clinical Genome Resource (ClinGen) Somatic Working Group (sWG) is a multi-institution team engaged in developing processes, resources, and standards to support accurate classification of somatic variants in cancer. Existing decision support resources in cancer knowledgebases are heavily skewed towards genes and variants relevant in adult cancers; however, information to support variant interpretation in childhood cancers is limited. Here we report on the goals and progress of the Pediatric Cancer Taskforce, created within the ClinGen sWG, to lead curation efforts of actionable alterations in childhood cancers. Methods: The ClinGen sWG Pediatric Cancer Taskforce (PCT) consists of a core group of twelve members comprising geneticists, pathologists, and oncologists with expertise in different pediatric cancers and with representation from 9 leading pediatric institutions. The taskforce has a total of 35 members including volunteer-curators who work under guidance of the expert members. Curation of childhood cancer variants is conducted in collaboration with the Clinical Interpretation of Variants in Cancer (CIViC) team at Washington University in Saint Louis, using the CIViC knowledgebase (civicdb.org) and the ClinVar database as open-access curation and data-sharing platforms. Diagnostic, prognostic, and therapeutic evidence is tiered according to the AMP/ASCO/CAP guidelines for the clinical interpretation of somatic variants. PCT members are assigned specific genetic variant-tumor type associations for curation, which are then reviewed in monthly conferences to finalize assertions in CIViC. Results: The PCT has prioritized 40 genetic alterations relevant to pediatric cancer for curation based on their clinical relevance and the lack of sufficient existing curated evidence in clinical knowledgebases. To date, 4 assertions have been created and added to the database: HEY1-NCOA2 fusion in mesenchymal chondrosarcoma, KIAA1549-BRAF fusion and ACVR1 p.G328V variant in pediatric glioma, and EBF1-PDGFRB fusion in pediatric B-cell precursor acute lymphoblastic leukemia. Active curation has been initiated for NTRK fusions agnostic of tissue histology, targetable kinase fusions in Ph-like B-lymphoblastic leukemia, and common variants in selected pediatric sarcomas and brain tumors, focusing heavily on driver gene fusions in childhood cancers. 119 evidence items have been created in CIViC by the members. The PCT also works to implement more standardized and accurate classification of pediatric cancers in CIViC and other cancer resources, and to enhance search for pediatric-specific data through appropriate tagging of evidence using ontology terms. Conclusions: As molecular alterations are increasingly relevant to the care of children with cancer, the ClinGen PCT will work to develop standards, processes, and resources for efficient and accurate determination of clinical relevance of pediatric cancer variants. Citation Format: Alanna J. Church, Shruti Rao, Deborah Ritter, Arpad Danos, Kilann Krysiak, Laura B. Corson, Kevin E. Fisher, Matthew Hiemenz, Katherine A. Janeway, Jianling Ji, Chimene A. Kesserwan, Theodore W. Laetsch, Donald W. Parsons, Ryan J. Schmidt, Kristen L. Sund, Wan-Hsin Lin, Malachi Griffith, Obi L. Griffith, Shashikant Kulkarni, Subha Madhavan, Angshumoy Roy, Gordana Raca. Curation of pediatric cancer variants within the Clinical Genome Resource (ClinGen) [abstract]. In: Proceedings of the AACR Special Conference on the Advances in Pediatric Cancer Research; 2019 Sep 17-20; Montreal, QC, Canada. Philadelphia (PA): AACR; Cancer Res 2020;80(14 Suppl):Abstract nr A58.

