Evaluation of phenotype-driven gene prioritization methods for Mendelian diseases

Area of Science:

• Genomics
• Medical Genetics
• Bioinformatics

Background:

• Identifying disease-causing genes from next-generation sequencing (NGS) data for Mendelian disorders presents significant challenges.
• Numerous phenotype-driven gene prioritization methods have been developed, utilizing patient genotype and phenotype information or phenotype data alone.
• Existing retrospective benchmarks often lack the statistical power to definitively differentiate the performance of these ranking methods.

Purpose of the Study:

• To benchmark the performance of ten recognized causal-gene prioritization methods.
• To evaluate these methods using a relatively unbiased methodology on a substantial dataset.
• To provide guidance for selecting appropriate tools for computer-assisted diagnosis in Mendelian diseases.

Main Methods:

• Benchmarking ten causal-gene prioritization methods.
• Utilizing 305 cases from the Deciphering Developmental Disorders (DDD) project and 209 in-house cases.
• Comparing methods based on input data types, including Human Phenotype Ontology (HPO) terms and Variant Call Format (VCF) files versus phenotypic data alone.

Main Results:

• Methods incorporating HPO terms and VCF files demonstrated superior overall performance compared to those relying solely on phenotypic data.
• LIRICAL and AMELIE emerged as top-performing methods in the benchmark.
• A complementary relationship was observed between LIRICAL and AMELIE, suggesting potential for an integrated approach.

Conclusions:

• The study provides a valuable reference for computer-assisted diagnosis in Mendelian diseases.
• Integrating HPO terms and VCF data enhances gene prioritization accuracy.
• Future research directions include exploring integrative approaches with methods like LIRICAL and AMELIE to improve diagnostic efficiency.