Estimation of cis-eQTL effect sizes using a log of linear model

Area of Science:

• Genomics
• Biomedicine
• Statistical Genetics

Background:

• Expression Quantitative Trait Loci (eQTL) detection is crucial in genomics.
• Estimating eQTL effect size is less studied than detection.
• Current methods use normalization leading to inaccurate effect size estimates.

Purpose of the Study:

• Propose a novel log-of-linear model (ACME) for eQTL action.
• Develop a method for accurate estimation of eQTL effect sizes.
• Provide a biologically coherent model for cis-acting eQTLs.

Main Methods:

• Developed the Additive Contribution Model Estimation (ACME) model.
• Employed a non-linear least-squares algorithm for maximum likelihood fitting.
• Validated using simulated data and Genotype Tissue Expression (GTEx) project data.

Main Results:

• ACME provides interpretable and accurate eQTL effect size estimates.
• Demonstrated minimal evidence for dominance effects, supporting a simple biological model.
• Showed Type-I error control and highlighted detrimental effects of standard normalization on power and accuracy.
• ACME analysis of GTEx data revealed significant differences in eQTL ranking and sign compared to standard methods.

Conclusions:

• The ACME model offers a biologically coherent and accurate approach to eQTL effect size estimation.
• Rank-based normalizations are unnecessary and can reduce power and accuracy.
• ACME provides a robust alternative for eQTL analysis, improving estimation and interpretation.