The Importance of Automation in Genetic Diagnosis: Lessons from Analyzing an Inherited Retinal Degeneration Cohort with the Mendelian Analysis Toolkit (MATK)

Abstract

Purpose: In Mendelian disease diagnosis, variant analysis is a repetitive, error-prone, and time-consuming process. To address this, we have developed the Mendelian Analysis Toolkit (MATK), a configurable automated variant ranking program. Methods: MATK aggregates variant information from multiple annotation sources and uses expert-designed rules with parameterized weights to produce a ranked list of potentially causal solutions. MATK performance was measured by a comparison of MATK-aided versus human domain-expert analyses of 1060 inherited retinal degeneration (IRD) families investigated with an IRD-specific gene panel (589 families) and exome sequencing (471 families). Results: When comparing MATK-assisted analysis to expert curation, we found that 97.3% (541/556) of potential solutions found by experts were also identified by the MATK-assisted analysis. Furthermore, MATK-assisted analysis identified 114 additional potential solutions. The software also showed utility in data reanalysis after remapping to the GRCh38 genome build. Conclusion: MATK expedites the process of identifying likely solving variants in Mendelian traits and helps to remove variability coming from human error and researcher bias. MATK facilitates data re-analysis to keep up with the constantly improving annotation sources and NGS processing pipelines. The software is open source and available at https://gitlab.partners.org/meei-ogi-bioinformatics/MendelAnalysis

StrVCTVRE: A supervised learning method to predict the pathogenicity of human genome structural variants

“…This variant was confirmed by CMG researchers as causative of the proband’s disease. 51 At the ClinVar 90% sensitivity threshold (StrVCTVRE score > 0.37), StrVCTVRE identified 31 of 34 disease-causing SVs (91%) as potentially pathogenic.…”

“…These SVs were recently identified through exome sequencing of cohorts with undiagnosed neuromuscular or retinal degeneration disorders. 48 , 49 , 50 , 51 , 52 Clinical researchers determined these rare SVs were disease causing or likely disease causing. To avoid overlap between these CMG clinical SVs and StrVCTVRE training SVs, we used a leave-one-chromosome-out approach in which 24 separate StrVCTVRE classifiers were developed, one for each chromosome.…”

“…This variant was confirmed by CMG researchers as causative of the proband's disease. 51 29 Because there is some overlap between training ClinVar SVs and DECIPHER SVs, we tested on DECIPHER by using a leave-one-chromosome-out approach, as described above. Additionally, to ensure this DECIPHER test set is independent from our ClinVar test set, we considered only DECIPHER SVs with a reciprocal overlap of less than 10% with any SV used in training or testing StrVCTVRE.…”

“…Under reasonable request, DECIPHER data can be requested from https://www.deciphergenomics.org/about/data-sharing . A subset of the recent CMG clinical SVs found to cause neuromuscular and retinal degeneration disorders have been made publicly available, 48 , 49 , 50 , 51 , 52 but the full dataset is not currently publicly available. These data can be made available by Anne O’Donnell on reasonable request.…”

A high throughput splicing assay to investigate the effect of variants of unknown significance on exon inclusion

“…Skipping of an in-frame exon, will not lead to transcript degradation but may lead to a non-functional protein if an important domain is deleted. Weighing these criteria, we evaluated variants that showed an effect on exon skipping in a cohort of ~3000 IRD cases analyzed by targeted sequencing or exome sequencing 13,38 . We found ten cases for whom the splicing results likely lead to new genetic diagnoses, including six cases with autosomal recessive gene solutions (ABCA4 88,89 , EYS 90,91 , FLVCR1 92 , USH2A 93 ) three with X-linked disease, (CHM 94 , RPGR 95 ) and one dominant haploinsufficient gene (JAG1 96,97 ) (Table 2).…”