TxNova

TxNova — short for Transcript Nova — recovers unannotated spliced residual loci from existing bulk RNA-seq BAMs. Cohort-recurrent residual splices become locus models. Class, counts, junctions, and bridges are recomputed from the BAM and the annotation. A control-versus-treat filter is optional.

Docs: Read the Docs.

What you need

• Python 3.10+
• Coordinate-sorted, indexed BAMs from STAR or HISAT2 (≥2 samples; control/treat optional)
• Genome FASTA + .fai, and a comprehensive gene GTF (mouse GENCODE M39 / GRCm39, or human GENCODE / GRCh38). species: auto infers mouse or human; you can set species: mouse or human. Only those two are supported.
• A YAML config and a sample sheet (txnova init writes starters)

Install

pip install txnova

Wheels cover Linux and macOS. On Windows use WSL2. Building from source needs a Rust toolchain; see Installation.

First run

txnova init -c config.yaml --samples samples.tsv
# edit paths, strandedness (rf / fr / unstranded), and sample rows
txnova preflight -c config.yaml
txnova run -c config.yaml

What to expect under output_dir:

• candidates/candidates.tsv — structure-pass residuals; treat-detected / control-silent when a contrast exists
• candidates/candidates.unnamed.tsv — structure-pass, also in control by TPM
• candidates/candidates.shared.tsv — structure-pass, splice in both groups
• report/report.html — start here
• quantify/ — full-universe counts, TPM, and DE

The three candidate tables share the structural gates. They answer different questions (induction vs unannotated structure in both groups). They are not a declaration of new genes.

Next: Preparing input BAMs · Quickstart · Output reference · FAQ

Status

0.1.x. Pin txnova==0.1.7 in Methods. Changelog.

Citation

For the software, cite the Zenodo DOI above. Pin the installed version in Methods (this tree is txnova==0.1.7). See CITATION.cff.

Li, Z. TxNova. Zenodo. doi:10.5281/zenodo.21970482

PyPI: https://pypi.org/project/txnova/.

License

Software: Apache License 2.0. The packaged hexamer tables come from CPAT (Wang et al. 2013); see docs/license.

Author

Zhao Li (李钊)
Email: leelieber@gmail.com