ROSeq

ROSeq performs rank-based differential gene expression analysis of single-cell transcriptomics data to detect genes differing between annotated cell groups using filtered and normalized read count matrices.

Key Features:

• Input: Accepts filtered and normalized read count matrices and cell annotations or conditions as input.
• Rank-based modeling: Uses expression ranks as surrogates for transcript abundance instead of direct expression estimates.
• Dropout robustness: Handles high dropout rates commonly observed in single-cell data compared to parametric models such as zero-inflated Poisson/negative binomial and log-normal distributions.
• Statistical framework: Models ranked expression using the discrete generalized beta distribution (DGBD).
• Hypothesis testing: Employs a Wald-type test for comparing gene expression across phenotypically divergent groups of single cells.
• Error balance: Achieves a balance between Type I and Type II error rates for differential expression calls.
• Scalability and robustness: Demonstrates robustness to expression noise and scalability with increasing sample sizes.
• Parallel processing: Supports specification of the number of computational cores for parallel computation.

Scientific Applications:

• Differential expression analysis: Identifies differentially expressed genes in single-cell transcriptomics datasets.
• Group comparisons: Compares gene expression profiles between contrasting or phenotypically divergent groups of single cells.
• Cell-to-cell variability: Models gene expression patterns to investigate cell-to-cell variability within complex tissues.

Methodology:

ROSeq takes filtered and normalized read count matrices and cell annotations, converts expression values to ranks, models ranked expression with the discrete generalized beta distribution (DGBD), and applies a Wald-type test to compare groups while supporting parallel execution via user-specified computational cores.