CorrAdjust unveils biologically relevant transcriptomic correlations by efficiently eliminating hidden confounders

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Confounding variables are often overlooked in RNA-RNA correlation analysis, potentially impacting results.
• Accurate correlation analysis is crucial for understanding gene regulation and biological pathways.

Purpose of the Study:

• To introduce CorrAdjust, a novel method for identifying and correcting hidden confounders in RNA-RNA correlation.
• To provide a robust tool for enhancing the accuracy and interpretability of RNA sequencing data analysis.

Main Methods:

• CorrAdjust selects principal components to residualize expression data by maximizing "reference pair" enrichment.
• Utilizes a novel enrichment-based metric tailored for correlation data, offering RNA-level interpretability.
• Evaluated on 25,063 human RNA-seq datasets from TCGA, GTEx, and Geuvadis.

Main Results:

• CorrAdjust outperforms current state-of-the-art methods in RNA-RNA correlation analysis.
• Significantly enhances the enrichment of validated miRNA targets among negatively correlated miRNA-mRNA pairs.
• Demonstrates superior performance in integrating small RNA and mRNA sequencing data.

Conclusions:

• CorrAdjust offers a powerful solution for addressing confounding variables in RNA-RNA correlation.
• The method provides enhanced accuracy and interpretability, particularly for integrated small and messenger RNA sequencing data.
• Available with documentation and tutorials for broader adoption in bioinformatics research.