An Improved Phenotype-Driven Tool for Rare Mendelian Variant Prioritization: Benchmarking Exomiser on Real Patient Whole-Exome Data

Abstract

Next-generation sequencing has revolutionized rare disease diagnostics, but many patients remain without a molecular diagnosis, particularly because many candidate variants usually survive despite strict filtering. Exomiser was launched in 2014 as a Java tool that performs an integrative analysis of patients' sequencing data and their phenotypes encoded with Human Phenotype Ontology (HPO) terms. It prioritizes variants by leveraging information on variant frequency, predicted pathogenicity, and gene-phenotype associations derived from human diseases, model organisms, and protein-protein interactions. Early published releases of Exomiser were able to prioritize disease-causative variants as top candidates in up to 97% of simulated whole-exomes. The size of the tested real patient datasets published so far are very limited. Here, we present the latest Exomiser version 12.0.1 with many new features. We assessed the performance using a set of 134 whole-exomes from patients with a range of rare retinal diseases and known molecular diagnosis. Using default settings, Exomiser ranked the correct diagnosed variants as the top candidate in 74% of the dataset and top 5 in 94%; not using the patients' HPO profiles (i.e., variant-only analysis) decreased the performance to 3% and 27%, respectively. In conclusion, Exomiser is an effective support tool for rare Mendelian phenotype-driven variant prioritization.

Keywords: bioinformatics; human phenotype ontology; inherited retinal disease; phenotypic similarity; rare disease; variant prioritization; whole-exome sequencing; whole-genome sequencing.

Figure 1

Overview of the Exomiser workflow analysis: The diagram depicts the two main steps in an Exomiser analysis: (a) variant filtering and (b) variant prioritization. A single-sample variant call format (VCF) file and corresponding list of Human Phenotype Ontology (HPO) terms are mandatory inputs. If a multi-sample VCF file (from a nuclear family) is used in the analysis, the user must provide a corresponding pedigree file. In the filtering step, variants are filtered according to type of variant, allele frequency in selected databases, and mode of inheritance as per user-defined options and values. In the prioritization step, a variant score is calculated based on allele frequency and pathogenicity as predicted by user-defined in silico algorithms, together with a gene-specific phenotype score based on the semantic similarity of the patient’s HPO terms and phenotypic annotation in known human disease gene, mouse, zebrafish and protein–protein interaction databases. Finally, the Exomiser score is obtained from the variant score and phenotypic score within a logistic regression classifier framework and used for variant prioritization. PED: Pedigree, OMIM: Online mendelian inheritance in man, IMPC: International mouse phenotyping consortium, STRING: Search tool for the retrieval of interacting genes/proteins.

Figure 2

HPO graphic visualization for the HPO-encoded clinical diagnosis Leber congenital amaurosis (Retinal dystrophy, HP:0000556; visual impairment, HP:0007758, undetectable electroretinogram, HP:0000550; and Nystagmus, HP:0000639).

Figure 3

Exomiser performance on the IRD patient dataset using different analysis settings. The left-hand side panel shows the categorical percentage distribution of the disease-causing variant ranking according to five mutually exclusive disease-causing ranking bins (“Top”, “2–5”, “6–10”, “>10”, and “Filtered out/Not prioritized”) per each analysis setting. The right-hand panel shows the corresponding cumulative percentage distributions.

Figure 4

Screenshot of the Exomiser HTML output file (from the DEFAULT analysis) for patient P127 who was clinically diagnosed with Usher syndrome type II and molecularly diagnosed with frameshift elongation c.920_923dup:p.(His308Glnfs*16) and inframe deletion c.3832_3834del:p.(Leu1278del) in USH2A (Table S1).