CellDiffusion: Enhancing scRNA-seq data integration with accurate inference of cell-cell association
6 affiliations
- 1Department of Medicine, Boston University Chobanian & Avedisian School of Medicine, Boston, MA, USA.
- 2Beijing Key Laboratory of Collaborative Innovation in Frontier Technologies for Population Quality, Beijing, China.
- 3State Key Laboratory of Female Fertility Promotion, Center for Reproductive Medicine, Department of Obstetrics and Gynecology, Peking University Third Hospital, Beijing, China.
- 4Department of Electrical Engineering and Computer Science, and Christopher S. Bond Life Sciences Center, University of Missouri, Columbia, MO, USA.
- 5Department of Computer Science and Faculty of Computing & Data Sciences, Boston University, Boston, MA, USA.
- 6Department of Biomedical Engineering and Informatics, Luddy School of Informatics, Computing, and Engineering, Indiana University Indianapolis, Indianapolis, IN, USA.
Abstract
Integrating single-cell transcriptomic datasets across laboratories, conditions, and technologies remains challenging because of cell type-specific batch effects and heterogeneous biological compositions. Current methods often fail to separate genuine biological variation from technical noise, misaligning distinct cell populations or leaving batch effects uncorrected. To address these challenges, we present CellDiffusion, an unsupervised graph-based model that propagates cellular features through a diffusion process to strengthen meaningful cell-cell associations in data integration. By learning a cell-cell attention network, CellDiffusion adaptively identifies corresponding populations across batches. Our approach applies cell-specific correction in data integration, generating cell embeddings that align cells across batches while preserving intrinsic cell types and developmental trajectories. We validate CellDiffusion across various datasets, including the human endometrium, one of the most dynamic and complex tissues featuring extensive cellular and molecular changes with the menstrual cycle. This framework also improves other essential analyses such as label transfer, clustering, and trajectory inference. CellDiffusion represents a powerful paradigm for accurate single-cell transcriptomic integrative analysis.
Related
- Enhancing Linear B-Cell Epitope Prediction via Evolutionary Scale Modeling (ESM-2) and Hybrid Feature Integration
- Integration of genome-wide association study and single-cell transcriptome analysis reveals the role of N6-methyladenosine in treatment resistance in lung cancer.
- Accurate calling of low-frequency somatic mutations by sample-specific modeling of error rates
- Accurate prediction of CRISPR editing outcomes in somatic cell lines and zygotes with few-shot learning with inDecay.
- Accurate quantification of protein concentrations and state populations by 2D 1H-N Q-BEST NMR
BibTeX
@article{liu2026celldiffusion,
title = {{CellDiffusion: Enhancing scRNA-seq data integration with accurate inference of cell-cell association}},
author = {Yu-Chen Liu and Ziying Huang and Simon Liang Lu and Lei Jiang and Zedong Lin and Jou-Hsuan Lee and Vijaya B. Kolachalama and Rong Li and Dong Xu and Juexin Wang and Heng Pan and Chao Zhang},
journal = {Science Advances},
year = {2026},
month = oct,
doi = {10.1126/sciadv.aed4155},
url = {https://www.science.org/doi/10.1126/sciadv.aed4155},
note = {CC BY-NC},
}
RIS
TY - JOUR
TI - CellDiffusion: Enhancing scRNA-seq data integration with accurate inference of cell-cell association
AU - Yu-Chen Liu
AU - Ziying Huang
AU - Simon Liang Lu
AU - Lei Jiang
AU - Zedong Lin
AU - Jou-Hsuan Lee
AU - Vijaya B. Kolachalama
AU - Rong Li
AU - Dong Xu
AU - Juexin Wang
AU - Heng Pan
AU - Chao Zhang
PY - 2026
DA - 2026/10/16
JO - Science Advances
DO - 10.1126/sciadv.aed4155
UR - https://www.science.org/doi/10.1126/sciadv.aed4155
AB - Integrating single-cell transcriptomic datasets across laboratories, conditions, and technologies remains challenging because of cell type-specific batch effects and heterogeneous biological compositions. Current methods often fail to separate genuine biological variation from technical noise, misaligning distinct cell populations or leaving batch effects uncorrected. To address these challenges, we present CellDiffusion, an unsupervised graph-based model that propagates cellular features through a diffusion process to strengthen meaningful cell-cell associations in data integration. By learning a cell-cell attention network, CellDiffusion adaptively identifies corresponding populations across batches. Our approach applies cell-specific correction in data integration, generating cell embeddings that align cells across batches while preserving intrinsic cell types and developmental trajectories. We validate CellDiffusion across various datasets, including the human endometrium, one of the most dynamic and complex tissues featuring extensive cellular and molecular changes with the menstrual cycle. This framework also improves other essential analyses such as label transfer, clustering, and trajectory inference. CellDiffusion represents a powerful paradigm for accurate single-cell transcriptomic integrative analysis.
SN - 2375-2548
ER -