RECOVAR

Tools for cryo-EM heterogeneity analysis

Get Started Web GUI

---

Why RECOVAR?

• Highest resolution on CryoBench across multiple datasets
• Conformational density and free-energy landscape estimation
• Cryo-ET support for tilt-series heterogeneity analysis
• Web GUI with interactive latent-space exploration and sub-particle selection
• Direct input from RELION (.star) and cryoSPARC (.cs) -- no format conversion
• No hallucinations -- kernel regression produces transparent, verifiable volumes

What you need before starting

RECOVAR starts after consensus refinement in RELION or cryoSPARC. You need:

1. A particle stack with poses and CTF — a RELION .star file or cryoSPARC .cs file
2. A solvent mask (.mrc) — or let RECOVAR generate one automatically
3. An NVIDIA GPU (any Volta or newer — V100, RTX 20/30/40-series, A100, H100)

RECOVAR outputs: mean reconstruction, variance maps, eigenvolumes, latent coordinates, k-means cluster volumes, UMAP embeddings, and trajectories — all exportable back to RELION/cryoSPARC.

---

Example output

Eigenvalue spectrum	UMAP latent space (EMPIAR-10076)

Inspect results directly in the browser -- 3D volume viewer with adjustable isosurface threshold:

---

Typical workflow

The easiest way to use RECOVAR is through the Web GUI -- launch it with recovar gui, then create jobs, explore the latent space, and generate volumes all from your browser.

Or use the command line:

# 1. Run the pipeline (~10 min for a small dataset)
recovar pipeline particles.star -o output --mask mask.mrc

# 2. Analyze results (k-means, trajectories, UMAP)
recovar analyze output --zdim=10

# 3. Explore interactively in the browser
recovar gui

---

How it works

Particles Mean reconstruction 3D Covariance PCA Embedding Volumes

RECOVAR estimates a regularized 3D covariance from your particle images, extracts principal components to build a low-dimensional latent space, and uses kernel regression to generate volumes at any point in that space.

For the full method, see the paper or recorded talk.

---

Citing RECOVAR

If you use RECOVAR in your research, please cite:

Citation

Gilles, M.A. & Singer, A. "Cryo-EM heterogeneity analysis using regularized covariance estimation and kernel regression." PNAS (2025). DOI: 10.1073/pnas.2419140122

@article{gilles2025recovar,
  title={Cryo-EM heterogeneity analysis using regularized covariance estimation and kernel regression},
  author={Gilles, Marc Aur{\`e}le and Singer, Amit},
  journal={Proceedings of the National Academy of Sciences},
  year={2025},
  doi={10.1073/pnas.2419140122}
}