MUREN: a robust and multi-reference approach of RNA-seq transcript normalization.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• RNA-sequencing (RNA-seq) data normalization is crucial for identifying biological expression differences.
• Traditional methods often rely on housekeeping genes, whose universal applicability across diverse samples is uncertain.
• Confounding factors can obscure true biological variations in RNA-seq data.

Purpose of the Study:

• To develop a robust RNA-seq normalization method that does not depend on the assumption of universal housekeeping genes.
• To accurately identify biological expression differentiation while removing unwanted confounding factors.
• To provide a reliable tool for analyzing RNA-seq data across varied experimental conditions.

Main Methods:

• Proposed a novel pairwise normalization approach using multiple reference samples.
• Integrated pairwise normalization results using a linear model to adjust for reference effects.
• Employed robust least trimmed squares regression for pairwise normalization, inspired by statistical counterparts of housekeeping genes.

Main Results:

• The proposed method, MUREN, was compared against existing tools on standard datasets.
• MUREN demonstrated effectiveness in preserving asymmetric biological differentiation, as shown in single-cell RNA-seq data of the cell cycle.
• The normalization goodness was evaluated by analyzing the densities of pairwise differentiations.

Conclusions:

• MUREN normalizes RNA-seq data using a two-step statistical regression approach.
• The method adjusts the mode of differentiation towards zero while preserving biological skewness.
• MUREN's robust integration of multiple references makes it immune to outlier samples, enhancing data reliability.