Tweedieverse

Tweedieverse applies Tweedie-distribution-based statistical models to perform differential analysis of multi-omics and single-cell RNA sequencing (scRNA-seq) data, modeling heavy tails, sparsity, and varying count distributions across plate- and droplet-based platforms.

Key Features:

• Generalized Linear Models with Tweedie Distribution: Employs GLMs based on the Tweedie distribution to flexibly model the wide dynamic range of scRNA-seq count data.
• Zero-Inflated Tweedie Model: Incorporates a zero-inflated Tweedie extension to account for excess zeros commonly observed in scRNA-seq datasets.
• Robustness to data characteristics: Explicitly addresses heavy tails, sparsity, and varying count distributions across different experimental platforms.
• Comprehensive Benchmarking: Performs systematic evaluations using synthetic and published plate- and droplet-based scRNA-seq datasets and demonstrates improved statistical power and false discovery rate control compared to over 10 representative differential expression methods.

Scientific Applications:

• Differential expression in single-cell transcriptomics: Identification of differentially expressed genes in scRNA-seq experiments.
• Cellular heterogeneity and gene expression dynamics: Analysis of cellular heterogeneity and gene expression changes at the single-cell level, including disease-related investigations.
• Multi-omics differential analysis: Application to multi-omics datasets requiring flexible count-distribution modeling.

Methodology:

Implements generalized linear models with Tweedie distributions, includes a zero-inflated Tweedie model for excess zeros, and conducts systematic benchmarking using synthetic and published plate- and droplet-based scRNA-seq datasets against more than 10 differential expression methods.