rMATS: robust and flexible detection of differential alternative splicing from replicate RNA-Seq data

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Ultra-deep RNA sequencing (RNA-Seq) is crucial for analyzing pre-mRNA alternative splicing across the genome.
• Multivariate Analysis of Transcript Splicing (MATS) was previously developed for differential splicing analysis between two RNA-Seq samples.

Purpose of the Study:

• To introduce replicate MATS (rMATS), a novel statistical model and program for detecting differential alternative splicing using replicate RNA-Seq data.
• To provide a flexible hypothesis-testing framework that accommodates user-defined splicing change magnitudes.
• To offer guidance on designing effective RNA-Seq studies for alternative splicing analysis.

Main Methods:

• Development of a hierarchical statistical model within rMATS to account for both sampling uncertainty and inter-replicate variability.
• Implementation of models for both unpaired and paired replicate RNA-Seq data analysis.
• Evaluation of rMATS performance using simulated and real RNA-Seq datasets, including comparison with existing methods.

Main Results:

• rMATS demonstrated superior performance compared to two existing methods across all simulation settings.
• Experimental validation using RT-PCR on a prostate cancer cell line RNA-Seq dataset achieved a high validation rate of 94%.
• Analysis highlighted the critical importance of incorporating biological replicates and cautioned against pooling or merging replicate data due to sensitivity to outliers.

Conclusions:

• rMATS is a powerful and accurate tool for differential alternative splicing detection in RNA-Seq studies with replicates.
• The study provides essential guiding principles for designing future RNA-Seq experiments, emphasizing the necessity of biological replicates.
• rMATS is expected to be a valuable resource for diverse RNA-Seq projects investigating alternative splicing.