AdaReg: data adaptive robust estimation in linear regression with application in GTEx gene expressions

Area of Science:

• Biostatistics
• Genomics
• Bioinformatics

Background:

• The Genotype-Tissue Expression (GTEx) project offers extensive gene expression data across diverse tissues.
• Estimating tissue-specific gene expression effects is challenging due to technical noise and biological heterogeneity.
• Existing methods struggle to robustly account for variations in gene expression across tissues.

Purpose of the Study:

• To develop a robust statistical method for accurately estimating tissue effects in gene expression data.
• To address the challenge of heterogeneous gene expression patterns across different tissue types.
• To improve the reliability of tissue effect estimation by adapting to data complexities.

Main Methods:

• Employed a robust estimation approach based on gamma-density-power-weighting.
• Developed a data-adaptive procedure to tune the gamma parameter, balancing bias and variance.
• Constructed a robust likelihood criterion using weighted densities within a mixture model framework.
• Introduced the Adaptive Regression (AdaReg) procedure for robust linear regression.

Main Results:

• The proposed data-adaptive gamma selection procedure effectively balances bias-variance trade-offs.
• Simulation studies demonstrated AdaReg's superior performance compared to fixed gamma and other robust methods.
• Real-data application on GTEx heart tissue samples confirmed AdaReg's advantage in estimating tissue effects.

Conclusions:

• The developed AdaReg method provides a robust and adaptive approach for analyzing gene expression data.
• This method significantly enhances the accuracy of tissue effect estimation, especially in the presence of noise and heterogeneity.
• AdaReg offers a valuable tool for researchers utilizing large-scale genomic datasets like GTEx.