A comprehensive evaluation of collapsing methods using simulated and real data: excellent annotation of functionality...

Area of Science:

• Genetics
• Statistical Genetics
• Genomic Analysis

Background:

• Next-generation sequencing (NGS) enables genome-wide rare variant analysis.
• Investigating the rare variant-common disease hypothesis requires specialized statistical methods.
• Existing methods often aggregate rare variants within genomic regions.

Purpose of the Study:

• To compare the performance of 15 rare variant collapsing methods.
• To evaluate statistical power and type I error rates across different genomic regions and variant frequencies.
• To validate findings using both simulated and real-world genetic data.

Main Methods:

• Extensive simulation study using Genetic Analysis Workshop 17 data.
• Comparison of 15 collapsing methods based on minor allele frequency and variant functionality.
• Real data analysis of SLCO1B1 gene variants and methotrexate clearance in acute lymphoblastic leukemia patients.

Main Results:

• Many collapsing methods exhibited substantially inflated type I error rates.
• Only four methods maintained valid type I errors across all tested scenarios.
• No single method consistently detected true associations in simulated data; variant functionality annotation is critical.
• Real data analysis confirmed simulation findings.

Conclusions:

• Many current rare variant collapsing methods lack statistical validity.
• Accurate variant functionality annotation is essential for detecting true genetic associations.
• There is a need for computationally efficient, powerful, and valid statistical tests for rare variant association studies.