ncRNA Search (bioSkills)

A Claude Code skill that runs Infernal covariance-model searches against Rfam properly — clan de-overlapping, curated bit-score cutoffs, and the E-value-depends-on-database-size trap handled rather than ignored.


Type	Claude Skill
Supplier	GPTomics bioSkills (community OSS, MIT)
Availability	GA — part of the bioSkills collection
Pricing	Free / OSS (MIT) — Infernal (BSD) and the Rfam database (CC0) are downloaded separately
Capabilities	Read/Write — Claude runs the skill’s workflow locally (Bash/Python), not as an MCP tool
Verified	works · 2026-08-13
Security	cleared · 2026-08-13 — GPTomics/bioSkills MIT, no external credentials

How to install

bioSkills is not an npm package — skills are plain markdown/code read directly by the agent. Clone the repo, then either run the installer for the whole category or copy the single skill directory.

Claude Code — clone and install via the bundled script:
```
git clone https://github.com/GPTomics/bioSkills
cd bioSkills
./install-claude.sh --categories "rna-structure"
```
The installer copies matching skills into ~/.claude/skills/ (default target). Use ./install-claude.sh --list to preview the skills first.
Claude Code / other agents — copy just this one skill:
```
cp -r bioSkills/rna-structure/ncrna-search ~/.claude/skills/
```
(run from inside your clone — the previous step left you in bioSkills/; otherwise replace bioSkills/ with the absolute path of your clone, e.g. /Users/you/repos/bioSkills).
Prerequisites — Infernal 1.1.4+ plus the Python helpers the skill’s examples use:
```
conda install -c conda-forge -c bioconda infernal
pip install "biopython>=1.83" "pandas>=2.2"
```
bioconda ships infernal 1.1.5 (checked 2026-08-08). Confirm with cmscan -h.

Rfam database — download and index it once (this is the skill’s own documented setup step):

wget https://ftp.ebi.ac.uk/pub/databases/Rfam/CURRENT/Rfam.cm.gz && gunzip Rfam.cm.gz
wget https://ftp.ebi.ac.uk/pub/databases/Rfam/CURRENT/Rfam.clanin
cmpress Rfam.cm

Rfam.cm ships pre-calibrated — run cmpress on it, and do not run cmcalibrate.

What it does

Search — cmscan for one sequence against the whole CM database, cmsearch for one covariance model against a sequence database; both score sequence and consensus secondary structure jointly, which is what a covariance model buys over an HMM.
De-overlap — resolves the redundant hits Rfam clans produce, using --fmt 2 --clanin Rfam.clanin and filtering the marked lines (grep -v ' = ').
Threshold choice — explains the three curated per-family cutoffs and when each applies: GA (gathering, the curated membership threshold and the default choice for Rfam), TC (trusted cutoff, the lowest known true-positive score, most conservative), NC (noise cutoff, the highest false-positive score, most permissive).
Structure recovery — cmalign to fold hits back against the model and recover the consensus secondary structure, which is the handoff into covariation testing or restrained folding.
Custom models — cmbuild → cmcalibrate → cmpress for a family not in Rfam, plus cmfetch to pull individual models out of the flatfile.
Specialized alternatives the skill routes to when a general CM search is the wrong tool: tRNAscan-SE 2.0 (tRNAs), barrnap and RNAmmer (rRNA), snoscan and snoReport 2.0 (snoRNAs), miRDeep2 (miRNAs).

Primary use cases: annotating structured ncRNAs in a new genome or contig set, assigning an unknown transcript to an Rfam family, recovering a consensus fold for a set of homologs.

Notes

The skill leads with a scoping rule that saves wasted runtime: a covariance model offers no advantage when there is little conserved secondary structure to exploit, and a CM built from a structure-free alignment (no real #=GC SS_cons pairs) collapses to an HMM — all the cost, none of the benefit. Long lncRNAs are the usual case where this bites.

Two reporting rules follow. First, a CM E-value scales linearly with the searched database size (Z), so the same hit gets a different E-value depending on what you searched; for reproducibility use --cut_ga or pin -Z <Mb> explicitly, and prefer bit scores (database-size independent) when working with a custom uncalibrated model. Second, a significant hit means the locus has sequence plus structure consistent with the family — it does not establish that the RNA is expressed, processed, or functional.

Upstream skill front-matter name is bio-rna-structure-ncrna-search; upstream directory rna-structure/ncrna-search. The natural next step after a hit is Covariation Analysis to test whether the recovered structure has evolutionary support, or ViennaRNA for thermodynamic folding of individual hits; Rfam is the same database as a hosted Claude connector if you want interactive family lookups without a local download, and RNA Structure Probing supplies the experimental complement.

Sources

Installed this tool?

Share feedback — install path, OS, errors, workarounds. The form opens with this tool pre-selected and a link back to this page.