Shiba: a versatile computational method for systematic identification of differential RNA splicing across platforms

Area of Science:

• Molecular Biology
• Bioinformatics
• Genomics

Background:

• Alternative pre-mRNA splicing (AS) generates transcript diversity crucial for cell type variation.
• Accurate quantification of AS events is essential for understanding gene regulation.
• Existing tools often suffer from low sensitivity, high false positives, or require numerous replicates.

Purpose of the Study:

• To develop a comprehensive computational method (Shiba) for analyzing alternative splicing events across RNA-seq platforms.
• To extend Shiba's capabilities to single-cell RNA-seq data (scShiba).
• To improve accuracy, sensitivity, and reproducibility in AS analysis, reducing false positives.

Main Methods:

• Development of Shiba, integrating transcript assembly, splicing event identification, read counting, and differential splicing analysis.
• Implementation of scShiba using a pseudobulk approach for cluster-level AS analysis in single-cell RNA-seq.
• Validation using simulated data and real n=1 RNA-seq datasets, and application to single-cell data.

Main Results:

• Shiba accurately captures annotated and unannotated AS events with high sensitivity and reproducibility.
• Shiba effectively addresses junction read imbalance, significantly reducing false positives.
• scShiba successfully identified AS regulation in dopaminergic neurons and neuronal subtypes.

Conclusions:

• Shiba and scShiba provide robust and reproducible quantification of alternative splicing events across diverse RNA-seq platforms.
• These tools are valuable for mechanistic exploration of RNA splicing complexity, even with limited sample sizes or single-cell data.
• The methods are available in containers and pipelines, ensuring accessibility and reproducibility.