Advertisement

HTAADVar: Aggregation and fully automated clinical interpretation of genetic variants in heritable thoracic aortic aneurysm and dissection

  • Wei-Zhen Zhou
    Correspondence
    Correspondence and requests for materials should be addressed to Wei-Zhen Zhou, Center of Laboratory Medicine, Fuwai Hospital, No. 167, Beilishi Road, Xicheng District, Beijing 100037, China
    Affiliations
    Center of Laboratory Medicine, State Key Laboratory of Cardiovascular Disease, Beijing Key Laboratory for Molecular Diagnostics of Cardiovascular Diseases, Fuwai Hospital, National Center for Cardiovascular Diseases, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for articles by this author
  • Yujing Zhang
    Affiliations
    Center of Laboratory Medicine, State Key Laboratory of Cardiovascular Disease, Beijing Key Laboratory for Molecular Diagnostics of Cardiovascular Diseases, Fuwai Hospital, National Center for Cardiovascular Diseases, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for articles by this author
  • Guoyan Zhu
    Affiliations
    Center of Laboratory Medicine, State Key Laboratory of Cardiovascular Disease, Beijing Key Laboratory for Molecular Diagnostics of Cardiovascular Diseases, Fuwai Hospital, National Center for Cardiovascular Diseases, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for articles by this author
  • Huayan Shen
    Affiliations
    Center of Laboratory Medicine, State Key Laboratory of Cardiovascular Disease, Beijing Key Laboratory for Molecular Diagnostics of Cardiovascular Diseases, Fuwai Hospital, National Center for Cardiovascular Diseases, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for articles by this author
  • Qingyi Zeng
    Affiliations
    Center of Laboratory Medicine, State Key Laboratory of Cardiovascular Disease, Beijing Key Laboratory for Molecular Diagnostics of Cardiovascular Diseases, Fuwai Hospital, National Center for Cardiovascular Diseases, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for articles by this author
  • Qianlong Chen
    Affiliations
    Center of Laboratory Medicine, State Key Laboratory of Cardiovascular Disease, Beijing Key Laboratory for Molecular Diagnostics of Cardiovascular Diseases, Fuwai Hospital, National Center for Cardiovascular Diseases, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for articles by this author
  • Wenke Li
    Affiliations
    Center of Laboratory Medicine, State Key Laboratory of Cardiovascular Disease, Beijing Key Laboratory for Molecular Diagnostics of Cardiovascular Diseases, Fuwai Hospital, National Center for Cardiovascular Diseases, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for articles by this author
  • Mingyao Luo
    Affiliations
    Center of Vascular Surgery, State Key Laboratory of Cardiovascular Disease, Fuwai Hospital, National Center for Cardiovascular Diseases, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for articles by this author
  • Chang Shu
    Affiliations
    Center of Vascular Surgery, State Key Laboratory of Cardiovascular Disease, Fuwai Hospital, National Center for Cardiovascular Diseases, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for articles by this author
  • Hang Yang
    Correspondence
    Hang Yang, Center of Laboratory Medicine, Fuwai Hospital, No. 167, Beilishi Road, Xicheng District, Beijing 100037, China
    Affiliations
    Center of Laboratory Medicine, State Key Laboratory of Cardiovascular Disease, Beijing Key Laboratory for Molecular Diagnostics of Cardiovascular Diseases, Fuwai Hospital, National Center for Cardiovascular Diseases, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for articles by this author
  • Zhou Zhou
    Correspondence
    Zhou Zhou, Center of Laboratory Medicine, Fuwai Hospital, No. 167, Beilishi Road, Xicheng District, Beijing 100037, China
    Affiliations
    Center of Laboratory Medicine, State Key Laboratory of Cardiovascular Disease, Beijing Key Laboratory for Molecular Diagnostics of Cardiovascular Diseases, Fuwai Hospital, National Center for Cardiovascular Diseases, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
    Search for articles by this author
Open AccessPublished:October 03, 2022DOI:https://doi.org/10.1016/j.gim.2022.08.024

      Abstract

      Purpose

      Early detection and pathogenicity interpretation of disease-associated variants are crucial but challenging in molecular diagnosis, especially for insidious and life-threatening diseases, such as heritable thoracic aortic aneurysm and dissection (HTAAD). In this study, we developed HTAADVar, an unbiased and fully automated system for the molecular diagnosis of HTAAD.

      Methods

      We developed HTAADVar (http://htaadvar.fwgenetics.org) under the American College of Medical Genetics and Genomics/Association for Molecular Pathology framework, with optimizations based on disease- and gene-specific knowledge, expert panel recommendations, and variant observations. HTAADVar provides variant interpretation with a self-built database through the web server and the stand-alone programs.

      Results

      We constructed an expert-reviewed database by integrating 4373 variants in HTAAD genes, with comprehensive metadata curated from 697 publications and an in-house study of 790 patients. We further developed an interpretation system to assess variants automatically. Notably, HTAADVar showed a multifold increase in performance compared with public tools, reaching a sensitivity of 92.64% and specificity of 70.83%. The molecular diagnostic yield of HTAADVar among 790 patients (42.03%) also matched the clinical data, independently demonstrating its good performance in clinical application.

      Conclusion

      HTAADVar represents the first fully automated system for accurate variant interpretation for HTAAD. The framework of HTAADVar could also be generalized for the molecular diagnosis of other genetic diseases.