BackgroundThe 2015 American College of Medical Genetics and Genomics (ACMG) and the Association for Molecular Pathology (AMP) guidelines for clinical sequence variant interpretation state that “well-established” functional studies can be used as evidence in variant classification. These guidelines articulated key attributes of functional data, including that assays should reflect the biological environment and be analytically sound; however, details of how to evaluate these attributes were left to expert judgment. The Clinical Genome Resource (ClinGen) designates Variant Curation Expert Panels (VCEPs) in specific disease areas to make gene-centric specifications to the ACMG/AMP guidelines, including more specific definitions of appropriate functional assays. We set out to evaluate the existing VCEP guidelines for functional assays.MethodsWe evaluated the functional criteria (PS3/BS3) of six VCEPs (CDH1, Hearing Loss, Inherited Cardiomyopathy-MYH7, PAH, PTEN, RASopathy). We then established criteria for evaluating functional studies based on disease mechanism, general class of assay, and the characteristics of specific assay instances described in the primary literature. Using these criteria, we extensively curated assay instances cited by each VCEP in their pilot variant classification to analyze VCEP recommendations and their use in the interpretation of functional studies.ResultsUnsurprisingly, our analysis highlighted the breadth of VCEP-approved assays, reflecting the diversity of disease mechanisms among VCEPs. We also noted substantial variability between VCEPs in the method used to select these assays and in the approach used to specify strength modifications, as well as differences in suggested validation parameters. Importantly, we observed discrepancies between the parameters VCEPs specified as required for approved assay instances and the fulfillment of these requirements in the individual assays cited in pilot variant interpretation.ConclusionsInterpretation of the intricacies of functional assays often requires expert-level knowledge of the gene and disease, and current VCEP recommendations for functional assay evidence are a useful tool to improve the accessibility of functional data by providing a starting point for curators to identify approved functional assays and key metrics. However, our analysis suggests that further guidance is needed to standardize this process and ensure consistency in the application of functional evidence.

Clinical cancer genomic analysis: data engineering required

Large language models (LLMs) have been extensively tested for incorporating into medical applications in recent years, yet their potential in clinical genetics, particularly in diagnosing rare diseases, remains underexplored. Recent advancements in LLMs have improved their reasoning capabilities and transparency, facilitating significant enhancements in clinical workflow designs. The open-sourced DeepSeek model also serves as a cost-effective alternative of top-ranked proprietary reasoning LLMs such as o3-mini-high for genome projects and hospitals that have specific needs in data security. In this study, we developed a framework that fully automates genetic variant classification according to the American College of Medical Genetics and Genomics (ACMG) and the Association for Molecular Pathology (AMP) guidelines and Clinical Genome Resource (ClinGen) recommendations. Two state-of-the art LLMs, DeepkSeek-R1 and o3-mini-high were tested for their performance in variant classification. We demonstrated that through careful prompt engineering and creation of ACMG-rule specific knowledgebases, DeepSeek-R1 outperformed o3-mini-high and achieved high sensitivity and 100% specificity in interpreting ACMG rules that require understanding literature-based evidence. Further testing using 150 variants curated by ClinGen experts, DeepSeek-R1 demonstrated performance on par with human curators. Finally, we showed the framework can be also used for reanalysis using 150 ClinVar variants with conflicting interpretations. Our study provided the first LLM framework capable of fully automated variant classification in the diagnosis of genetic diseases and variant reanalysis.

Background In precision oncology, the accurate and reproducible classification of variant oncogenicity is fundamental for therapy decision making. In 2022, a set of guidelines for the classification of oncogenicity of somatic variants in cancer were defined by the Clinical Genome Resource, the Cancer Genomics Consortium, and the Variant Interpretation for Cancer Consortium. However, to date an implementation that automates the evaluation of these criteria does not exist. Furthermore, for the majority of the criteria, the interpretation of the criterion-associated textual indication and the choice of publicly available resources to gather criterion-supporting information depends on the user. Thus, the risk of a variant being classified differently by independent laboratories is relevant, with implications for the management of patient care. Methods Here, we developed Oncogenicity Variant Interpreter (OncoVI), a fully-automated Python-based implementation of the oncogenicity guidelines. First, each criterion was interpreted and publicly available resources were identified to be utilised as reference. Then, criteria were implemented in OncoVI and the information reported by the associated resources were integrated. OncoVI is part of a broader Python-framework that, starting from the genomic position of the variant, automatically performs functional annotation, collects the available evidence from the reference resources, and provides a classification of oncogenicity. Results On the set of 93 somatic variants provided by the guidelines OncoVI achieved an overall accuracy of 80%, with a sensitivity of 88% in the classification of Oncogenic/Likely Oncogenic variants. On a real-world data set of 7,802 variants from 557 patients previously evaluated within the Molecular Tumour Board (MTB) Erlangen, an agreement of 79% was observed between the oncogenicity classification of OncoVI and the MTB pathogenicity assessment. In addition, the pathogenicity classification of 135 variants of the MTB data set was re-assessed by expert biologists adhering solely to the oncogenicity guidelines. This re-evaluation confirmed the validity of OncoVI"s interpretation of the resources chosen as reference, but it also underlined the ability of experts in solving conflicting evidence. Conclusions Taken together, OncoVI provides an effective implementation of the oncogenicity guidelines, thus facilitating their adoption and supporting the reproducible and harmonised oncogenicity classification of somatic variants between institutions.

