RNA Conformation Viewer

RCV is a free desktop application for predicting, exploring and presenting the conformations of an RNA. It combines the rnafold3d prediction engine with an interactive 3D/2D viewer, and runs entirely on your own computer.

Download RCV 0.5.0  Get started

Main features

• Prediction from sequence. Paste a sequence or a multi-sequence FASTA file; RCV returns a ranked, validated all-atom ensemble.
• Open your own structures. NMR ensembles, MD snapshots, PDB/mmCIF files from any predictor.
• Pull to switch. Drag a residue and the structure moves to the nearest stable computed conformation.
• Linked 2D and 3D. The secondary structure follows the 3D view, with shared colours, hover and selection.
• Switch analysis. Boltzmann populations of 2D states and a point-mutation scan.
• Command console. A selection language for RNA ensembles and familiar molecular-viewer commands.
• Publication output. Figures, figure panels, transition videos, interactive HTML pages and methods text.
• One file per RNA. The open .rcv format stores all conformations, 10–14× smaller than PDB.

Download

RCV is free and open-source software, released under the GNU General Public License v3.0. The desktop installers include Python, the prediction engine, ViennaRNA and the demo data; nothing else needs to be installed.

RCV 0.5.0 (23 October 2026)

Older versions are available on the GitHub releases page. Earlier releases were named RNAconfSwitch (0.4) and RNAconformer Viewer (0.3); their settings, plugins and projects are read by RCV 0.5.

System requirements

License

RCV is distributed under the GPL-3.0 license. ViennaRNA is a separate package under its own license. three.js (MIT) is included in the viewer.

Install

Linux and installation from source

Requirements: Python 3.10 or later with numpy and scipy. ViennaRNA is needed for prediction from sequence and for the switch analysis.

git clone https://github.com/PrinceWang2018/RNAconformer.git
cd RNAconformer
pip install -e .                        # installs the rcv and rnafold3d commands
conda install -c bioconda viennarna     # or: pip install ViennaRNA
pip install pywebview                   # optional: native window

Check the installation

rcv --demo        # opens RCV with the HIV-1 TAR NMR ensemble and two predictions
rcv               # opens the start page

Without pywebview, RCV opens in an Edge or Chrome application window. Useful options: --port 8765, --projects-dir DIR, --shell auto|webview|window|browser.

Where RCV keeps its files

Predictions are saved in Documents/RCV/projects by default (View › Preferences).

Uninstall

Windows: Settings › Apps › RCV › Uninstall. macOS: move RCV from Applications to the Trash. Source install: pip uninstall rna-conformation-viewer. Your projects and settings are kept unless you delete the folders listed above.

Tutorial

This tutorial covers the main tasks in RCV. It takes about 20 minutes, most of which is waiting for a prediction. For every menu, option and command, see the full user guide.

Contents: Start page · Predict · The window · Pull to switch · Switch analysis · Console · Export · Shortcuts

The start page

When RCV starts, it shows the start page. Paste a sequence to predict its conformations, or click Open files… to open structures you already have (.rcv files, PDB/mmCIF, multi-model NMR files, prediction folders). Fetch PDB… downloads an entry from the PDB. Recent predictions are listed below. To look around first, click Demo ensembles.

Predict from a sequence

1. Paste an RNA sequence (plain text or FASTA; T is read as U). Leave the box empty to run the demo sequence (HIV-1 TAR, 29 nt).
2. Choose the number of conformations, the sampling depth (Fast, Standard or Thorough) and the CPU cores. RCV shows an estimated run time.
3. Optional, under Advanced: a known secondary structure in dot-bracket notation and the random seed.
4. Click Predict. The progress view shows each stage, the time left and notes on what the engine is doing.

A multi-record FASTA file starts a batch (up to 100 sequences). When the batch ends, choose which result to open; the others are saved as .rcv files with a summary table.

Typical run times on an 8-core laptop: a 12–14 nt hairpin with Fast sampling, 10–60 s; a 29 nt hairpin with Standard sampling, 5–7 minutes.

Explore the ensemble

The conformations are listed on the left, ordered by energy. Hover a conformation to preview it; click it to morph the structure. The timeline below the 3D view plays through all conformations. The right panel shows the energy profile and terms (Energy), the switch analysis (Switch) and full details (Details).

• Mouse: drag to rotate, right-drag to pan, wheel to zoom, double-click to centre, right-click a residue for its menu.
• The sequence bar shows the sequence and dot-bracket of the current conformation; changed pairs are underlined.
• The 2D window (D) draws the secondary structure in the same colours as the 3D model and morphs with it.

