SwarnSeq

SwarnSeq performs differential expression and differential zero inflation analysis for single-cell RNA sequencing (scRNA-seq) data to model dropout events and molecular capture processes that affect gene-level measurements.

Key Features:

• Differential Expression Analysis: Performs DE analysis using a statistical approach that accounts for biological processes leading to dropout events in scRNA-seq data.
• Differential Zero Inflation Analysis (DZI): Implements DZI analysis to model and compare zero inflation (dropout) across conditions.
• Modeling Molecular Capture Process: Integrates the molecular capture process into the statistical framework for more accurate expression modeling.
• Dropout-aware Statistical Modeling: Explicitly models dropout events to reduce bias in expression estimates and downstream comparisons.
• Benchmarking Performance: Evaluated against 11 existing methods across 10 real scRNA-seq datasets under three comparison settings.
• External Spike-ins: Can leverage external spike-in data to refine model behavior and performance.

Scientific Applications:

• Single-cell differential expression: Identifies genes with differential expression between conditions in scRNA-seq experiments while accounting for dropouts.
• Zero-inflation comparison: Detects differences in zero inflation (dropout rates) across cell populations or conditions.
• Characterization of cellular heterogeneity: Supports analyses that reveal variability in gene expression across individual cells.
• Identification of novel cell types: Aids in discriminating cell types based on dropout-aware expression signatures.
• Cross-protocol comparative analysis: Applicable to comparisons across diverse experimental conditions and scRNA-seq protocols.

Methodology:

Uses statistical modeling that accounts for biological causes of dropout and the molecular capture process, performs differential zero inflation and differential expression analyses, can incorporate external spike-ins, and was benchmarked against 11 methods across 10 real scRNA-seq datasets under three comparison settings.