AutoCNV: a semiautomatic CNV interpretation system based on the 2019 ACMG/ClinGen Technical Standards for CNVs

Abstract

Background The American College of Medical Genetics and Genomics (ACMG) and the Clinical Genome Resource (ClinGen) presented technical standards for interpretation and reporting of constitutional copy-number variants in 2019 (the standards). Although ClinGen developed a web-based CNV classification calculator based on scoring metrics, it can only track and tally points that have been assigned based on observed evidence. Here, we developed AutoCNV (a semiautomatic automated CNV interpretation system) based on the standards, which can automatically generate predictions on 18 and 16 criteria for copy number loss and gain, respectively. Results We assessed the performance of AutoCNV using 72 CNVs evaluated by external independent reviewers and 20 illustrative case examples. Using AutoCNV, it showed that 100 % (72/72) and 95 % (19/20) of CNVs were consistent with the reviewers’ and ClinGen-verified classifications, respectively. AutoCNV only required an average of less than 5 milliseconds to obtain the result for one CNV with automated scoring. We also applied AutoCNV for the interpretation of CNVs from the ClinVar database and the dbVar database. We also developed a web-based version of AutoCNV (wAutoCNV). Conclusions AutoCNV may serve to assist users in conducting in-depth CNV interpretation, to accelerate and facilitate the interpretation process of CNVs and to improve the consistency and reliability of CNV interpretation.

Combination of Expert Guidelines-based and Machine Learning-based Approaches Leads to Superior Accuracy of Automated Prediction of Clinical Effect of Copy Number Variations

“…MarCNV is a new automated tool for evaluation of mainly the first three sections of the ACMG criteria 2 (Table 1). It is similar to the already published ClassifyCNV 4 and AutoCNV 5 , which, however, were not available at the time of the implementation.…”

“…At present, several approaches are available to clinicians in the decision-making process regarding interpretation of CNVs' clinical significance/impact. In the main approach, CNVs can be classified following a recently published joint consensus recommendation of the ACMG and the Clinical Genome Resource (ClinGen) (in brief: ACMG criteria) 2 , by selecting choices/options and respective score values for individual guideline categories [3][4][5] . These professional standards tend to encourage consistency and transparency of CNV evaluation across clinical laboratories.…”

“…This has driven the effort to create semi-automated tools that can partially eliminate the tedious process of annotating and assigning a certain score for CNVs by automating the search for information related to the genomic content of CNVs from various sources/databases. Examples of such classification tools that help to automate CNV annotation and classification of individual sections of ACMG criteria include AnnotSV 3 , ClassifyCNV 4 and AutoCNV 5 . These tools slightly differ in the data sources they use, in the categories from five sections of ACMG criteria that they are able to evaluate (Table 1), and in the way they interpret the details of the ACMG criteria.…”

CNV-ClinViewer: Enhancing the clinical interpretation of large copy-number variants online

“…Therefore, in the development of the CNV-ClinViewer, we emphasized interactive analyses of the CNVs beyond a classification score and allow the user to dynamically inspect and compare the results in detail. In comparison, existing open-source tools [3][4][5] focus more on the classification scores than on the interactive inspection and further investigations. We believe that the comprehensive approach of the CNV-ClinViewer will empower users with more trust in the results, the agency to test hypotheses, and enable them to apply their domain knowledge while simultaneously leveraging the benefits of automation.…”

“…This specifically affects data that cannot be extracted from public online resources, such as detailed information about the phenotype, including family history, CNV inheritance pattern, and many other criteria currently not integrated in the clinical significance classification. For the semiautomated tools 5 , the extracted and manually entered information is not explorable within the same web-application interface and comparative visual inspection of CNVs with various genomic and clinical data sources is not possible because current tools do not provide such functionality [3][4][5] . While the current focus of those existing tools is on clinical significance classification, they are not designed to perform ad-hoc evaluations of new data such as genotype-phenotype analyses to expose undiscovered patterns of CNV localization in patient cohorts.…”