Pull to switch

Press G (or hold Alt) and drag a residue. The structure stretches, and the computed conformation that best matches the pull is shown as an orange preview with its energy change. Release to morph to it. Explore new conformations runs the engine with the pull as a temporary bias and adds the new stable conformations to the ensemble.

Switch analysis

The Switch tab lists the distinct secondary structures in the ensemble with their free energy and Boltzmann population at a chosen temperature (0–95 °C).

Choose two states, A and B, to scan all point mutations. The heat map shows how each mutation shifts the balance: blue favours A, orange favours B. Click a cell to select the residue, copy the mutant sequence, or Predict in 3D.

This analysis uses ViennaRNA at the secondary-structure level. Populations are estimates, not experimental values.

Command console

Open the console with `. Type one command per line; Tab completes and help lists all commands. Some examples:

select loop, resi 23-27 and unpaired
color orange, loop
show sticks, switching
distance d1, 22/N1, 27/N3
orient helix H2
png 3000, 2000, 600

Commands can be saved in a script file (.rcvs) and run with run script.rcvs. Every command is stored in the .rcv file as a reproducible record.

Save and export

Output	How
.rcv file	Ctrl+S. All conformations, energies, 2D structures and your current view in one file.
Image	Ctrl+E. PNG, TIFF or PDF, 300–1200 dpi, optional legend and energy panel.
Figure panels	Export › Figure panels. Several conformations in the same orientation, each with its 2D structure.
Video	Export › Video or Transition A → B. MP4, WebM or PNG frames, 720p to 4K.
Interactive HTML	Export › Interactive HTML page. One file that opens in any browser, for supplements and reviewers.
Methods text	Export › Methods text. A methods paragraph and citation generated from the ensemble record.

Keyboard shortcuts

G pull	D 2D window	M measure	` console
←/→ previous/next	Space play	1–5 representation	C colour scheme
R reset view	S snapshot	Ctrl+O open	? help

Command line

RCV also works without the window:

rcv batch my_rnas.fasta                 # predict every sequence, one .rcv each
rcv convert models/*.pdb -o my_rna.rcv  # structures to one .rcv file
rcv info my_rna.rcv --methods           # contents and methods paragraph
rcv extract my_rna.rcv -o pdb/          # back to PDB files
rcv html my_rna.rcv -o my_rna.html      # interactive web page

Plugins

RCV can be extended with two kinds of plugins. Place the plugin file in the plugins folder of the RCV user data folder and restart RCV. Plugins appear in the Plugins menu, as tabs in the right panel, or as new metrics in the energy panel.

Available plugins

Have you written a plugin that others could use? Send it to us and we will list it here (see Co-creation Program).

Writing an interface plugin (JavaScript)

An interface plugin is an ES module whose default export has an activate(api) function. The API gives read-only access to the current ensemble and selection, events (ensemble, conformation, morph, selection, …), panels, menu items, overlays, colour schemes and metrics.

// my-plugin.js
export default {
  id: 'my-plugin',
  name: 'My plugin',
  version: '1.0',
  description: 'Short description shown in the Plugins menu.',
  activate(api) {
    api.metrics.register({
      key: 'n_residues', label: 'Residues', unit: '', lower_is_better: null,
      description: 'Number of residues (example).',
      compute: (ens, k) => ens.residues.length,
    });
  },
};

Writing an analysis plugin (Python)

An analysis plugin defines a PLUGIN dictionary and a compute(sequence, residues) function that returns one value per metric. residues is a list (5′ to 3′) of {atom_name: xyz} dictionaries after superposition.

# end_distance.py
import numpy as np

PLUGIN = {
    'id': 'end_distance',
    'name': 'End distance',
    'metrics': [{'key': 'end_to_end', 'label': "5'-3' distance", 'unit': 'Å',
                 'lower_is_better': None, 'description': "C4'-C4' distance of the chain ends."}],
}

def compute(sequence, residues):
    a, b = residues[0].get("C4'"), residues[-1].get("C4'")
    if a is None or b is None:
        return {'end_to_end': None}
    return {'end_to_end': round(float(np.linalg.norm(a - b)), 2)}

The plugin API is version 1 and unchanged since RNAconfSwitch 0.4. The built-in plugins in src/rcv/static/plugins are complete examples.