      Keywords

      Introduction

      With the rapid development of next-generation sequencing technologies, genetic testing has been extensively used clinically for Mendelian diseases because it plays substantial roles in clinical management, including diagnosis, treatment decisions, family screening, and reproductive guidance. A major challenge is obtaining an efficient and accurate interpretation of variant pathogenicity owing to the time-consuming and laborious processes involved and the inconsistencies across different laboratories.
      • Amendola L.M.
      • Jarvik G.P.
      • Leo M.C.
      • et al.
      Performance of ACMG-AMP variant-interpretation guidelines among nine laboratories in the clinical sequencing exploratory research consortium.
      In 2015, the American College of Medical Genetics and Genomics/Association for Molecular Pathology (ACMG/AMP) promulgated their guidelines to address the issue of variant interpretation standardization.
      • Richards S.
      • Aziz N.
      • Bale S.
      • et al.
      Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology.
      However, inconsistencies still remain
      • Amendola L.M.
      • Jarvik G.P.
      • Leo M.C.
      • et al.
      Performance of ACMG-AMP variant-interpretation guidelines among nine laboratories in the clinical sequencing exploratory research consortium.
      in that researchers have subjective understandings of the ACMG/AMP guidelines and have varying abilities to access information from the literature and databases.
      InterVar was first developed to facilitate variant interpretation for the full spectrum of diseases, but it does not perform well for specific diseases.
      • Li Q.
      • Wang K.
      InterVar: clinical interpretation of genetic variants by the 2015 ACMG-AMP guidelines.
      Thus, several pilot studies have been conducted to develop disease-specific interpretation tools, such as Variant Interpretation Platform for genetic Hearing Loss,
      • Peng J.
      • Xiang J.
      • Jin X.
      • et al.
      VIP-HL: semi-automated ACMG/AMP variant interpretation platform for genetic hearing loss.
      Variant Interpretation for Cancer,
      • He M.M.
      • Li Q.
      • Yan M.
      • et al.
      Variant Interpretation for Cancer (VIC): a computational tool for assessing clinical impacts of somatic variants.
      neXtProt for BRCA1,
      • Cusin I.
      • Teixeira D.
      • Zahn-Zabal M.
      • et al.
      A new bioinformatics tool to help assess the significance of BRCA1 variants.
      CardioClassifier
      • Whiffin N.
      • Walsh R.
      • Govind R.
      • et al.
      CardioClassifier: disease- and gene-specific computational decision support for clinical genome interpretation.
      and Cardio Variant Interpreter
      • Nicora G.
      • Limongelli I.
      • Gambelli P.
      • et al.
      CardioVAI: an automatic implementation of ACMG-AMP variant interpretation guidelines in the diagnosis of cardiovascular diseases.
      for cardiovascular diseases; however, these semiautomated tools still interpret variants largely dependent on user-curated data. Moreover, public databases such as ClinVar,
      • Landrum M.J.
      • Chitipiralla S.
      • Brown G.R.
      • et al.
      ClinVar: improvements to accessing data.
      Human Gene Mutation Database (HGMD),
      • Stenson P.D.
      • Mort M.
      • Ball E.V.
      • et al.
      The Human Gene Mutation Database: towards a comprehensive repository of inherited mutation data for medical research, genetic diagnosis and next-generation sequencing studies.
      and Universal Mutation Database (UMD)
      • Collod-Béroud G.
      • Le Bourdelles S.
      • Ades L.
      • et al.
      Update of the UMD-FBN1 mutation database and creation of an FBN1 polymorphism database.
      ,
      • Frederic M.Y.
      • Hamroun D.
      • Faivre L.
      • et al.
      A new locus-specific database (LSDB) for mutations in the TGFBR2 gene: UMD-TGFBR2.
      are widely used for current tools. However, these databases have inevitable limitations, such as nonstandard sequencing variant nomenclature, inconsistent variant classifications generated by different criteria, and a lack of key information for interpretation, such as family- and case-level data and functional assay results.
      Thoracic aortic aneurysm (TAA) is an insidious and life-threatening vascular disease that is difficult to detect and diagnose before catastrophic complications, aortic dissection or rupture.
      • Kuzmik G.A.
      • Sang A.X.
      • Elefteriades J.A.
      Natural history of thoracic aortic aneurysms.
      Approximately 95% of the patients with TAA are asymptomatic.
      • Elefteriades J.A.
      • Sang A.
      • Kuzmik G.
      • Hornick M.
      Guilt by association: paradigm for detecting a silent killer (thoracic aortic aneurysm).
      When dissection or rupture occurs abruptly, approximately 22% of patients die before reaching the hospital.
      • Cheung K.
      • Boodhwani M.
      • Chan K.L.
      • Beauchesne L.
      • Dick A.
      • Coutinho T.
      Thoracic aortic aneurysm growth: role of sex and aneurysm etiology.
      However, if TAA is promptly recognized and managed surgically before dissection, patients have excellent survival rates with limited complications.
      • Zafar M.A.
      • Farkas E.A.
      • Javier A.
      • Anderson M.
      • Gilani O.
      • Elefteriades J.A.
      Are thromboembolic and bleeding complications a drawback for composite aortic root replacement?.
      Therefore, early detection and diagnosis are crucial for disease monitoring and surgical management to prevent devastating events. Because it is difficult to detect TAA based on symptoms alone, genetic testing is used to facilitate the establishment of a definitive diagnosis in patients and the identification of at-risk relatives.
      • Milewicz D.M.
      • Guo D.
      • Hostetler E.
      • Marin I.
      • Pinard A.C.
      • Cecchi A.C.
      Update on the genetic risk for thoracic aortic aneurysms and acute aortic dissections: implications for clinical care.
      To improve the reliability and efficiency of genetic testing in heritable TAA and dissection (HTAAD), we developed HTAADVar, an unbiased and fully automated system for variant interpretation, comprising a self-built variant database and interpretation programs. HTAADVar shows multifold higher sensitivity and specificity than other tools and produces a diagnostic yield in a real sequencing cohort that is highly comparable to the manual interpretation and previous reports. To facilitate HTAADVar use, we built a web server with a friendly interactive interface and powerful browse, search, and variant interpretation functions. Stand-alone programs are also provided for customized interpretations and generalization of the HTAADVar framework to other diseases.

      Materials and Methods

      Data collection

      We first selected 18 HTAAD genes as our targeted genes according to the clinical validity of genes for HTAAD evaluated by the expert panel of the Aortopathy Working Group (Supplemental Methods). Then, we retrieved 1564 publications from PubMed using the query statement described in the Supplemental Methods. For completeness, 617 publications collected from ClinVar and/or HGMD but missed in the PubMed query were also retrieved. We excluded all publications reporting only large sequence variations (>50 basepairs) or performing analyses based on the variants in public databases. Moreover, we also integrated the variants in targeted genes identified in 790 Chinese probands by our laboratory (Supplemental Methods). Then, comprehensive metadata were manually collected and double-checked independently by experts. Functional annotations for genes were integrated following a previous procedure.
      • Zhou W.Z.
      • Li W.
      • Shen H.
      • et al.
      CHDbase: A comprehensive knowledgebase for congenital heart disease-related genes and clinical manifestations.

      Variant database comparison

      To ensure a fair comparison, we compared our variant database with ClinVar (v2021-11-08), HGMD (v2021.2), and UMD (v2021-12-06) based on the single-nucleotide variants and insertions-deletions (≤50 basepairs) from studies published by October 20, 2020. We also excluded studies based on database variants.

