rseqc-redux

rseqc-redux is a modernized fork of RSeQC (RNA-seq Quality Control), originally by Liguo Wang. It updates the RSeQC 5.0.1 codebase with modern Python packaging, comprehensive tests, and CI — while preserving the original functionality.

Requirements

• Python 3.10 or higher
• Indexed BAM files (.bam with corresponding .bam.bai index files)
• A BED-format gene model / reference annotation

Dependencies

Core dependencies are automatically installed:

• pysam — BAM/SAM file handling
• bx-python — Interval indexing and overlap
• numpy — Numerical operations
• pyBigWig — BigWig file I/O
• matplotlib / logomaker — Plotting

Installation

From PyPI (Recommended)

pip install rseqc-redux

From Source

git clone https://github.com/semenko/rseqc-redux.git
cd rseqc-redux
pip install .

Development Installation

git clone https://github.com/semenko/rseqc-redux.git
cd rseqc-redux
uv sync

Quick Start

# Basic BAM statistics
bam_stat -i sample.bam

# Infer RNA-seq strandedness
infer_experiment -r gene_model.bed -i sample.bam

# Transcript integrity number
tin -i sample.bam -r gene_model.bed

# Gene body coverage
geneBody_coverage -r gene_model.bed -i sample.bam -o output

# Read distribution over genomic features
read_distribution -r gene_model.bed -i sample.bam

# Junction annotation
junction_annotation -r gene_model.bed -i sample.bam -o output

Available Tools

Performance

Benchmarked against RSeQC 5.0.4 on the official RSeQC test data (chr22 subset, ~430K paired-end reads). See benchmark-data/BENCHMARK_REPORT.md for the full report with profiling details.

Key optimizations: pysam count_coverage() replacing manual pileup iteration (tin, geneBody_coverage), numpy vectorization (read_NVC, RPKM_saturation), single-pass CIGAR loops (clipping/insertion profiles), and dead code removal (bam_stat).

Upgrading from RSeQC 5.x or earlier rseqc-redux releases

Soft clipping bug fix (post-6.1.0)

fetch_exon() in bam_cigar.py had a long-standing bug inherited from the original RSeQC: soft clip CIGAR operations incorrectly advanced the reference coordinate, shifting exon boundaries rightward for reads with leading soft clips. The sibling function fetch_intron() already handled this correctly. This affects output from any script that resolves read-to-genome exon coordinates: bam2wig, read_distribution, read_duplication, RPKM_saturation, FPKM_count, and inner_distance.

Should I re-run my analyses? Almost certainly not. The shifts are small (1–10 bp, matching the soft clip length) and only affect the ~5% of reads with leading soft clips. Gene-level counts, coverage profiles, and quality metrics change by amounts well below meaningful thresholds. See CHANGES.md for a per-script impact table.

Contributing

Contributions are welcome!

Development Commands

# Install development dependencies
uv sync

# Run tests
uv run pytest

# Run full test matrix
uv run nox

# Lint and format
uv run ruff check .
uv run ruff format .

# Type check
uv run mypy rseqc/

License

Distributed under the terms of the GPLv3+ license, rseqc-redux is free and open source software.

Issues

If you encounter any problems, please file an issue with a description of the problem, steps to reproduce, and relevant error messages.

Credits

Original RSeQC by Liguo Wang — rseqc.sourceforge.net

Modernization by Nick Semenkovich (@semenko).