CNV Classification

Recently the ACMG and ClinGen published new technical standards and guidelines for CNV classification to assist clinical laboratories in the classification and reporting of CNVs.

These standards use a quantitative evidence-based evaluation framework (score-based).

The recommendations are divided into five main evidence criteria/sections:

Section 1: Is there any gene or the functional region that overlaps with CNV, or is it just an intergenic region?

Section 2: Is the CNV overlap with a known dosage-sensitive gene/region and what kind of overlap?

Section 3: Number of coding genes overlap with the CNV

Section 4: Additional literature/external evidence about the overlapping region

Section 5: Additional evidence from the evaluated patient

As described above, the new recommendations are a score based classification, and include five different sections, while each section contains different evidence rules (e.g 1A, 1B, 2A,2B 2C, etc.).

Franklin evaluates each evidence criterion, and if the rule is met, it scores it based on the recommendations.

After evaluating all evidence rules, the final score should be accumulated from all the evidence scores and based on that decide on the final classification when:

Score > 0.99 => Pathogenic

Score between [0.9,0.99] => Likely pathogenic

Score between [-0.9,0.9] => Variant of Uncetain Signficance

Score between [-0.9, -0.9-] => Likely Benign

Score < -0.99 => Benign

Franklin automatically implements the majority of the different criteria based on the new recommendations and reducing the burden of their computational and technical challenges. aiVCE’s overall performance incorrectly categorizing CNVs per the new ACMG/ClinGen CNV standards and guidelines were evaluated utilizing a dataset from the original paper which contains 114 CNVs — 58 deletions and 56 duplications — accompanied with classification findings from two independent clinical Reviewer Labs.

More details can be found in the poster “Implementation of ACMG/ClinGen Standards for CNV Classification Using AI: Benchmark Results” we presented in the 2020 ACMG conference.