Semi-supervised clustering with knowledge-guided representation learning in cryo-electron tomography
5 affiliations
- 1Department of Computer Vision, Mohamed bin Zayed University of Artificial Intelligence, Abu Dhabi, United Arab Emirates
- 2Research Office, Mohamed bin Zayed University of Artificial Intelligence, Abu Dhabi, United Arab Emirates
- 3Department of Radiology, Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Boston, Massachusetts, United States of America
- 4Department of Radiology, Harvard Medical School, Boston, Massachusetts, United States of America
- 5Department of Machine Learning, Mohamed bin Zayed University of Artificial Intelligence, Abu Dhabi, United Arab Emirates
Abstract
The automated discovery of structural patterns in macromolecular complexes remains a central challenge in cryo-electron tomography, particularly in highly heterogeneous datasets. Although fully unsupervised clustering methods have shown promise in grouping subtomograms by structural similarity, they often ignore a crucial source of information: the partial ground truth routinely available to structural biologists from prior studies or manual annotations. In this work, we propose a semi-supervised structural discovery framework that utilizes partial supervision to guide clustering without compromising the ability to uncover previously unknown structures. At the core of our method is a label-anchored probabilistic clustering mechanism that seeds the latent space using a small subset of labeled examples and refines it through a multi-resolution consensus strategy based on PCA-space voting. This is complemented by an entropy-based confidence scoring scheme that attenuates the influence of ambiguous samples, as well as a feature propagation procedure that extends structural labels to low-confidence regions using local similarity in feature space. Together, these components create a stable and adaptive pipeline capable of discovering both known and novel structures. Our approach is efficient, requires as little as 1% of labeled data per class, and consistently produces clearer, more interpretable feature embeddings compared to fully unsupervised methods, with well-separated clusters from the very first iterations. Extensive experiments on simulated and realistic tomographic datasets demonstrate that this semi-supervised strategy significantly improves clustering performance, robustness, and biological relevance in cryo-electron tomography analysis. These methods are integrated as extensions to the existing Deep Iterative Subtomogram Clustering Approach pipeline, enhancing its capability for guided structural discovery. Author summary: Cryo-electron tomography (cryo-ET) enables the visualization of macromolecular complexes in 3D inside cells, but each tomogram contains numerous noisy, overlapping structures that are difficult to sort automatically. Existing clustering methods usually treat every subtomogram as completely unknown, while many biologists already have some prior knowledge from earlier studies or manual annotations. In this work, we introduce a semi-supervised clustering framework that can use a very small number of labeled examples—about 1% of subtomograms per known class—to guide the discovery of structural patterns without blocking the appearance of new, previously unseen structures. The method anchors the clustering process with these trusted labels, scores the confidence of each assignment, and propagates labels only where the data support them. Across challenging simulated datasets and real experimental tomograms, our approach produces clearer feature representations and more accurate clusters than a fully unsupervised state-of-the-art method, even under severe noise and limited tilt conditions. This enables more reliable, scalable mining of large cryo-ET collections where most subtomograms remain unlabeled, helping turn growing data resources into biologically meaningful structural insight.
Related
- Metabolite-centric Knowledge Graph Unveils Predictive and Mechanistic Insights through Self-supervised Multimodal Learning
- Physics-aware resolution enhancement of soft X-ray tomography with measurement-supervised learning
- Fish 3D Clustering App (F3DCA): A Non-Supervised Clustering Method with a Machine Learning Autoencoder for Analyzing 3D Trajectories of Three Fish Species.
- Dual-domain self-supervised feature alignment via spectral–spatial representation learning for deepfake anomaly detection
- An urban patrol case urgency classification method based on self-supervised clustering of a graph neural network
BibTeX
@article{kassab2026semisupervised,
title = {{Semi-supervised clustering with knowledge-guided representation learning in cryo-electron tomography}},
author = {Mohamad Kassab and Chengzhi Cao and Vincent Yao and Xiangrui Zeng and Qirong Ho and Min Xu},
journal = {PLOS Digital Health},
year = {2026},
month = aug,
doi = {10.1371/journal.pdig.0001619},
url = {https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001619},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Semi-supervised clustering with knowledge-guided representation learning in cryo-electron tomography
AU - Mohamad Kassab
AU - Chengzhi Cao
AU - Vincent Yao
AU - Xiangrui Zeng
AU - Qirong Ho
AU - Min Xu
PY - 2026
DA - 2026/08/31
JO - PLOS Digital Health
DO - 10.1371/journal.pdig.0001619
UR - https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001619
AB - The automated discovery of structural patterns in macromolecular complexes remains a central challenge in cryo-electron tomography, particularly in highly heterogeneous datasets. Although fully unsupervised clustering methods have shown promise in grouping subtomograms by structural similarity, they often ignore a crucial source of information: the partial ground truth routinely available to structural biologists from prior studies or manual annotations. In this work, we propose a semi-supervised structural discovery framework that utilizes partial supervision to guide clustering without compromising the ability to uncover previously unknown structures. At the core of our method is a label-anchored probabilistic clustering mechanism that seeds the latent space using a small subset of labeled examples and refines it through a multi-resolution consensus strategy based on PCA-space voting. This is complemented by an entropy-based confidence scoring scheme that attenuates the influence of ambiguous samples, as well as a feature propagation procedure that extends structural labels to low-confidence regions using local similarity in feature space. Together, these components create a stable and adaptive pipeline capable of discovering both known and novel structures. Our approach is efficient, requires as little as 1% of labeled data per class, and consistently produces clearer, more interpretable feature embeddings compared to fully unsupervised methods, with well-separated clusters from the very first iterations. Extensive experiments on simulated and realistic tomographic datasets demonstrate that this semi-supervised strategy significantly improves clustering performance, robustness, and biological relevance in cryo-electron tomography analysis. These methods are integrated as extensions to the existing Deep Iterative Subtomogram Clustering Approach pipeline, enhancing its capability for guided structural discovery. Author summary: Cryo-electron tomography (cryo-ET) enables the visualization of macromolecular complexes in 3D inside cells, but each tomogram contains numerous noisy, overlapping structures that are difficult to sort automatically. Existing clustering methods usually treat every subtomogram as completely unknown, while many biologists already have some prior knowledge from earlier studies or manual annotations. In this work, we introduce a semi-supervised clustering framework that can use a very small number of labeled examples—about 1% of subtomograms per known class—to guide the discovery of structural patterns without blocking the appearance of new, previously unseen structures. The method anchors the clustering process with these trusted labels, scores the confidence of each assignment, and propagates labels only where the data support them. Across challenging simulated datasets and real experimental tomograms, our approach produces clearer feature representations and more accurate clusters than a fully unsupervised state-of-the-art method, even under severe noise and limited tilt conditions. This enables more reliable, scalable mining of large cryo-ET collections where most subtomograms remain unlabeled, helping turn growing data resources into biologically meaningful structural insight.
SN - 2767-3170
ER -