MOCCASIN: a method for correcting for known and unknown confounders in RNA splicing analysis

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• High-throughput sequencing technologies like RNA sequencing (RNA-Seq) have enabled large-scale gene expression analysis.
• Confounding factors are known to influence gene expression measurements, but their impact on RNA splicing quantification is less understood.
• Existing tools for analyzing RNA splicing variations do not adequately address confounding effects.

Purpose of the Study:

• To assess the impact of confounding factors on both gene expression and RNA splicing quantifications using large public RNA-Seq datasets.
• To develop a computational method for correcting confounding effects in RNA splicing data.
• To provide a validated tool and resources for more accurate RNA splicing analysis.

Main Methods:

• Analysis of two large public RNA-Seq datasets (TARGET and ENCODE) to evaluate confounding factor effects.
• Development of MOCCASIN, a novel computational method designed to correct for known and unknown confounders in RNA splicing quantification.
• Validation of MOCCASIN using both synthetic and real-world RNA-Seq data.

Main Results:

• Confounding factors were found to affect RNA splicing quantifications as much as, or more than, gene expression quantifications.
• Unwanted sources of variation were identified in both gene expression and splicing data from the TARGET and ENCODE datasets.
• MOCCASIN demonstrated effectiveness in correcting confounding effects on RNA splicing quantification.

Conclusions:

• Confounding factors represent a significant challenge in RNA splicing analysis, comparable to their impact on gene expression.
• The developed MOCCASIN method effectively corrects for confounding variations in RNA splicing data.
• MOCCASIN provides a valuable resource for improving the accuracy and reliability of RNA splicing quantification in bioinformatics research.