MUREN

MUREN performs multi-reference pairwise normalization of RNA sequencing (RNA-seq) data to mitigate confounding factors and preserve biological expression asymmetries using robust least trimmed squares regression and a linear model adjusted for reference effects.

Key Features:

• Multi-Reference Pairwise Normalization: Performs pairwise normalization with respect to multiple reference samples selected from representative datasets to avoid dependence on a single reference.
• Robust Least Trimmed Squares Regression: Integrates pairwise-normalized data through a linear model adjusted for reference effects using robust least trimmed squares regression to reduce the influence of outliers.
• Preservation of Biological Asymmetry: Maintains skewness inherent in biological data to preserve asymmetric differentiation signals such as those observed in single-cell RNA-seq cell-cycle datasets.
• Evaluation of Normalization Quality: Assesses normalization quality by examining densities of pairwise differentiations and adjusting them toward zero while preserving essential biological asymmetries.
• Outlier Immunity: Achieves resilience to individual outlier samples by robustly integrating pre-normalized counts across multiple references.

Scientific Applications:

• Cross-condition RNA-seq normalization: Enables identification of biological expression differences between samples across varied experimental conditions without relying solely on housekeeping genes.
• Single-cell transcriptomics: Supports analyses that require preservation of asymmetric expression patterns and subtle differences in single-cell RNA-seq studies, including cell-cycle–related asymmetries.
• Large-scale heterogeneous datasets: Provides robust normalization for large and heterogeneous datasets by reducing the impact of outlier samples.

Methodology:

Performs pairwise normalization relative to multiple reference samples, integrates pairwise-normalized counts via a linear model adjusted for reference effects, combines results using robust least trimmed squares regression, and evaluates normalization by inspecting densities of pairwise differentiations and adjusting them toward zero while preserving skewness.