      Rule optimization and implementation

      There are 28 criteria under the ACMG/AMP framework,
      • Richards S.
      • Aziz N.
      • Bale S.
      • et al.
      Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology.
      and the ClinGen Sequence Variant Interpretation working group suggested removing PP5 and BP6.
      • Biesecker L.G.
      • Harrison S.M.
      ClinGen Sequence Variant Interpretation Working Group. The ACMG/AMP reputable source criteria for the interpretation of sequence variants.
      PM3 and BP1 were deemed not applicable because our targeted genes cause HTAAD in a dominant mode, and truncating variants are not the only pathogenicity mechanism of HTAAD. Therefore, we established HTAAD-specific and gene-specific rules and implementation methods for 24 criteria of the ACMG/AMP guidelines (Supplemental Methods). The HTAADVar implementation process is also described in the Supplemental Methods.

      Benchmarking and comparative analysis

      To evaluate the performance of HTAADVar, we used pathogenic or likely pathogenic (P/LP) and benign or likely benign (B/LB) variants by multiple submitters with no conflicts in ClinVar (v2021-11-08) as a benchmark data set.
      We performed automated variant interpretation using InterVar and CardioClassifier using default parameters without manual adjustment. To ensure interpretation under the same framework, we recalibrated the InterVar result after removing PP5 and BP6. We compared the final classifications and activated rules using different tools.

      Web server and stand-alone programs

      We constructed a MongoDB database to store and manage the collected data. For interpretation, we provided a web server and Perl programs to support users with different requirements. The web server has a user-friendly interface developed using Java, HTML, CSS, and PHP.

      Results

      HTAAD variant database

      A professional database dedicated to variant interpretation was first developed to create a fully automated system. We constructed an HTAAD variant database for subsequent interpretation. For 18 HTAAD genes, we searched PubMed, ClinVar, and HGMD to retrieve related studies (Materials and Methods). In total, 2181 research articles were retrieved. After scrutinizing the abstracts to exclude irrelevant studies, 697 articles published between January 1989 and October 2020 were retained to be manually curated. We also integrated variants identified in 790 in-house patients to improve the interpretation power (Supplemental Methods; Supplemental Table 1). Finally, 4373 unique variants were collected in the database (Figure 1).
      Figure thumbnail gr1
      Figure 1The framework overview of HTAADVar. HTAADVar consists of a self-built variant database that integrates manually curated literature and in-house data and interpretation programs that implement automated interpretation based on all ACMG criteria to obtain a final classification under the 5-tier system. In the interpretation process, the criteria above the dotted line are scored on the basis of the variant database and in-house controls, whereas those under the dotted line are scored based on the external data sets. ACMG, American College of Medical Genetics and Genomics; AAF, alternative allele frequency; gnomAD, Genome Aggregation Database.
      For each study, comprehensive metadata were extracted and proofread manually, as shown in Supplemental Tables 2 to 8. Particularly, we integrated the key information for variant interpretation, such as variant observations in unrelated patients, the heterozygosity and parental origin of variants, including mosaic and de novo status, patient phenotypes, family history, familial segregation as scored following Jarvik and Browning’s guidelines,
      • Jarvik G.P.
      • Browning B.L.
      Consideration of cosegregation in the pathogenicity classification of genomic variants.
      variant effect on gene function to which specific PS3 strength level was assigned according to experimental model and phenotypic expressions (Supplemental Methods), and variant effect on splicing to which specific PVS1 strength level was assigned according to ClinGen PVS1 guidelines.
      • Abou Tayoun A.N.
      • Pesaran T.
      • DiStefano M.T.
      • et al.
      Recommendations for interpreting the loss of function PVS1 ACMG/AMP variant criterion.
      To avoid repeatedly recording the same variant because of the variety of naming conventions, variants were standardized into a format with the chromosome and position (in GRCh37/hg19), reference allele, and alternative allele and then annotated at transcript-based DNA and protein levels following the nomenclature recommendations of the Human Genome Variation Society by Ensembl Variant Effect Predictor.
      • McLaren W.
      • Gil L.
      • Hunt S.E.
      • et al.
      The Ensembl variant effect predictor.
      Compared with ClinVar, HGMD, and UMD, as shown in Table 1, our variant database compiles a larger set of published variants in HTAAD genes. UMD contains the fewest variants because it has been adapted to only 7 targeted genes. In contrast to the user submission of some databases, we manually curated comprehensive information from the literature and double-checked it to ensure the high quality of the data. Notably, the standardized information critical to variant interpretation is available only in our variant database. Furthermore, more comprehensive annotations for variants, such as the alternative allele frequency with sequencing depth in Genome Aggregation Database (gnomAD),
      • Karczewski K.J.
      • Francioli L.C.
      • Tiao G.
      • et al.
      The mutational constraint spectrum quantified from variation in 141,456 humans.
      predictive algorithms for variant effects, protein domain, and pathogenicity classifications generated by our unified automated interpretation programs, were also integrated into our database. Overall, the HTAAD variant database can provide a reliable resource to support the in-depth interpretation of HTAAD variants.
      Table 1HTAAD variant database in comparison with existing databases
      CategoryHTAAD Variant Database
      Data from studies published by October 20, 2020.
      ClinVar
      Data from studies published by October 20, 2020.
      HGMD
      Data from studies published by October 20, 2020.
      UMD
      Data sourceManually curatedUser submittedManually curatedManually curated, user submitted
      HTAAD genesAll targeted genesAll targeted genesAll targeted genesFBN1, TGFBR1, TGFBR2, SMAD3, ACTA2, MYH11, and MYLK
      No. of HTAAD variants4077214838512072
      Collected informationStudy, variant, functional result, sample and family informationVariant informationVariant informationVariant and sample information
      Information specific to the interpretationDe novo and family history for PS2/PM6, functional evidence for PS3/BS3, number of unrelated probands with the variant for PS4, familial segregation for PP1/BS4, phenotype specificity for PP4Unstructured evidence may be described by some submittersNoneDe novo information for PS2/PM6, but without a family history
      Pathogenicity classificationUnified interpretation by HTAADVarUser submitted according to own criteriaConclusions of the original publicationsPathogenicity predicted by UMD-predictor
      No. of variant annotation items2213412
      HGMD, Human Gene Mutation Database; HTAAD, heritable thoracic aortic aneurysm and dissection; UMD, Universal Mutation Database.
      a Data from studies published by October 20, 2020.

