CRISPRtracrRNA

CRISPRtracrRNA is a multi-evidence pipeline from the Bioinformatics Group at the University of Freiburg that detects tracrRNA candidates in nucleotide 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.

---

GitHub 10 GitHub 10 Publication Publication Cite Cite Tool Source Tool Source Open as Notebook Open as Notebook Open on Proto Open on Proto

BackofenLab/CRISPRtracrRNA

10 stars

View repo

CRISPRtracrRNA: robust approach for CRISPR tracrRNA detection

Alexander Mitrofanov, Marcus Ziemann, … Rolf Backofen

Bioinformatics (2022)

Read paper

@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}
}

Copy citation

proto-bio/proto-tools/proto_tools/tools/gene_annotation/crispr_tracr_rna

View source

Open Notebook

Open notebook

Coming soon!

Run this tool directly in Proto with no setup required.

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: GPL-3.0

Review each source’s terms before commercial use or redistribution.

Background

CRISPRtracrRNA (Mitrofanov et al., 2022) 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 effector systems that depend on one, namely Type II (Cas9) and the tracrRNA-bearing Type V 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 used in modern genome editing. Because tracrRNAs share little primary-sequence conservation across families, single-model approaches such as Infernal covariance-model search alone miss divergent tracrRNAs in newly sequenced and metagenomic genomes.

Internally, the pipeline runs an array-detection step with CRISPRidentify (machine learning), a Cas-cassette step with 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, 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 (Bioinformatics Group Freiburg) - official repository with installation instructions, the canonical configuration surface, and the curated covariance models distributed with the tool.
• 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 (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

Input: CrisprTracrRNAInput

sequences · List[string] · required

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

Config: CrisprTracrRNAConfig

model_type · enum · default: "II"

CRISPR model type.

Available options: II, all

run_type · enum · default: "complete_run"

Pipeline mode.

Available options: complete_run, model_run

num_workers · integer

Parallel workers across input sequences (defaults to 1).

anti_repeat_similarity_threshold · number · default: "0.7"

Minimum anti-repeat ↔ repeat similarity (0-1).

anti_repeat_coverage_threshold · number · default: "0.6"

Minimum anti-repeat alignment coverage (0-1).

weight_crispr_array_score · number · default: "0.5"

Ranking weight for CRISPR array confidence.

weight_anti_repeat_sim · number · default: "0.5"

Ranking weight for anti-repeat similarity.

weight_anti_repeat_coverage · number · default: "0.5"

Ranking weight for anti-repeat coverage.

weight_anti_sim_coverage · number · default: "0.5"

Ranking weight for similarity x coverage.

weight_interaction_score · number · default: "0.6"

Ranking weight for IntaRNA interaction energy.

weight_model_hit_score · number · default: "0.9"

Ranking weight for the covariance-model tail hit.

weight_terminator_hit_score · number · default: "0.9"

Ranking weight for erpin terminator score.

weight_consistency_orientation · number · default: "0.1"

Ranking weight for orientation consistency.

weight_consistency_anti_repeat_tail · number · default: "0.1"

Ranking weight for anti-repeat ↔ tail consistency.

weight_consistency_tail_terminator · number · default: "0.1"

Ranking weight for tail ↔ terminator consistency.

perform_type_v_anti_repeat_analysis · boolean · default: "False"

Type V (Cas12) anti-repeat search.

verbose · integer · default: "0"

Verbosity level (0=quiet, 1=info, 2=debug, 3=raw subprocess stderr). True is coerced to 1 and False to 0.

device · string · default: "cpu"

Device to run the tool on.

timeout · integer · default: "600"

Maximum execution time in seconds. None waits indefinitely.

seed · integer

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

Output: CrisprTracrRNAOutput

results · List[CrisprTracrRNASequenceResult]

One result per input sequence, each carrying all candidate hits upstream produced for that sequence (top-ranked first).

Show CrisprTracrRNASequenceResult

sequence_id · string · required

ID of the input sequence.

candidates · List[CrisprTracrRNAPrediction]

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 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 knobs 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 for a copy-paste-ready walkthrough.

Infrastructure Guides

The following guides cover how to run tools efficiently and at scale.

Tool Persistence

Keep a tool’s model warm across calls instead of reloading it every invocation.

Device Management

How GPUs are allocated to tools and how to target specific devices.

Parallel Execution

Fan a batch of inputs out across multiple GPUs.

Cloud Inference

Run tools on managed cloud infrastructure with no local setup.