MutationDistiller: user-driven identification of pathogenic DNA variants.

Daniela Hombach,Markus Schuelke,Nadja Ehmke,Jana Marie Schwarz,Dominik Seelow,Ellen Knierim,Björn Fischer-Zirnsak

doi:10.1093/nar/gkz330

Abstract

MutationDistiller is a freely available online tool for user-driven analyses of Whole Exome Sequencing data. It offers a user-friendly interface aimed at clinicians and researchers, who are not necessarily bioinformaticians. MutationDistiller combines MutationTaster's pathogenicity predictions with a phenotype-based approach. Phenotypic information is not limited to symptoms included in the Human Phenotype Ontology (HPO), but may also comprise clinical diagnoses and the suspected mode of inheritance. The search can be restricted to lists of candidate genes (e.g. virtual gene panels) and by tissue-specific gene expression. The inclusion of GeneOntology (GO) and metabolic pathways facilitates the discovery of hitherto unknown disease genes. In a novel approach, we trained MutationDistiller's HPO-based prioritization on authentic genotype–phenotype sets obtained from ClinVar and found it to match or outcompete current prioritization tools in terms of accuracy. In the output, the program provides a list of potential disease mutations ordered by the likelihood of the affected genes to cause the phenotype. MutationDistiller provides links to gene-related information from various resources. It has been extensively tested by clinicians and their suggestions have been valued in many iterative cycles of revisions. The tool, a comprehensive documentation and examples are freely available at https://www.mutationdistiller.org/

Highlights

Generation Sequencing has led to a large advance in the elucidation of monogenic diseases
As the plethora of data entry options might be overwhelming for first-time users, we provide a number of dedicated user modes for different interest groups, limiting the
We are planning to incorporate a number of features if new fitting data resources are developed: As MutationDistiller bases its predictions on MutationTaster, it can currently only detect variants located in protein-coding transcripts

Summary

INTRODUCTION

Generation Sequencing has led to a large advance in the elucidation of monogenic diseases. The data sources integrated into MutationDistiller allow assessing a case from various angles, making the tool attractive to a variety of user groups: Clinical symptoms or diagnoses can be entered via the widely used resources HPO [11], OMIM [21] or Orphanet [22]. We assessed the resulting weight combinations to find a balanced solution that would consistently rank the causative gene highly while showing a low rate of unsolved cases; and after careful consideration we settled for a weight of 5 for direct matches, 0.05 for descendants and 2 for ancestor terms We chose this approach rather than dynamically searching for the optimal weight distribution to avoid overfitting on this relatively small data set. A more in-depth description of the comparison together with the results can be found on our web site and in the supplement

Findings

DISCUSSION AND OUTLOOK

CONCLUSION