      Substantial improvement in the performance of HTAADVar

      Given that the ACMG/AMP framework can fully exert its power in a disease- and gene-specific manner, optimizing rules for specific disease–gene pairs is particularly important to achieve accurate classification. Our panel consists of experts in genetics, molecular biology, clinical diagnosis, and treatment for HTAAD. The experts in this panel are members of the Precision Medicine Group, established in 2019 and affiliated with the National Society of Vascular Surgery. In the same year, our experts issued the “Chinese expert consensus on the genetic testing and clinical management of heritable thoracic aortic aneurysm/dissection” to promote HTAAD molecular diagnosis standardization in China.
      National Society of Vascular Surgery
      Chinese expert consensus on the genetic testing and clinical management of heritable thoracic aortic aneurysm/dissection.
      Accordingly, we refined the ACMG/AMP rules by combining HTAAD- and gene-specific knowledge with released recommendations from expert panels such as ClinGen SVI for PVS1, PS2, PM5, PM6, PP1, PP5, and BP6 (Supplemental Methods; Supplemental Tables 9 and 10) and developed a suite of methods to implement the rules, leveraging the HTAAD variant database and public data sets (Figure 1; Supplemental Figure 1). Using our automated interpretation system, 4373 variants in our database were assessed, and their classifications are summarized in Supplemental Table 11.
      To evaluate the performance of our system, we compared HTAADVar with InterVar and CardioClassifier, which are freely accessible and commonly used in ACMG/AMP-based interpretation for HTAAD variants. Among our targeted genes, only FBN1 can also be analyzed by CardioClassifier; thus, we selected 611 nonconflicting variants in FBN1 from ClinVar as a benchmark set, consisting of 299 P/LP and 312 B/LB variants. For a fair comparison, we recalibrated InterVar classifications after removing the reputable source criteria (PP5 and BP6) because of their questionable utility
      • Biesecker L.G.
      • Harrison S.M.
      ClinGen Sequence Variant Interpretation Working Group. The ACMG/AMP reputable source criteria for the interpretation of sequence variants.
      rendering them inapplicable for HTAADVar and CardioClassifier. Based on an automated interpretation step with default parameters, InterVar and CardioClassifier classified 109 (36.45%) and 113 (37.79%) P/LP variants, respectively, as having the same pathogenicity, whereas HTAADVar reproduced the classifications for 277 P/LP variants with a sensitivity of 92.64%. InterVar and CardioClassifier classified 52 (16.67%) and 17 (5.45%) B/LB variants, respectively, as having the same pathogenicity, whereas HTAADVar reproduced the classifications for 221 B/LB variants with a specificity of 70.83% (Figure 2A). Notably, all misclassified variants were interpreted as variant of uncertain significance (VUS) by the tools.
      Figure thumbnail gr2
      Figure 2Comparison of HTAADVar with existing tools. A. Performance comparison of HTAADVar with InterVar and CardioClassifier on 299 P/LP and 312 B/LB variants in FBN1 from ClinVar. B. Frequency of rules activated by HTAADVar, InterVar, and CardioClassifier for 299 P/LP (upper panel) and 312 B/LB (lower panel) in FBN1 from ClinVar. Only rules activated at least once by one of the tools are shown. The x-axis labels indicate the activated rules at any strength level. The number in the heatmap cell indicates the proportion of variants applied to the rule. Note that in B/LB group, the activated rules of CardioClassifier are only shown for the variants that were upgraded to VUS by CardioClassifier because this tool did not provide the activated rules for the variant classified as B/LB. B/LB, benign/likely benign; P/LP, pathogenic/likely pathogenic; VUS, variant of uncertain significance.
      The high performance of HTAADVar is not limited to FBN1. In comparison with InterVar, we further extended the benchmark set to 1764 nonconflicting variants in ClinVar across all targeted genes. Based on this larger set, HTAADVar still reached 2 to 3 times the sensitivity (HTAADVar: 85.10% vs InterVar: 35.10%) and specificity (HTAADVar: 76.41% vs InterVar: 28.12%) compared with InterVar (Supplemental Table 12), although, the InterVar classifications might be slightly biased toward those of ClinVar because InterVar refers to ClinVar information during interpretation.
      • Li Q.
      • Wang K.
      InterVar: clinical interpretation of genetic variants by the 2015 ACMG-AMP guidelines.
      As shown in Figure 2B, Supplemental Figure 2, and Supplemental Table 13, the evidence, such as de novo observations, functional studies, familial segregation, and phenotype data for PS2/PM6, PS3, PP1, and PP4, automatically scored by HTAADVar but absent in other tools contribute substantially to the interpretation of P/LP variants. Similarly, PS4 and PM5 were also often triggered by HTAADVar but not activated by the other tools. Benefiting from the self-built database, we could count the variant observations in unrelated patients to apply PS4 following the ClinGen General Sequence Variant Curation standard operating procedure.

      The Clinical Genome Resource ClinGen Variant Curation SOP Committee. ClinGen general sequence variant curation process standard operating procedure. Version 2.0. Clinical Genome Resource. Published January 2021. Accessed March 17, 2021. https://clinicalgenome.org/site/assets/files/5933/variant_curation_sop_2_0_jan_2021.pdf

      This is more suitable than the odds ratio that is used by InterVar because rare variants are typically difficult to identify with genome-wide significance because of their low allele frequencies. However, CardioClassifier does not implement PS4 for our targeted diseases. For PM5, InterVar and CardioClassifier only refer to variants without conflicting annotations in ClinVar to check previous pathogenic changes; whereas in HTAADVar, our variant database provides a more reliable and complete reference because all variant classifications were generated using our unified interpretation process and the reasoning behind the classifications can be traced. Finally, because disease and gene specificity are not considered, InterVar inappropriately inactivated PM1 and activated BP1, further leading to the misclassification of P/LP variants as VUS.
      For B/LB variants, we mainly compared the activated rules between HTAADVar and InterVar because 86.22% of variants could not be analyzed using CardioClassifier. As shown in Figure 2B, Supplemental Figure 2, and Supplemental Table 13, the misclassification of B/LB variants as VUS is largely owing to PM2 activation. Although HTAADVar adopts a more stringent rule for PM2 than InterVar, requiring not only the variant absent in gnomAD but also a sufficient read depth (≥10×) for an accurate call, PM2 activation still reduced the specificity. Allele frequency evidence (BA1, BS1, and BS2) also contributes to the discrepancy. For BA1 and BS1, HTAADVar used lower but more suitable thresholds for HTAAD (>1% for BA1 and >0.02% for BS1) compared with InterVar (>5% for BA1 and >1% for BS1). For BS2, InterVar applied it for the variant in any targeted gene if it was observed in the 1000 Genomes Project, whereas HTAADVar only applied it for syndromic TAAD and genes with full penetrance on the basis of an in-house control cohort of 460 adults confirmed to be free of cardiovascular diseases. However, given that in 312 B/LB variants in FBN1, only 11 were applied to BS2 by HTAADVar, the control sample size needs to be further expanded.