Release Notes

RCV 0.5.0 — 23 October 2026

Changed

• Renamed to RCV (RNA Conformation Viewer) with a new logo. Command rcv, user folders RCV. Settings, plugins and projects of RNAconfSwitch 0.4 are still read.
• Ctrl+S now saves an .rcv file; the JSON session is now called a workspace (Ctrl+Shift+S).

New

• .rcv file format: all conformations of one RNA with energies, 2D structures, validation, provenance and the viewer state, 10–14× smaller than PDB. Command-line tools rcv convert, rcv info, rcv extract.
• 2D structure window linked to the 3D view, morphing in step with it.
• Transition video A → B, directly or through intermediate conformations.
• Batch prediction from a multi-sequence FASTA file (rcv batch).
• Command console with an RNA selection language and about 45 commands; script files.
• Switch analysis: 2D-state populations and point-mutation scan.
• Figure panels, interactive HTML export and methods text.

rnafold3d engine 2.0.1 — 22 October 2026

• Valid all-atom ensembles from sequence alone for 26/26 benchmark targets (14–192 nt).
• About 2–4 Å accuracy for hairpins and small motifs. Large multi-branched RNAs are not yet predicted accurately (no tertiary-contact term).

Earlier versions

• 0.4 — released as RNAconfSwitch.
• 0.3 — released as RNAconformer Viewer.

Full history: CHANGELOG.md.

Statistics

Usage of this website and downloads of RCV. The numbers are updated once a day.

Downloads by platform

Downloads by release

Last updated: –. Visits are counted with GoatCounter, which uses no cookies and stores no personal data; a visit is one visitor per page per day, and countries are derived from the visitor’s network address, which is not stored. Downloads are the installer downloads from GitHub releases; copies of the source code are not counted.

Co-creation Program

RCV is developed together with the people who use it. The Co-creation Program is open to research groups, students, educators and developers who want to help shape the software. Participation is free.

Ways to take part

What participants receive

• Early access to new versions and direct contact with the developer.
• Influence on the development plan; requests from participants are considered first.
• Acknowledgement in the release notes and on this website.
• For substantial scientific contributions, co-authorship on related publications may be discussed.

How to join

1. Send an email to princewang2018@gmail.com with the subject “RCV Co-creation Program”.
2. Tell us briefly who you are, what you work on, and how you would like to take part.
3. We will reply within a few working days.

You can also start directly by opening an issue or a pull request on GitHub.

About

RCV (RNA Conformation Viewer) predicts, explores and presents the conformations of an RNA. Many RNAs do not have a single structure but switch between several; RCV is built to look at these ensembles as a whole rather than at one model at a time.

How it works

The rnafold3d engine is physics-inspired and does not use deep learning:

sequence ─► ViennaRNA secondary-structure hypotheses (MFE + distinct suboptimals)
         ─► rigid all-atom A-form stems (calibrated geometry)
         ─► fold-tree search in torsion space (simulated annealing)
         ─► loop closure and clash relief
         ─► independent all-atom validation
         ─► ranked, structurally distinct ensemble

Predictions run on your CPU cores and the 3D view is rendered on your GPU. No data leave your computer. Runs are deterministic for a given seed and version.

Limitations

• Accuracy is good for hairpins and small motifs (about 2–4 Å) but limited for large multi-branched RNAs.
• Switch analysis is at the secondary-structure level; 3D energies are model scores, not free energies.
• Molecular surfaces, non-canonical base-pair annotation and multi-chain complexes are not yet supported.

Benchmark details: engine status and benchmark.

How to cite

If you use RCV in your work, please cite:

Wang Z. RCV: RNA Conformation Viewer, version 0.5.0 (2026).
https://github.com/PrinceWang2018/RNAconformer

Please also cite ViennaRNA when you use predictions from sequence:
Lorenz R. et al. ViennaRNA Package 2.0. Algorithms for Molecular Biology 6, 26 (2011).

The methods text generated by RCV (Export › Methods text) includes the exact versions and parameters of each run.

Developer

RCV is developed by Zixiang Wang.

Acknowledgements

RCV uses ViennaRNA, NumPy, SciPy, three.js and pywebview. Benchmark structures are from the RCSB Protein Data Bank.

Contact

Reporting a problem

Please include:

• the RCV version (Help › About) and your operating system;
• what you did, what you expected, and what happened;
• the log file from the RCV user data folder (see Install);
• if possible, the sequence or structure file that shows the problem.