> ## Documentation Index
> Fetch the complete documentation index at: https://proto.evodesign.org/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# CRISPRtracrRNA
> [CRISPRtracrRNA](https://github.com/BackofenLab/CRISPRtracrRNA) is a multi-evidence pipeline from the [Bioinformatics Group at the University of Freiburg](https://www.bioinf.uni-freiburg.de/) that detects [tracrRNA](https://en.wikipedia.org/wiki/Trans-activating_crRNA) candidates in nucleotide [CRISPR](https://en.wikipedia.org/wiki/CRISPR) loci. It combines covariance-model search, CRISPR array detection, Cas-effector cassette detection, anti-repeat similarity, RNA-RNA interaction prediction, and transcription-terminator detection into a single weighted ranking score per candidate.
This toolkit is open source. Any third-party models, product names, or trademarks referenced are the property of their respective owners, and Proto is not affiliated with them.
```bibtex theme={null}
@article{mitrofanov2022crisprtracrna,
title={CRISPRtracrRNA: robust approach for CRISPR tracrRNA detection},
author={Mitrofanov, Alexander and Ziemann, Marcus and Alkhnbashi, Omer S and Hess, Wolfgang R and Backofen, Rolf},
journal={Bioinformatics},
volume={38},
number={Supplement\_2},
pages={ii42--ii48},
year={2022},
publisher={Oxford University Press},
doi={10.1093/bioinformatics/btac466}
}
```
| Function | Description | |
| ------------------------ | ------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| `run_crispr_tracr_rna()` | Predict tracrRNA sequences from nucleotide CRISPR loci | Docs Source |
**License:** CRISPRtracrRNA's own code is licensed under MIT, and it federates over bundled data sources and components, each under its own license terms.
Bundled dependencies, each under its own license:
* [CRISPRcasIdentifier](https://github.com/BackofenLab/CRISPRcasIdentifier/blob/master/LICENSE.txt): GPL-3.0
Review each source's terms before commercial use or redistribution.
## Background
[CRISPRtracrRNA](https://github.com/BackofenLab/CRISPRtracrRNA) ([Mitrofanov et al., 2022](https://doi.org/10.1093/bioinformatics/btac466)) detects trans-activating CRISPR RNA (tracrRNA) sequences in nucleotide CRISPR loci. A tracrRNA is a small non-coding RNA that base-pairs with the precursor crRNA in the [Class 2](https://en.wikipedia.org/wiki/CRISPR) effector systems that depend on one, namely [Type II](https://en.wikipedia.org/wiki/Cas9) (Cas9) and the tracrRNA-bearing [Type V](https://en.wikipedia.org/wiki/Cas12a) subtypes such as Cas12b, Cas12c, and Cas12e, and the resulting RNA duplex licenses Cas-mediated cleavage of target DNA. It is also the second component fused into the [single-guide RNA](https://en.wikipedia.org/wiki/Guide_RNA) used in modern genome editing. Because tracrRNAs share little primary-sequence conservation across families, single-model approaches such as [Infernal](https://github.com/EddyRivasLab/infernal) covariance-model search alone miss divergent tracrRNAs in newly sequenced and metagenomic genomes.
Internally, the pipeline runs an array-detection step with [CRISPRidentify](https://github.com/BackofenLab/CRISPRidentify) (machine learning), a Cas-cassette step with [CRISPRcasIdentifier](https://github.com/BackofenLab/CRISPRcasIdentifier) (HMM and machine learning), a tracrRNA candidate scan with Infernal `cmsearch` against curated covariance models, an anti-repeat alignment step using fasta36, vmatch, Clustal Omega, and BLAST, an RNA-RNA interaction step with [IntaRNA](https://github.com/BackofenLab/IntaRNA), and a transcription-terminator step with erpin. A final ranking step combines the per-candidate features into a single weighted score, and a faster `model_run` mode performs only the covariance-model scan and skips the validation evidence and the ranking step.
### Learning Resources
* [BackofenLab/CRISPRtracrRNA](https://github.com/BackofenLab/CRISPRtracrRNA) (Bioinformatics Group Freiburg) - official repository with installation instructions, the canonical configuration surface, and the curated covariance models distributed with the tool.
* [EddyRivasLab/infernal](https://github.com/EddyRivasLab/infernal) (The Eddy/Rivas Laboratory, Harvard) - official repository and User's Guide for the covariance-model search engine and the `cmsearch` E-value statistics that score tracrRNA candidates.
* [BackofenLab/IntaRNA](https://github.com/BackofenLab/IntaRNA) (Bioinformatics Group Freiburg) - official repository for the RNA-RNA interaction predictor that scores the anti-repeat to repeat duplex.
## Tools
### CRISPRtracrRNA Prediction (`crispr-tracr-rna`)
Predicts tracrRNA candidates from one or more nucleotide sequences and returns, per input sequence, a list of `CrisprTracrRNAPrediction` rows sorted by ranking score. Each row carries the candidate position and sequence, CRISPR array context, anti-repeat similarity and coverage, predicted RNA-RNA interaction with the repeat, terminator location and score, distance to the nearest Cas-effector cassette, and a single weighted multi-evidence score.
