AdaReg

AdaReg implements data-adaptive robust estimation for linear regression on RNA-seq gene expression data as an R package, providing robust parameter estimates in the presence of sequencing background noise and outliers for analyses such as those of GTEx.

Key Features:

• Robust Estimation Approach: Employs a γ-density-power-weight robust estimation method to manage technical variation, sequencing background noise, and unknown outlier effects in RNA-seq data.
• Data-Adaptive Parameter Tuning: Automatically selects the exponent parameter γ to optimize the bias–variance trade-off under mixture distributions.
• Robust Likelihood Criterion: Constructs a robust likelihood criterion based on weighted densities within a mixture model combining a Gaussian population distribution and an unknown outlier distribution.
• Heuristic Analysis and Performance: Provides heuristic analysis showing that the selected γ closely follows trends observed in Mean Squared Error (MSE) simulations across settings.
• Application to GTEx Data: Validated with simulation studies and real-data application to GTEx heart samples, showing advantages over fixed-γ procedures and other robust methods.

Scientific Applications:

• RNA-seq gene expression analysis: Robust linear regression estimation for large-scale RNA-seq gene expression datasets.
• Outlier and noise mitigation: Handling technical variation and sequencing background noise to improve reliability of transcriptomic inference.
• Population-scale studies: Analysis of cohort datasets such as GTEx, including tissue-specific applications (e.g., heart samples).

Methodology:

Uses γ-density-power-weight robust estimation; automatically selects γ to balance bias and variance under a mixture model comprising a Gaussian population distribution and an unknown outlier distribution; constructs a weighted-density robust likelihood criterion; evaluates γ selection via heuristic analysis and Mean Squared Error (MSE) simulations; validated with simulation studies and GTEx heart-sample analyses.