      Improved efficiency with comparable diagnostic yield in comparison to manual interpretation

      We compared the classifications of 547 variants from 790 in-house probands obtained by HTAADVar and manual interpretation to further assess the clinical utility of HTAADVar. Of the 300 P/LP variants interpreted using HTAADVar, the pathogenicity was the same as the manual interpretation in 99.33% of variants. Among the 55 B/LB variants, the consistency was lower (58.18%), largely because of the different strategies used; manual interpretation applied more stringent BP4 rules requiring all 4 categories of predictions, including the conservation, function, meta, and splicing site predictions, to support the benign classification, whereas HTAADVar assigns a variant as benign only if the predictions from most of these categories are supportive. To evaluate which rule is more appropriate, we used the aforementioned 1348 B/LB benchmark variants for testing. Using manual interpretation rules resulted in the application of only 38 (2.82%) B/LB variants to BP4, which is far below that of HTAADVar (80.12%), indicating that the BP4 rule of manual interpretation may be too stringent. Finally, of the 192 VUS variants in HTAADVar, 20 were manually upgraded to LP, and 2 were downgraded to LB. This change is mainly because (1) HTAADVar adopted the ClinGen SVI recommendations to refine the ACMG/AMP criteria. However, some of these recommendations are generally too complex to follow manually, such as the loss-of-function PVS1 criterion.
      • Abou Tayoun A.N.
      • Pesaran T.
      • DiStefano M.T.
      • et al.
      Recommendations for interpreting the loss of function PVS1 ACMG/AMP variant criterion.
      In addition, some of them were released recently, such as the de novo PS2 and PM6 criteria,

      Clinical Genome Resource. ClinGen Sequence Variant Interpretation Recommendation for de novo Criteria (PS2/PM6). Version 1.1. Clinical Genome Resource. Published March 18, 2018. Updated May 5, 2021. Accessed August 20, 2021. https://clinicalgenome.org/docs/ps2-pm6-recommendation-for-de-novo-ps2-and-pm6-acmg-amp-criteria-version-1.0

      and were typically not used for the manual interpretation. (2) Manual interpretation primarily uses ClinVar and UMD to quickly retrieve the necessary evidence, including unpublished user-submitted data. HTAADVar only reviewed and included published and our in-house data. Although, the HTAADVar strategy better ensures data quality, some evidence are missed. Therefore, we will continually update HTAADVar with data from our laboratory and the user community to alleviate this problem.
      Among 790 in-house patients, P/LP variants were associated with 332 by HTAADVar, resulting in an overall molecular diagnostic yield of 42.03%. This yield is highly consistent with that obtained by manual interpretation (44.56%) and comparable with previously reported data from studies evaluating a similar patient population (30%-43%).
      • Yang H.
      • Luo M.
      • Fu Y.
      • et al.
      Genetic testing of 248 Chinese aortopathy patients using a panel assay.
      • Li J.
      • Yang L.
      • Diao Y.
      • et al.
      Genetic testing and clinical relevance of patients with thoracic aortic aneurysm and dissection in northwestern China.
      • Duan D.M.
      • Chiu H.H.
      • Chen P.L.
      • et al.
      Clinical manifestations and genetic characteristics in the Taiwan thoracic aortic aneurysm and dissection cohort – a prospective cohort study.
      Notably, the diagnostic yield among patients with suspected or confirmed Marfan, Loeys-Dietz, or vascular Ehlers-Danlos syndrome (256/334, 76.65%) was much higher than that for patients with nonsyndromic TAAD (76/456, 16.67%), which is also consistent with previous reports.
      • Yang H.
      • Luo M.
      • Fu Y.
      • et al.
      Genetic testing of 248 Chinese aortopathy patients using a panel assay.
      ,
      • Arnaud P.
      • Hanna N.
      • Benarroch L.
      • et al.
      Genetic diversity and pathogenic variants as possible predictors of severity in a French sample of nonsyndromic heritable thoracic aortic aneurysms and dissections (nshTAAD).
      ,
      • Li Y.
      • Kong Y.
      • Duan W.
      • et al.
      Evaluating the monogenic contribution and genotype-phenotype correlation in patients with isolated thoracic aortic aneurysm.
      Finally, we assessed the HTAADVar speed on a machine with 16 GB of memory and an Intel(R) Xeon(R) Gold 6151 CPU (4 cores). For a variant in the database, users can receive an immediate response (<1 second) regardless of whether additional evidence is submitted. A longer time (∼10 seconds) will be needed for a variant that is not in the database. Thus, the run time for a batch task varies with the preinterpreted variant proportion. For example, to interpret 100 variants including 36 that are novel, HTAADVar will take approximately 13 seconds. If all 100 variants are novel, it will take approximately 35 seconds. Compared with manual interpretation, HTAADVar substantially improved the efficiency of the interpretation on the premise of the reliability of classification.

