Machine learning-based reclassification of germline variants of unknown significance: The RENOVO algorithm

Abstract

The increasing scope of genetic testing allowed by next-generation sequencing (NGS) dramatically increased the number of genetic variants to be interpreted as pathogenic or benign for adequate patient management. Still, the interpretation process often fails to deliver a clear classification, resulting in either variants of unknown significance (VUSs) or variants with conflicting interpretation of pathogenicity (CIP); these represent a major clinical problem because they do not provide useful information for decision-making, causing a large fraction of genetically determined disease to remain undertreated. We developed a machine learning (random forest)-based tool, RENOVO, that classifies variants as pathogenic or benign on the basis of publicly available information and provides a pathogenicity likelihood score (PLS). Using the same feature classes recommended by guidelines, we trained RENOVO on established pathogenic/benign variants in ClinVar (training set accuracy ¼ 99%) and tested its performance on variants whose interpretation has changed over time (test set accuracy ¼ 95%). We further validated the algorithm on additional datasets including unreported variants validated either through expert consensus (ENIGMA) or laboratory-based functional techniques (on BRCA1/2 and SCN5A). On all datasets, REN-OVO outperformed existing automated interpretation tools. On the basis of the above validation metrics, we assigned a defined PLS to all existing ClinVar VUSs, proposing a reclassification for 67% with >90% estimated precision. RENOVO provides a validated tool to reduce the fraction of uninterpreted or misinterpreted variants, tackling an area of unmet need in modern clinical genetics.

Classification of 101 BRCA1 and BRCA2 variants of uncertain significance by cosegregation study: A powerful approach

“…However, AI variant classification still requires critical examination by geneticists, especially with regard to the quality of the data on which AI tools are trained. 66…”

Analysis of missense variants in the human genome reveals widespread gene-specific clustering and improves prediction of pathogenicity

“…Some of these tools exploit these features per se, 24 , 25 , 26 , 27 , 28 , 29 , 30 whereas others take advantage of machine learning approaches and are trained on sets of pathogenic versus benign DNA changes. 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 , 40 , 41 , 42 , 43 , 44 , 45 Finally, meta-predictors offer an optimized combination of existing tools, such as those mentioned above, 46 , 47 , 48 , 49 , 50 , 51 , 52 , 53 , 54 , 55 although they sometimes suffer from circularity issues that are prone to producing falsely optimistic results (e.g., Grimm et al. 56 ).…”

Validation and Data-Integration of Yeast-Based Assays for Functional Classification of BRCA1 Missense Variants

“…FYI-BRCA1 classified 2355 variants by integrating data from a set of 22 already validated functional assays [ 17 ]. RENOVO, a machine learning-based approach, has proposed a classification of 67% VUS reported in ClinVar [ 25 ]. BRCA-ML, another machine learning-based strategy, was developed to evaluate the functional impact and classify thousands of variants that were localized in the RING and BRCT domain that are reported in several databases [ 26 ].…”