Arrhythmogenic right ventricular cardiomyopathy (ARVC) is an inherited disease characterized by ventricular arrhythmias and progressive ventricular dysfunction. Genetic testing is recommended, and a pathogenic variant in an ARVC-associated gene is a major criterion for diagnosis according to the 2010 Task Force Criteria. As incorrect attribution of a gene to ARVC can contribute to misdiagnosis, we assembled an international multidisciplinary ARVC Clinical Genome Resource Gene Curation Expert Panel to reappraise all reported ARVC genes. Following a comprehensive literature search, six 2-member teams conducted blinded independent curation of reported ARVC genes using the semiquantitative Clinical Genome Resource framework. Of 26 reported ARVC genes, only 6 (PKP2, DSP, DSG2, DSC2, JUP, and TMEM43) had strong evidence and were classified as definitive for ARVC causation. There was moderate evidence for 2 genes, DES and PLN. The remaining 18 genes had limited or no evidence. RYR2 was refuted as an ARVC gene since clinical data and model systems exhibited a catecholaminergic polymorphic ventricular tachycardia phenotype. In ClinVar, only 5 pathogenic/likely pathogenic variants (1.1%) in limited evidence genes had been reported in ARVC cases in contrast to 450 desmosome gene variants (97.4%). Using the Clinical Genome Resource approach to gene-disease curation, only 8 genes (PKP2, DSP, DSG2, DSC2, JUP, TMEM43, PLN, and DES) had definitive or moderate evidence for ARVC, and these genes accounted for nearly all pathogenic/likely pathogenic ARVC variants in ClinVar. Therefore, only pathogenic/likely pathogenic variants in these 8 genes should yield a major criterion for ARVC diagnosis. Pathogenic/likely pathogenic variants identified in other genes in a patient should prompt further phenotyping as variants in many of these genes are associated with other cardiovascular conditions.

BackgroundIn 2015, the American College of Medical Genetics and Genomics (ACMG) and the Association for Molecular Pathology (AMP) developed standardized variant curation guidelines for Mendelian disorders. Although these guidelines have been widely adopted, they are not gene- or disease-specific. To mitigate classification discrepancies, the Clinical Genome Resource FBN1 variant curation expert panel (VCEP) was established in 2018 to develop adaptations to the ACMG/AMP criteria for FBN1 in association with Marfan syndrome.MethodsThe specific recommendations were developed through literature review, surveys, online expert panel discussions, and pilot testing of a set of 60 different variants. Consensus among experts was considered reached if at least 75% of the members agreed with a given rule specification. The final set of rules received approval from the ClinGen Sequence Variant Interpretation Working Group.ResultsThe developed specifications introduce modifications to 14 of the 28 ACMG/AMP evidence criteria and deem 6 criteria non-applicable. Some of these specifications include refining the minor allele frequency thresholds, creating a FBN1-specific flowchart for PVS1, defining functional domains of the protein, developing a point-based system of counting probands and instances of de novo occurrences, recommending a points-based method of accounting for family segregation data, and clarifying the applicable functional assays that should be considered. To date, this VCEP has curated 120 variants which have been deposited to ClinVar with the 3-star review status.ConclusionsEstablishing specific adaptations for FBN1 has provided a framework to foster greater classification concordance among clinical laboratories, ultimately improving clinical care for patients with Marfan syndrome.

Clingen Coagulation Factor Deficiency Variant Curation Expert Panel: Meeting the Need for Recommendations to Curate Variants in the Coagulation Factor Genes

Introduction:The Clinical Genome Resource (ClinGen) is creating a central resource of clinically relevant genetic knowledge to improve genomic medicine. Dissemination and use of the ClinGen Resource is essential to ensure broad community uptake. We report on experiences and sustained use of ClinGen tools through engaging international genetics groups based in India, Africa and Singapore in variant classification training workshops using the ClinGen Variant Curation Interface (VCI).Methods:We developed pre and post workshop questionnaires and analyzed ClinGen tool use following the workshops. We evaluated organizational aspects and costs of creating a dedicated ClinGen VCI instance for each workshop.Results:The workshops yielded >200 participants, with local scientists as essential participants. While ~55% of participants were unfamiliar with variant classification, we found ~79% were likely to use the VCI after the workshop. Further, we identified about ~10% of workshop participants created permanent accounts. We estimate costs at ~$3 per VCI instance.Discussion:Our efforts highlight the yield of international workshops to sustained use of ClinGen’s curation tools, and identify areas for future consideration such as creating user-groups by experience level, and the importance of local scientist engagement in workshop deployment and organizational aspects.