      Comparative analysis with ClinVar

      ClinVar is a freely accessible archive of genetic variants and their clinical significance to disease and is thus widely used in variant interpretation.
      • Landrum M.J.
      • Chitipiralla S.
      • Brown G.R.
      • et al.
      ClinVar: improvements to accessing data.
      To assess the concordance rate and investigate the reasons for discordance, we compared variant classifications made by HTAADVar with those in ClinVar based on 1977 shared variants. For 1160 P/LP variants in HTAADVar, ClinVar also annotated 1096 (94.48%) as P/LP and none as B/LB. For 93 B/LB variants in HTAADVar, 65 (69.89%) were also annotated as B/LB in ClinVar, and none were annotated as P/LP (Figure 3). This shows that ClinVar is highly consistent with HTAADVar in P/LP interpretation but lower in B/LB interpretation. For the variants annotated as B/LB in HTAADVar but VUS in ClinVar, most of them were applied to BP5 by HTAADVar because the allelic and alternative locus data available in our database enable HTAADVar to automatically assess BP5, which is difficult in manual assessment.
      Figure thumbnail gr3
      Figure 3Comparison of HTAADVar with ClinVar. Upset diagram shows the difference in classifications between HTAADVar and ClinVar for 1977 their shared variants. B/LB, benign/likely benign; P/LP, pathogenic/likely pathogenic; VUS, variant of uncertain significance.
      Half of the 724 VUS variants classified by HTAADVar (369, 50.97%) were upgraded to P/LP and 21 (2.90%) were downgraded to B/LB in ClinVar (Figure 3). The possible reasons for the discrepancy are as follows: (1) ClinVar submitters obtained the variant classifications based on other criteria, such as Blueprint, Baylor, and Laboratory for Molecular Medicine genetics variant classification, or with no explicit criteria. (2) If ACMG/AMP guidelines were adopted, submitters may not have adjusted the strength level for specific criteria following the expert panel recommendations. For example, in ClinVar, PVS1 may be incorrectly applied for genes in which the loss-of-function is not a well-established disease mechanism, such as MFAP5 and TGFB3, or submitters did not determine PVS1’s appropriate strength level following the ClinGen SVI recommendations,
      • Abou Tayoun A.N.
      • Pesaran T.
      • DiStefano M.T.
      • et al.
      Recommendations for interpreting the loss of function PVS1 ACMG/AMP variant criterion.
      such as NM_000138.4:c.164+1G>A of FBN1. PP1 may be activated for variants that segregate with disease in a family, but the segregation score does not meet the minimum requirement of PP1 quantitative criteria,
      • Jarvik G.P.
      • Browning B.L.
      Consideration of cosegregation in the pathogenicity classification of genomic variants.
      such as NM_003238.6:c.1013C>A (p.Pro338His) of TGFB2, with a segregation score of 1/4. (3) The thresholds for allele frequency evidence (BA1, BS1, and PM2) and the annotations of gene functional regions used by the submitters might differ from those of HTAADVar. In our interpretation, PS4 and PP1 must be applied together with PM2. If the variant is observed in gnomAD even at a low frequency, PS4 and PP1 cannot be applied. ClinVar submitters may not use such a stringent threshold; thus, rare variants may be applied to PM2, PS4, and PP1 and upgraded to P/LP, such as NM_000138.4:c.7754T>C (p.Ile2585Thr) of FBN1. Another example is NM_000138.4:c.5513G>A (p.Gly1838Asp), which is outside the well-established functional domains of FBN1. The submitter still applied PM1 because several variants in its nearby residues have been associated with Marfan syndrome. (4) Finally, submitters may include additional evidence from their unpublished data during assessment as mentioned above. For example, NM_001613.4:c.772C>T (p.Arg258Cys) of ACTA2 has been reported as de novo in a TAAD family by Guo et al,
      • Guo D.C.
      • Papke C.L.
      • Tran-Fadulu V.
      • et al.
      Mutations in smooth muscle alpha-actin (ACTA2) cause coronary artery disease, stroke, and Moyamoya disease, along with thoracic aortic disease.
      which HTAADVar also collects. Another de novo event of this variant in the submitter’s laboratory further supports its pathogenic role. Overall, when using HTAADVar or ClinVar to interpret variants, users should pay attention to interpretation and classification differences and their possible reasons.

      Web interface and customized running using stand-alone programs

      HTAADVar provides a simple-to-use web server (http://htaadvar.fwgenetics.org) to facilitate access to our variant database and interpret any variant of interest. In this article, the powerful browse, search, and interpretation functions of the HTAADVar webserver are introduced (Figure 4).
      Figure thumbnail gr4
      Figure 4Web interface of HTAADVar. The Gene page, Variant page, and Source page display the data of the variant database at the gene, variant, and study levels, respectively. For each study, detailed information can be accessed by clicking on the link on the Source page. To facilitate the users in querying data of interest, the basic search mode is provided on the Home page and the top navigation bar, and the advanced search mode is embedded into the Variant page and Source page. The Interpretation page provides an interactive interface for users to interpret a single variant or a batch of variants. The interpretation report includes the final classification and the reasoning behind it with all supporting evidence. ACMG, American College of Medical Genetics and Genomics; HTAAD, heritable thoracic aortic aneurysm and dissection.
      Users can browse the gene, variant, and study levels as follows: Gene page first displays the overview, which includes gene–HTAAD validity as evaluated by ClinGen, the first publication and total number of publications reporting gene–HTAAD associations, the total number of variants in the gene reported in HTAAD, and the gene function summary. Variant distributions grouped by transcript, exon/domain, and pathogenicity are sequentially shown. Furthermore, users can click on the “Gene Annotation” label in the top left to view comprehensive gene annotations. On the Variant page, users can browse the variant information flexibly. Notably, HTAADVar interprets the variants at the transcript level. The “IC” column indicates the consistency of classifications between different transcripts. Users can click on the “+” sign to view the classifications on hidden transcripts. Users can scrutinize the reasoning behind the interpretation and all supporting data by clicking on the evidence codes in the “Evidence” column. The sample information will be shown when users click on the “Probands” column. The Source page summarizes literature information, including the PubMed identifier, title, authors, journal, publication date, study type, and count of variants reported. By clicking on the “Study Type” column, users can view the detailed information for “Sequencing Study” and “Functional Study.”
      In the web server, 2 search modes are available. The basic search box on the home page and the top navigation bar enables users to search for genes, variants, and literature of interest in the variant database. The gene symbol and aliases, variants in the format of “chromosome: start position-end position: reference allele > alternative allele,” Human Genome Variation Society expression at the complementary DNA and protein levels, and PubMed identifier can be recognized. Fuzzy matching is allowed for variant and literature queries. On the Variant and Source pages, search boxes under the table header enable users to filter data using an advanced mode.
      In addition to searching for preinterpreted variants in the database, users can interpret any variant of interest. The web server provides a friendly interactive interface to enable users to submit data easily for single-variant interpretation. The interpretation of multiple variants supports variant submission in variant call format. The results are rendered as HTML pages and/or Excel files.
      In additional, HTAADVar can be run in Perl, which is more suitable for large tasks. In the stand-alone version, users can refine the thresholds of PVS1, PS2, PS4, PM2, PM5, PM6, BS1, BA1, and computational predictors in the configure file and update the variant classifications in the database. Users can interpret any variant on the basis the updated database using newly defined thresholds.

