Selection-adjusted inference: an application to confidence intervals for cis-eQTL effect sizes.

Area of Science:

• Genetics
• Bioinformatics
• Statistical Genomics

Background:

• Expression quantitative trait loci (eQTL) studies link genetic variants to gene expression levels.
• High-throughput technologies enable large-scale gene expression and polymorphism data generation.
• Limited sample sizes, especially for critical human tissues, pose a significant challenge in eQTL studies.

Purpose of the Study:

• To address the under-addressed problem of estimating effect sizes for genetic variants in eQTL studies, accounting for selection bias.
• To develop a reliable method for calculating confidence intervals for eQTL effect sizes, crucial for medical relevance.

Main Methods:

• A novel conditional inference approach is proposed for eQTL effect size estimation.
• The method employs a randomized hierarchical strategy, incorporating selection steps and using randomness instead of data-splitting.
• The approach aims to maximize the utilization of limited available data.

Main Results:

• Naive confidence intervals for eQTL effect sizes may not accurately cover true values.
• The proposed method provides reliable confidence intervals for eQTL effect sizes.
• Analysis suggests potential underestimation of genetic polymorphisms influencing gene expression levels.

Conclusions:

• Accurate estimation of eQTL effect sizes is critical, especially with limited sample sizes.
• The conditional inference approach offers a robust solution for reliable eQTL effect size estimation.
• This methodology enhances the understanding of genetic contributions to gene expression in medically relevant tissues.