BackgroundMultiplexed Assays of Variant Effects (MAVEs) can test all possible single variants in a gene of interest. The resulting saturation-style functional data may help resolve variant classification disparities between populations, especially for Variants of Uncertain Significance (VUS).MethodsWe analyzed clinical significance classifications in 213,663 individuals of European-like genetic ancestry versus 206,975 individuals of non-European-like genetic ancestry from All of Us and the Genome Aggregation Database. Then, we incorporated clinically calibrated MAVE data into the Clinical Genome Resource’s Variant Curation Expert Panel rules to automate VUS reclassification for BRCA1, TP53, and PTEN.ResultsUsing two orthogonal statistical approaches, we show a higher prevalence (p ≤ 5.95e − 06) of VUS in individuals of non-European-like genetic ancestry across all medical specialties assessed in all three databases. Further, in the non-European-like genetic ancestry group, higher rates of Benign or Likely Benign and variants with no clinical designation (p ≤ 2.5e − 05) were found across many medical specialties, whereas Pathogenic or Likely Pathogenic assignments were increased in individuals of European-like genetic ancestry (p ≤ 2.5e − 05). Using MAVE data, we reclassified VUS in individuals of non-European-like genetic ancestry at a significantly higher rate in comparison to reclassified VUS from European-like genetic ancestry (p = 9.1e − 03) effectively compensating for the VUS disparity. Further, essential code analysis showed equitable impact of MAVE evidence codes but inequitable impact of allele frequency (p = 7.47e − 06) and computational predictor (p = 6.92e − 05) evidence codes for individuals of non-European-like genetic ancestry.ConclusionsGeneration of saturation-style MAVE data should be a priority to reduce VUS disparities and produce equitable training data for future computational predictors.

Comparison of somatic variant oncogenicity classification using ClinGen/CGC/VICC guidelines and QIAGEN Clinical Insight Interpret decision support software.

Purpose: The addition of Pompe disease (Glycogen Storage Disease Type II) to the Recommended Uniform Screening Panel in the United States has led to an increase in the number of variants of uncertain significance (VUS) and novel variants identified in the GAA gene. This presents a diagnostic challenge, especially in the setting of late-onset Pompe disease when symptoms are rarely apparent at birth. There is an unmet need for validated functional studies to aid in classification of GAA variants. Methods: We developed an in vitro mammalian cell expression and functional analysis system based on guidelines established by the Clinical Genome Resource (ClinGen) Sequence Variant Interpretation Working Group for PS3/BS3. We validated the assay with 12 control variants and subsequently analyzed eight VUS or novel variants in GAA identified in patients with a positive newborn screen for Pompe disease without phenotypic evidence of infantile-onset disease. Results: The control variants were analyzed in our expression system and an activity range was established. The pathogenic controls had GAA activity between 0% and 11% of normal. The benign or likely benign controls had an activity range of 54%–100%. The pseudodeficiency variant had activity of 17%. These ranges were then applied to the variants selected for functional studies. Using the threshold of <11%, we were able to apply PS3_ supporting to classify two variants as likely pathogenic (c.316C > T and c.1103G > A) and provide further evidence to support the classification of likely pathogenic for two variants (c.1721T > C and c.1048G > A). One variant (c.1123C > T) was able to be reclassified based on other supporting evidence. We were unable to reclassify three variants (c.664G > A, c.2450A > G, and c.1378G > A) due to insufficient or conflicting evidence. Conclusion: We investigated eight GAA variants as proof of concept using our validated and reproducible in vitro expression and functional analysis system. While additional work is needed to further refine our system with additional controls and different variant types in order to apply the PS3/BS3 criteria at a higher level, this tool can be utilized for variant classification to meet the growing need for novel GAA variant classification in the era of newborn screening for Pompe disease.

Gene-specific ACMG/AMP classification criteria for germline APC variants: Recommendations from the ClinGen InSiGHT Hereditary Colorectal Cancer/Polyposis Variant Curation Expert Panel

Multi-Platform Curation in the Development of ACMG/AMP Specifications for Von Hippel-Lindau (VHL) Disease.