      Discussion

      Genetic testing has been widely used in clinical practice, and variant interpretation is a crucial but challenging step. In this study, we developed an unbiased interpretation system that can automatically implement all applicable ACMG/AMP guideline criteria. HTAADVar was demonstrated to be superior in accuracy and efficiency, indicating that it can be effectively applied for molecular diagnostics.
      Compared with existing tools, HTAADVar has the following advantages: (1) It is the first fully automated system of variant interpretation specific to HTAAD that does not require users to manually curate the items of evidence from massive publications in variant annotations. Metadata from publications, in-house sequencing data, and public data sets have been integrated and organized to support a comprehensive annotation in HTAADVar. (2) Based on the 2015 ACMG/AMP guidelines, the refined rules combining disease- and gene-specific knowledge with expert panel recommendations for PVS1, PS2, PM5, PM6, PP1, PP5, and BP6 substantially increase the interpretation power. (3) HTAADVar allows users to interpret variants with customized criteria, which is useful in such a rapidly evolving field of variant interpretation. (4) Finally, HTAADVar provides a user-friendly web interface, and users can interactively interpret the variants or search for metadata for variants of interest in our expert-reviewed variant database.
      However, HTAADVar has the following limitations: (1) HTAADVar only supports automated interpretation for variants in HTAAD genes. However, its framework could be generalized to other inherited diseases and genes. Users would need to build a variant database and define the interpretation parameters specific to their targeted diseases and genes of interest. Then, the interpretation program can be easily applied. (2) The current rules and thresholds were determined according to previous recommendations and our expert panel’s knowledge, but a consensus has not yet been reached in the field. Users can update the thresholds in the configuration file to interpret variants according to their preferences. Alternatively, the transparent result report enables users to manually adjust the classification.
      We will continuously improve HTAADVar as genetic studies on HTAAD progress. In the future, we will update HTAADVar regularly by curating new publications, incorporating new in-house data, including HTAADVar user community data, updating the list of disease-causing genes, and adjusting rules, thresholds, or methodologies for interpretation according to new recommendations from expert panels. HTAADVar is a valuable system that can greatly facilitate clinician and researcher assessment of variant pathogenicity. Its framework and methodologies could substantially improve standardization for the molecular diagnosis of genetic diseases.

      Data Availability

      All data collected for the HTAAD variant database in this study are available on the web server without login requirements (http://htaadvar.fwgenetics.org/).

      Acknowledgments

      We acknowledge Yandong Cao, Kaituo Mi, Kunpeng He, Yulei Liu, and Wuqiang Zhang at Anngeen Technology Co Ltd for their support and contribution to the web server building and Ge Gao at Peking University for assistance with database comparison. This work was supported by the Chinese Academy of Medical Sciences ( CAMS ) Innovation Fund for Medical Sciences (CIFMS) (2021-I2M-1-008) and the Young Scientists Fund of the National Natural Science Foundation of China (31801103).

      Author Information

      Conceptualization: Z.Z., W.-Z.Z.; Supervision: W.-Z.Z., H.Y.; Data Curation: W.-Z.Z., H.Y., G.Z., H.S., Q.Z., Q.C.; Formal Analysis: W.-Z.Z., Y.Z., G.Z.; Funding Acquisition: Z.Z., W.-Z.Z.; Investigation: W.-Z.Z., Y.Z., H.Y., G.Z.; Methodology: W.-Z.Z., Y.Z., H.Y., M.L., C.S.; Visualization: W.-Z.Z., W.L.; Writing-original draft: W.-Z.Z., H.Y.; Writing-review and editing: W.-Z.Z., Z.Z.

      Ethics Declaration

      The study was approved by the Ethics Committee of Fuwai Hospital (approval number: 2017-877). All patients and their relatives in the in-house cohort signed informed consent forms.

      Conflict of Interest

      The authors declare no conflict of interest.

