SCATS

SCATS detects differential alternative splicing events in single-cell RNA-seq (scRNA-seq) data by modeling technical noise and aggregating exon-level signals to increase sensitivity for splicing detection.

Key Features:

• Noise modeling: Models capture efficiency and amplification bias, uses external spike-ins for non-UMI data, and accounts for transcriptional burstiness in UMI data to mitigate technical variability.
• Exon grouping: Groups exons originating from the same isoform(s) and aggregates spliced reads across exons to improve detection power at low sequencing depth.
• UMI versatility: Supports scRNA-seq data with and without Unique Molecular Identifiers (UMIs).
• Statistical performance: Employs a statistical framework that controls type I error rates and achieves superior power for detecting small splicing differences, with comparative analyses against Census and DEXSeq.

Scientific Applications:

• Detection of subtle splicing differences: Identifies differential alternative splicing across cell types and conditions where changes are small and sequencing depth is limited.
• Complex tissue analysis: Applied to studies of complex tissues such as the mouse brain to resolve splicing heterogeneity across cellular populations.
• Benchmarking and validation: Used with simulated and real scRNA-seq datasets for method validation and performance evaluation.

Methodology:

Implements a scRNA-seq–tailored statistical framework that models capture efficiency and amplification bias, uses external spike-ins for non-UMI data, accounts for transcriptional burstiness in UMI data, groups exons from the same isoform(s) to aggregate spliced reads, and performs statistical testing that controls type I error with demonstrated higher power relative to Census and DEXSeq.