#### API Reference
Source
Nucleotide sequence(s) to predict tracrRNA from. Each sequence should contain a CRISPR locus. Labeled positionally (`seq_0`, `seq_1`, ...); results are returned in input order.
Source
CRISPR model type.
Available options: `II`, `all`
Pipeline mode.
Available options: `complete_run`, `model_run`
Parallel workers across input sequences (defaults to 1).
Minimum anti-repeat ↔ repeat similarity (0-1).
Minimum anti-repeat alignment coverage (0-1).
Ranking weight for CRISPR array confidence.
Ranking weight for anti-repeat similarity.
Ranking weight for anti-repeat coverage.
Ranking weight for similarity x coverage.
Ranking weight for IntaRNA interaction energy.
Ranking weight for the covariance-model tail hit.
Ranking weight for erpin terminator score.
Ranking weight for orientation consistency.
Ranking weight for anti-repeat ↔ tail consistency.
Ranking weight for tail ↔ terminator consistency.
Type V (Cas12) anti-repeat search.
Verbosity level (0=quiet, 1=info, 2=debug, 3=raw subprocess stderr). `True` is coerced to `1` and `False` to `0`.
Device to run the tool on.
Maximum execution time in seconds. `None` waits indefinitely.
Random seed. When set, tools run reproducibly up to small GPU float noise (see `BaseToolOutput.approx_equal`), and the seed participates in cache keys. When None, cacheable seed-sensitive tools skip cache until seeded.
Source
One result per input sequence, each carrying all candidate hits upstream produced for that sequence (top-ranked first).
ID of the input sequence.
All candidate hits for this sequence, top-ranked first; empty when upstream found nothing.
#### Applications
Use this to confirm and characterize Type II and Type V CRISPR-Cas loci, since a detected tracrRNA is the component that completes a functional Class 2 locus and distinguishes a Cas9 or Cas12 system from an unaccompanied CRISPR array. Pair it with [`minced`](https://bio-pro.mintlify.app/tools/gene-annotation/minced) on a confirmed array to recover the crRNA spacers, then design a single-guide RNA by fusing a spacer-bearing crRNA with the detected tracrRNA scaffold for genome-editing experiments. Run it across metagenomes and uncultured genomes to discover novel Cas9 or Cas12 systems whose tracrRNAs are too divergent to be caught by covariance-model search alone.
#### Usage Tips
* **Provide each CRISPR locus with at least 5 kb of flanking sequence on either side.** The multi-evidence pipeline needs adjacent context to locate the Cas cassette and the downstream transcription terminator. Loci submitted as narrow windows lose those evidence channels and fall back to a covariance-model-only score.
* **`model_type` defaults to `"II"`, which only screens for Cas9 systems.** To also screen tracr-bearing Type V (Cas12b, Cas12c, Cas12e, ...) loci, set `model_type="all"` and `perform_type_v_anti_repeat_analysis=True`. The Type V path is off by default because it is slower and irrelevant when only Cas9 loci are of interest.
* **Type I and Type III CRISPR systems do not use a tracrRNA.** A `complete_run` on such a locus returns array context and Cas annotations but empty tracrRNA fields, with the ranking score reflecting only the partial evidence.
* **The ten `weight_*` ranking parameters interact.** Sweep them together against a held-out positive and negative set rather than tuning a single weight in isolation, and keep upstream's documented defaults when there is no specific objective to optimize for.
* **`run_type="model_run"` is the high-throughput pre-filter, not the final answer.** It runs only the Infernal `cmsearch` step and returns candidates with E-values but none of the array, interaction, or terminator evidence, so re-run promising candidates through `complete_run` before drawing conclusions.
## Toolkit Notes
These apply to every CRISPRtracrRNA tool in this toolkit (`crispr-tracr-rna`).
* **Runs on CPU only.** The pipeline drives Infernal, IntaRNA, fasta36, vmatch, Clustal Omega, BLAST, and erpin, all CPU-based programs. There is no GPU acceleration to enable.
* **Initial install pulls model archives from Google Drive.** `complete_run` mode requires the CRISPRcasIdentifier ML and HMM archives, which the standalone install fetches once. Google Drive rate-limits anonymous fetches, so on a failed install retry after a minute or follow the upstream README to place the two archives in the CRISPRcasIdentifier directory by hand. After install the runtime needs no further network access.
* **`num_workers` parallelizes across input sequences, not within a sequence.** Each worker runs the full pipeline in its own working directory to avoid file-name collisions between concurrent jobs. The default of 1 is single-process; set it explicitly when batch-scanning many loci. The wrapper caps the effective worker count at `len(sequences)`, so over-provisioning is safe.
**Example notebook:** See the [full working example](https://github.com/evo-design/proto-tools/blob/47e34afa5ea240a3b406e323dc38aa5dc85f223e/proto_tools/tools/gene_annotation/crispr_tracr_rna/examples/example.ipynb) for a copy-paste-ready walkthrough.
## Infrastructure Guides
The following guides cover how to run tools efficiently and at scale.
Keep a tool's model warm across calls instead of reloading it every invocation.How GPUs are allocated to tools and how to target specific devices.Fan a batch of inputs out across multiple GPUs.
evo-design/proto-tools/proto\_tools/tools/gene\_annotation/crispr\_tracr\_rna
