MIRAGE: A Bayesian statistical method for gene-level rare-variant analysis incorporating functional annotations.

Area of Science:

• Human Genetics
• Bioinformatics
• Statistical Genomics

Background:

• Rare variants in whole-exome or genome sequencing studies offer larger effect sizes and can pinpoint causal genes.
• Existing gene-based rare-variant association methods often rely on unrealistic assumptions, leading to underpowered analyses.
• There is a need for more powerful and flexible methods to analyze rare variants in genetic studies.

Purpose of the Study:

• To propose a novel Bayesian method, MIRAGE (mixture-model-based rare-variant analysis on genes), for enhanced rare-variant association analysis.
• To address the limitations of current methods by capturing the heterogeneity of variant effects within genes.
• To improve the power and accuracy of identifying genes associated with complex diseases using rare variants.

Main Methods:

• MIRAGE employs a mixture-model approach to differentiate between risk and non-risk variants within a gene.
• It analyzes summary statistics from trio sequencing or case-control studies.
• Prior probabilities for variants being risk variants are modeled using external genetic information.

Main Results:

• Simulations and analysis of an autism exome-sequencing dataset demonstrate MIRAGE's superior performance over current methods.
• MIRAGE significantly improves the power of rare-variant association analysis.
• Top genes identified by MIRAGE showed significant enrichment for known or plausible autism-risk genes.

Conclusions:

• MIRAGE offers a powerful and flexible Bayesian framework for rare-variant gene-level association testing.
• The method effectively handles the heterogeneity of variant effects, leading to improved discovery of disease-associated genes.
• MIRAGE represents a significant advancement in analyzing rare variants for human genetic research, particularly in complex disorders like autism.