Clinical exome (CE) sequencing has become a first-tier diagnostic test for hereditary diseases; however, its diagnostic rate is around 30–50%. In this study, we aimed to increase the diagnostic yield of CE using a custom reanalysis algorithm. Sequencing data were available for three cohorts using two commercial protocols applied as part of the diagnostic process. Using these cohorts, we compared the performance of general and clinically relevant variant calling and the efficacy of an in-house bioinformatic protocol (FJD-pipeline) in detecting causal variants as compared to commercial protocols. On the whole, the FJD-pipeline detected 99.74% of the causal variants identified by the commercial protocol in previously solved cases. In the unsolved cases, FJD-pipeline detects more INDELs and non-exonic variants, and is able to increase the diagnostic yield in 2.5% and 3.2% in the re-analysis of 78 cancer and 62 cardiovascular cases. These results were considered to design a reanalysis, filtering and prioritization algorithm that was tested by reassessing 68 inconclusive cases of monoallelic autosomal recessive retinal dystrophies increasing the diagnosis by 4.4%. In conclusion, a guided NGS reanalysis of unsolved cases increases the diagnostic yield in genetic disorders, making it a useful diagnostic tool in medical genetics.

BackgroundThe introduction of next-generation sequencing in the diagnosis of genetic diseases has increased the known repertoire of causal variants and genes involved, as well as the amount of genomic information produced, that is not always shared or reused.MethodsWe built an allelic frequency database for a heterogeneous cohort of genetic diseases to explore the aggregated genomic information and boost the diagnosis in inherited retinal dystrophies (IRD). We retrospectively selected 5683 index-cases with clinical exome available, 1766 with IRD, and the rest with diverse genetic diseases. In our IRD cohort, 46% of the patients do not have conclusive diagnosis at the time of writing. We calculated the specific allele-frequencies of the solved and non-solved IRD subcohorts and compared them with suitable pseudocontrols that were used to prioritize variants. In addition, we developed a method to highlight genes with more frequent pathogenic variants in non-solved IRD cases than in pseudocontrols weighted by the increment of benign variants in the same comparison. Our resource was also used to calculate the carrier frequency of deleterious variants in IRD genes.ResultsPrioritized variants were significantly enriched in deleterious variants in non-solved IRD cases but not in solved. Focusing on non-solved IRD cases, we prioritized variants with a significant increment of frequencies in cases compared to pseudocontrols. Among them, eight variants may contribute to explain the phenotype of 10 cases. Applied to variants of uncertain significance (VUS) monitored in our laboratory, we detected 11 more frequent in IRD than in pseudocontrols, and 10 of them were reclassified as likely-pathogenic according to ACMG guidelines. We also identified 18 genes with an accumulated pathogenicity in non-solved IRD samples for further studies that provided new insights in five cases. Most prevalent genes carrying pathogenic mutations are ABCA4 (∼7%) and USH2A (∼3%).ConclusionsA cohort-specific database of allele frequencies is able to diagnose cryptic non-solved IRD cases, reclassify VUS, propose candidate genes, and calculate CF on genes of interest. The database operates as an engine providing new hypotheses in non-solved cases as well as offering new resources for genetic counselling.