RIsearch1 (tauso fork)

RIsearch1: RNA–RNA, RNA-DNA and DNA-DNA interaction prediction using a simplified nearest-neighbor energy model.

This fork carries RIsearch1 only. RIsearch2 and the siRNA off-target pipeline live upstream at RTH-tools/risearch; they were removed here because nothing downstream of this fork uses them. Their history is still in this repository if you need it.

Main changes of this fork

The fork is actively maintained, and focusing on the RISearch1 executable, especially the standard linSpace path. The main goals of the fork are:

• Improve correctness and bugfixes (SIGSEGV, memory leaks, hangs etc)
• Improve readability (DRY, functions, classes, naming etc)
• Convert to lightweight C++
• Add tests, both logical and regression
• MAIN GOAL: improve speed with known available techniques.

As a former C++ dev that does not have experience with SIMD, I focused on testing, and having readable code so I understand it. After that, we cumulatively improved about ~2.6x the performance from upstream fork while being bit identical. All code except SIMD was reviewed, and SIMD will be reviewed and tested in the following days.

Python package

A precompiled binary from this fork is published to PyPI as risearch-tauso, so RIsearch is a normal Python dependency for tauso and any other downstream tool:

pip install risearch-tauso

import risearch_tauso, subprocess
subprocess.run([risearch_tauso.executable_path(), "-q", "query.fa", "-t", "target.fa"])

Or as a CLI shim, which forwards all arguments straight to the bundled binary:

risearch-tauso -q query.fa -t target.fa
python -m risearch_tauso -q query.fa -t target.fa

The PyPI package is not canonical upstream RIsearch — it is the tauso-team fork. Use upstream if you want the unmodified tool.

Differences from upstream

Bugs fixed relative to upstream are recorded in BUGFIXES.md.

Building from source

Requires a C++17 compiler and CMake 3.20 or newer.

cmake -S RIsearch1 -B RIsearch1/build -DCMAKE_BUILD_TYPE=Release
cmake --build RIsearch1/build -j

That produces RIsearch1/bin/RIsearch and RIsearch1/bin/RIsearch.dbg, the second with the debug tracing compiled in. The wheel build runs the same CMake via setup.py.

Tests

./RIsearch1/build/tests/risearch1_tests

Unit tests for the nucleotide coding, the min/max helpers, the alignment symbols, the energy matrix and the FASTA reader, including death tests for the inputs that are refused; end-to-end tests that run main() in process with a constructed argv and assert on its output; and throughput benchmarks under Performance.*. Pass -DRISEARCH_BUILD_TESTS=OFF to skip googletest entirely, as the wheel build does.

Running

RIsearch -q query.fa -t target.fa

Both files may hold several sequences; RIsearch scans all against all. Single sequences can be given directly with -Q acgu -T acgu. See RIsearch1/Manual.pdf for the full option list.

Copyright

Copyright 2021 by the contributors; see RIsearch1/README.

RIsearch1 is released under the GNU General Public License version 3. This is free software: you can redistribute it and/or modify it under the terms of that licence, either version 3 or (at your option) any later version. You should have received a copy of the GNU General Public License along with RIsearch — see the file COPYING. If not, see http://www.gnu.org/licenses/.

This software is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

Citation

If you use RIsearch in a publication, please cite:

RIsearch: fast RNA-RNA interaction search using a simplified nearest-neighbor energy model. Wenzel A, Akbasli E, Gorodkin J. Bioinformatics. 2012 Nov 1;28(21):2738-46.

Contact

For problems with this fork, open an issue here. For upstream RIsearch: software+crispron@rth.dk