Quantile regression for challenging cases of eQTL mapping.

Area of Science:

• Genomics
• Statistical Genetics

Background:

• Expression quantitative trait loci (eQTL) mapping links genetic variants to gene expression.
• RNA sequencing (RNA-seq) data presents analysis challenges like overdispersion and dropouts.
• Conventional linear models assume Gaussian errors, which are often violated by RNA-seq data, leading to increased errors.

Purpose of the Study:

• To propose and evaluate quantile regression as a robust alternative for eQTL mapping.
• To address the limitations of traditional methods in handling RNA-seq data complexities.
• To improve the accuracy and reliability of eQTL detection.

Main Methods:

• Utilized quantile regression to model gene expression in the presence of overdispersion and dropouts.
• Conducted simulation studies to assess the robustness and performance of quantile regression.
• Analyzed real RNA-seq data to compare quantile regression with conventional linear models.

Main Results:

• Quantile regression demonstrated robustness to outliers and dropouts in simulations.
• The method significantly improved eQTL mapping accuracy compared to standard approaches.
• Real data analysis revealed distinct eQTL discoveries between quantile regression and linear models, especially with high overdispersion or dropout effects.

Conclusions:

• Quantile regression provides a more reliable and accurate approach for eQTL mapping, particularly for genes with complex expression patterns.
• This method is well-suited for large-scale eQTL studies dealing with RNA-seq data challenges.
• Quantile regression warrants increased attention in the field of genetic variant analysis.