RECOVAR: Cryo-EM Heterogeneity Analysis

Background

Cryogenic electron microscopy (cryo-EM) and cryo-electron tomography (cryo-ET) have revolutionized structural biology by enabling the visualization of complex biological macromolecules. However, biological molecules are rarely static; they exist in a continuous spectrum of dynamic states crucial to their function. Understanding this conformational heterogeneity is essential for mapping free energy landscapes, identifying metastable states, and modeling complex molecular motions. Consequently, there is a critical need for advanced computational methods capable of accurately extracting and analyzing these diverse 3D conformational states directly from noisy 2D projection images to fully comprehend molecular mechanisms.

Despite the importance of mapping molecular dynamics, current analytical approaches suffer from significant limitations. Traditional discrete 3D classification methods force continuous motions into rigid categories, failing to capture complex, nonlinear conformational changes. Conversely, modern neural network-based techniques rely on “black-box” architectures that lack mathematical interpretability and regularization. These deep-learning models are prone to “hallucinating” artificial structural features and introducing artifacts during volume reconstruction, particularly given the highly variable signal-to-noise ratios inherent in cryo-EM data. Furthermore, many existing methods fail to preserve true distances and densities when embedding data, distorting the underlying free energy landscapes and preventing reliable uncertainty quantification.

Technology

RECOVAR is a computational framework that analyzes conformational heterogeneity in cryogenic electron microscopy and tomography datasets. It utilizes principal component analysis with a regularized covariance estimator to calculate 3D covariance matrices directly from 2D projection images. To handle variable signal-to-noise ratios, it employs Fourier Shell Correlation regularization. A Bayesian framework embeds images into a low-dimensional latent space, enabling the estimation of free energy landscapes. For 3D volume reconstruction, RECOVAR applies adaptive kernel regression to average images across specific spatial domains. The toolkit also supports targeted analysis via focus masks and models complex molecular motions.

This solution is differentiated by its mathematically grounded, “white-box” statistical approach, which overcomes the limitations of neural network-based methods. Unlike deep-learning models that risk “hallucinating” artificial features, RECOVAR’s PCA-based embedding preserves true distances and densities. This allows for accurate uncertainty quantification and reliable validation of reconstructions. Furthermore, its adaptive kernel regression avoids artifacts commonly introduced by neural network decoders, achieving superior resolution in medium-to-high-density regions compared to state-of-the-art tools. By ensuring reconstructed states are strictly supported by underlying data, it provides a highly interpretable and robust framework for structural biology.

Benefits

• Delivers high-resolution, artifact-free 3D volume reconstructions by utilizing adaptive kernel regression, avoiding the “hallucinated” features often produced by black-box neural networks.
• Provides a transparent, mathematically grounded “white-box” statistical approach that ensures high interpretability and reliable uncertainty quantification.
• Accurately models complex, nonlinear molecular motions and conformational changes by preserving data density to estimate true free energy landscapes.
• Integrates seamlessly into existing structural biology workflows by directly accepting standard metadata formats from RELION and cryoSPARC without extensive preprocessing.
• Offers versatile support for both single-particle analysis (SPA) and cryo-electron tomography (cryo-ET), including the use of focus masks to isolate specific structural regions like active sites.
• Ensures high performance and ease of use through GPU acceleration, a Python API, a command-line interface, and a browser-based GUI for 3D visualization and job management.

Applications

• Structure-based drug design
• Protein engineering and design
• Cryo-EM data processing
• Molecular dynamics and modeling
• Targeted binding site analysis

Additional info

RECOVAR GitHub Repository

Copyright 2026 The Trustees of Princeton University. All rights reserved.

Permission is hereby granted, free of charge, to non-profit academic and research institutions (“Academic Users”) to use, reproduce, and modify this software for internal, non-commercial academic and research purposes only, subject to the following conditions:

This license applies solely to internal use for academic research or educational purposes. Any other use, including but not limited to:

   - use by or on behalf of a for-profit organization,

   - incorporation into a commercial product or service,

 - or use in connection with commercial research or fee-for-service activities, requires a separate commercial license. To inquire about a commercial license, please contact: Cortney Cavanaugh at ccavanaugh@princeton.edu or otl@princeton.edu

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: 

* Redistributions of source code must retain the above copyright notice, this list of conditions, and the following disclaimer.

* Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.

* Neither The Trustees of Princeton University nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.

NO EXPRESS OR IMPLIED LICENSES TO ANY PARTY’S PATENT RIGHTS ARE GRANTED BY THIS LICENSE. THIS SOFTWARE IS PROVIDED BY THE TRUSTEES OF PRINCETON UNIVERSITY AND CONTRIBUTORS “AS IS” AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE TRUSTEES OF PRINCETON UNIVERSITY OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

===========================================================

NOTIFICATION OF COMMERCIAL USE

===========================================================

This software is intended for research and instructional purposes only.

To use this software for commercial purposes, please contact: 

Cortney Cavanaugh 

New Ventures and Licensing Associate

Technology Licensing & New Ventures, Princeton University

ccavanaugh@princeton.edu or otl@princeton.edu

===========================================================