      References

        • Amendola L.M.
        • Jarvik G.P.
        • Leo M.C.
        • et al.
        Performance of ACMG-AMP variant-interpretation guidelines among nine laboratories in the clinical sequencing exploratory research consortium.
        Am J Hum Genet. 2016; 98 (Published correction appears in Am J Hum Genet. 2016;99(1):247): 1067-1076
        • Richards S.
        • Aziz N.
        • Bale S.
        • et al.
        Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology.
        Genet Med. 2015; 17: 405-424https://doi.org/10.1038/gim.2015.30
        • Li Q.
        • Wang K.
        InterVar: clinical interpretation of genetic variants by the 2015 ACMG-AMP guidelines.
        Am J Hum Genet. 2017; 100: 267-280https://doi.org/10.1016/j.ajhg.2017.01.004
        • Peng J.
        • Xiang J.
        • Jin X.
        • et al.
        VIP-HL: semi-automated ACMG/AMP variant interpretation platform for genetic hearing loss.
        Hum Mutat. 2021; 42: 1567-1575https://doi.org/10.1002/humu.24277
        • He M.M.
        • Li Q.
        • Yan M.
        • et al.
        Variant Interpretation for Cancer (VIC): a computational tool for assessing clinical impacts of somatic variants.
        Genome Med. 2019; 11: 53https://doi.org/10.1186/s13073-019-0664-4
        • Cusin I.
        • Teixeira D.
        • Zahn-Zabal M.
        • et al.
        A new bioinformatics tool to help assess the significance of BRCA1 variants.
        Hum Genomics. 2018; 12: 36https://doi.org/10.1186/s40246-018-0168-0
        • Whiffin N.
        • Walsh R.
        • Govind R.
        • et al.
        CardioClassifier: disease- and gene-specific computational decision support for clinical genome interpretation.
        Genet Med. 2018; 20: 1246-1254https://doi.org/10.1038/gim.2017.258
        • Nicora G.
        • Limongelli I.
        • Gambelli P.
        • et al.
        CardioVAI: an automatic implementation of ACMG-AMP variant interpretation guidelines in the diagnosis of cardiovascular diseases.
        Hum Mutat. 2018; 39: 1835-1846https://doi.org/10.1002/humu.23665
        • Landrum M.J.
        • Chitipiralla S.
        • Brown G.R.
        • et al.
        ClinVar: improvements to accessing data.
        Nucleic Acids Res. 2020; 48: D835-D844https://doi.org/10.1093/nar/gkz972
        • Stenson P.D.
        • Mort M.
        • Ball E.V.
        • et al.
        The Human Gene Mutation Database: towards a comprehensive repository of inherited mutation data for medical research, genetic diagnosis and next-generation sequencing studies.
        Hum Genet. 2017; 136: 665-677https://doi.org/10.1007/s00439-017-1779-6
        • Collod-Béroud G.
        • Le Bourdelles S.
        • Ades L.
        • et al.
        Update of the UMD-FBN1 mutation database and creation of an FBN1 polymorphism database.
        Hum Mutat. 2003; 22: 199-208https://doi.org/10.1002/humu.10249
        • Frederic M.Y.
        • Hamroun D.
        • Faivre L.
        • et al.
        A new locus-specific database (LSDB) for mutations in the TGFBR2 gene: UMD-TGFBR2.
        Hum Mutat. 2008; 29: 33-38https://doi.org/10.1002/humu.20602
        • Kuzmik G.A.
        • Sang A.X.
        • Elefteriades J.A.
        Natural history of thoracic aortic aneurysms.
        J Vasc Surg. 2012; 56: 565-571https://doi.org/10.1016/j.jvs.2012.04.053
        • Elefteriades J.A.
        • Sang A.
        • Kuzmik G.
        • Hornick M.
        Guilt by association: paradigm for detecting a silent killer (thoracic aortic aneurysm).
        Open Heart. 2015; 2e000169https://doi.org/10.1136/openhrt-2014-000169
        • Cheung K.
        • Boodhwani M.
        • Chan K.L.
        • Beauchesne L.
        • Dick A.
        • Coutinho T.
        Thoracic aortic aneurysm growth: role of sex and aneurysm etiology.
        J Am Heart Assoc. 2017; 6e003792https://doi.org/10.1161/JAHA.116.003792
        • Zafar M.A.
        • Farkas E.A.
        • Javier A.
        • Anderson M.
        • Gilani O.
        • Elefteriades J.A.
        Are thromboembolic and bleeding complications a drawback for composite aortic root replacement?.
        Ann Thorac Surg. 2012; 94: 737-743https://doi.org/10.1016/j.athoracsur.2012.04.007
        • Milewicz D.M.
        • Guo D.
        • Hostetler E.
        • Marin I.
        • Pinard A.C.
        • Cecchi A.C.
        Update on the genetic risk for thoracic aortic aneurysms and acute aortic dissections: implications for clinical care.
        J Cardiovasc Surg (Torino). 2021; 62: 203-210https://doi.org/10.23736/S0021-9509.21.11816-6
        • Zhou W.Z.
        • Li W.
        • Shen H.
        • et al.
        CHDbase: A comprehensive knowledgebase for congenital heart disease-related genes and clinical manifestations.
        Genomics Proteomics Bioinformatics. 2022; S1672-0229: 00093-00096https://doi.org/10.1016/j.gpb.2022.08.001
        • Biesecker L.G.
        • Harrison S.M.
        ClinGen Sequence Variant Interpretation Working Group. The ACMG/AMP reputable source criteria for the interpretation of sequence variants.
        Genet Med. 2018; 20: 1687-1688https://doi.org/10.1038/gim.2018.42
        • Jarvik G.P.
        • Browning B.L.
        Consideration of cosegregation in the pathogenicity classification of genomic variants.
        Am J Hum Genet. 2016; 98: 1077-1081https://doi.org/10.1016/j.ajhg.2016.04.003
        • Abou Tayoun A.N.
        • Pesaran T.
        • DiStefano M.T.
        • et al.
        Recommendations for interpreting the loss of function PVS1 ACMG/AMP variant criterion.
        Hum Mutat. 2018; 39: 1517-1524https://doi.org/10.1002/humu.23626
        • McLaren W.
        • Gil L.
        • Hunt S.E.
        • et al.
        The Ensembl variant effect predictor.
        Genome Biol. 2016; 17: 122https://doi.org/10.1186/s13059-016-0974-4
        • Karczewski K.J.
        • Francioli L.C.
        • Tiao G.
        • et al.
        The mutational constraint spectrum quantified from variation in 141,456 humans.
        Nature. 2020; 581 (Published correction appears in Nature. 2021;590(7846):E53. Published correction appears in Nature. 2021;597(7874):E3-E4): 434-443
        • National Society of Vascular Surgery
        Chinese expert consensus on the genetic testing and clinical management of heritable thoracic aortic aneurysm/dissection.
        Chin Circ J. 2019; 34: 319-325https://doi.org/10.3969/j.issn.1000-3614.2019.04.003
      1. The Clinical Genome Resource ClinGen Variant Curation SOP Committee. ClinGen general sequence variant curation process standard operating procedure. Version 2.0. Clinical Genome Resource. Published January 2021. Accessed March 17, 2021. https://clinicalgenome.org/site/assets/files/5933/variant_curation_sop_2_0_jan_2021.pdf

      2. Clinical Genome Resource. ClinGen Sequence Variant Interpretation Recommendation for de novo Criteria (PS2/PM6). Version 1.1. Clinical Genome Resource. Published March 18, 2018. Updated May 5, 2021. Accessed August 20, 2021. https://clinicalgenome.org/docs/ps2-pm6-recommendation-for-de-novo-ps2-and-pm6-acmg-amp-criteria-version-1.0

        • Yang H.
        • Luo M.
        • Fu Y.
        • et al.
        Genetic testing of 248 Chinese aortopathy patients using a panel assay.
        Sci Rep. 2016; 633002https://doi.org/10.1038/srep33002
        • Li J.
        • Yang L.
        • Diao Y.
        • et al.
        Genetic testing and clinical relevance of patients with thoracic aortic aneurysm and dissection in northwestern China.
        Mol Genet Genomic Med. 2021; 9e1800https://doi.org/10.1002/mgg3.1800
        • Duan D.M.
        • Chiu H.H.
        • Chen P.L.
        • et al.
        Clinical manifestations and genetic characteristics in the Taiwan thoracic aortic aneurysm and dissection cohort – a prospective cohort study.
        J Formos Med Assoc. 2022; 121: 1093-1101https://doi.org/10.1016/j.jfma.2021.08.016
        • Arnaud P.
        • Hanna N.
        • Benarroch L.
        • et al.
        Genetic diversity and pathogenic variants as possible predictors of severity in a French sample of nonsyndromic heritable thoracic aortic aneurysms and dissections (nshTAAD).
        Genet Med. 2019; 21: 2015-2024https://doi.org/10.1038/s41436-019-0444-y
        • Li Y.
        • Kong Y.
        • Duan W.
        • et al.
        Evaluating the monogenic contribution and genotype-phenotype correlation in patients with isolated thoracic aortic aneurysm.
        Eur J Hum Genet. 2021; 29: 1129-1138https://doi.org/10.1038/s41431-021-00857-2
        • Guo D.C.
        • Papke C.L.
        • Tran-Fadulu V.
        • et al.
        Mutations in smooth muscle alpha-actin (ACTA2) cause coronary artery disease, stroke, and Moyamoya disease, along with thoracic aortic disease.
        Am J Hum Genet. 2009; 84: 617-627https://doi.org/10.1016/j.ajhg.2009.04.007