Evaluation of phenotype-driven gene prioritization methods for Mendelian diseases

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Phenotype-driven gene prioritization is crucial for diagnosing Mendelian diseases.
• Previous evaluations of tools like Exomiser and PhenIX have shown varying diagnostic yields.
• Discrepancies in reported performance may stem from differences in tool configuration.

Purpose of the Study:

• To identify key differences in the configuration of gene prioritization tools.
• To demonstrate the impact of recommended settings on the performance of Exomiser and PhenIX.
• To reconcile discrepancies in reported diagnostic yields for Mendelian disease gene prioritization.

Main Methods:

• Comparison of default tool settings with recommended configurations for Exomiser and PhenIX.
• Analysis of variant frequency, quality, and predicted pathogenicity filtering parameters.
• Evaluation of tool performance on two independent clinical datasets, including 161 singleton samples.

Main Results:

• Three critical differences in parameter settings were identified, impacting variant filtering and prioritization.
• Implementing recommended settings increased Exomiser's diagnostic performance from 34% to 72% on a test set.
• The findings suggest that previously reported lower diagnostic yields were likely due to suboptimal tool configuration.

Conclusions:

• The configuration of phenotype-driven gene prioritization tools significantly affects diagnostic performance.
• Recommended settings for Exomiser and PhenIX substantially improve the accuracy of identifying disease-causing genes.
• Reassessment of existing datasets with optimized parameters is warranted to accurately evaluate these tools.