RIsearch for Python

RIsearch predicts RNA and DNA interactions: RNA-RNA, DNA-DNA and RNA-DNA hybrids. You give it query sequences, such as miRNAs or siRNAs, and a set of targets, such as a transcriptome. It reports every duplex whose predicted free energy is at or below a threshold, with coordinates, strand, energy and the pairing pattern, as a Polars DataFrame.

This package wraps the same Rust library as the risearch command-line tool, a rewrite of RIsearch2. It is a pre-release: options and output may still change before 3.0.0.

Installation

pip install risearch

or uv add risearch. Wheels are built for Linux x86_64 (manylinux) and macOS on Apple Silicon, for Python 3.10 and newer. On other platforms pip builds from the source distribution, which needs Rust 1.88 or newer.

On older x86-64 CPUs, or x86-64 Python running under Rosetta, install the compatibility Polars runtime; the module is still imported as polars:

pip install "risearch[lts-cpu]"

The package does not install the risearch command-line program; its releases ship prebuilt binaries.

Quick start

import risearch

risearch.index("targets.fa", "targets.idx")  # build once
targets = risearch.TargetRegistry.open("targets.idx")  # reuse for every search

hits = risearch.search("mirnas.fa", targets)

hits has one row per interaction. With the example in the repository, two human miRNAs against the 19 transcripts of RHOC in tests/data:

>>> hits.sort("energy", "target_name", "t_start").select(
...     "query_name", "target_name", "t_start", "t_end", "strand", "energy"
... ).head(3)
shape: (3, 6)
┌────────────────────────────┬─────────────────────────────────┬─────────┬───────┬────────┬──────────┐
│ query_name                 ┆ target_name                     ┆ t_start ┆ t_end ┆ strand ┆ energy   │
│ ---                        ┆ ---                             ┆ ---     ┆ ---   ┆ ---    ┆ ---      │
│ str                        ┆ str                             ┆ u64     ┆ u64   ┆ str    ┆ f64      │
╞════════════════════════════╪═════════════════════════════════╪═════════╪═══════╪════════╪══════════╡
│ hsa-miR-24-3p MIMAT0000080 ┆ ENSG00000155366|ENST00000527563 ┆ 481     ┆ 505   ┆ +      ┆ -25.3257 │
│ hsa-miR-24-3p MIMAT0000080 ┆ ENSG00000155366|ENST00000285735 ┆ 305     ┆ 322   ┆ +      ┆ -23.0914 │
│ hsa-miR-24-3p MIMAT0000080 ┆ ENSG00000155366|ENST00000285735 ┆ 1373    ┆ 1398  ┆ -      ┆ -21.7137 │
└────────────────────────────┴─────────────────────────────────┴─────────┴───────┴────────┴──────────┘

Coordinates are 0-based and inclusive, unlike the 1-based CLI output. The row order can change from run to run, so sort when you need a stable order.

The documentation covers every function and option, the result columns, the energy models and logging.

Citation

If you use RIsearch, please cite:

Roncelli S, Favaro L, Anthon C, Gorodkin J. RIsearch and RIOT: An integrated, high-performance framework for RNA-RNA interaction and siRNA off-target prediction. Bioinformatics.

License

Business Source License 1.1.