Identifying multi-omics spatial domains via masked autoencoder and diffusion-based latent regularization with SpaMEDM
Abstract
Spatial multi-omics data provide complementary molecular views of tissue organization. However, their integration remains challenging due to data sparsity, technical noise, and modality-specific distributions. Here, we present Spatial Masked autoEncoder and Diffusion-based Model (SpaMEDM), an end-to-end graph representation learning framework that augments a dual-attention graph-integration backbone with masked reconstruction and diffusion-based latent noise-prediction regularization. In the masked track, randomly masked spot features are recovered by leveraging modality- and graph-specific neighborhood information. In parallel, Gaussian perturbations are applied to modality-specific latent representations at randomly sampled noise levels, and a spatial-graph-informed network is trained to predict the injected noise. This diffusion-based objective is used exclusively during training to encourage robustness to latent perturbations, rather than to perform iterative reverse diffusion or sampling during inference. Furthermore, cosine-based paired-spot directional alignment and adjacency reconstruction encourage compatible cross-modality representations and preservation of the selected graph structure. SpaMEDM flexibly supports purely spatial, dual feature–spatial, and modality-specific feature-pruned spatial graph configurations. Evaluations across diverse spatial multi-omics datasets demonstrate competitive spatial-domain identification performance and produce integrated embeddings that facilitate downstream exploratory analyses.
Related
- Identification of transmembrane domains that regulate spatial arrangements and activity of prokineticin receptor 2 dimers
- Morphodynamic domains enable integration of live morphometrics and spatial transcriptomics
- STRIVE: interpretable multimodal subtype discovery in Women’s cancers using an attention-weighted joint variational autoencoder
- scDAEC: deep clustering of single-cell data based on dual autoencoder
- Attention-based hierarchical graph autoencoder for dose-specific single-cell resistance dynamics
BibTeX
@article{zhou2026identifying,
title = {{Identifying multi-omics spatial domains via masked autoencoder and diffusion-based latent regularization with SpaMEDM}},
author = {Chang Zhou and Mingxin Lu and Yilei Meng and Sangyu Hu and Yuping Zhu and Yizhuo Sun and Jianing Gu and Zhaoxuan Tian and Fei Yang and Jun Wang and Li Li},
journal = {Briefings in Bioinformatics},
year = {2026},
month = nov,
doi = {10.1093/bib/bbag520},
url = {https://doi.org/10.1093/bib/bbag520},
note = {CC BY-NC 4.0},
}
RIS
TY - JOUR
TI - Identifying multi-omics spatial domains via masked autoencoder and diffusion-based latent regularization with SpaMEDM
AU - Chang Zhou
AU - Mingxin Lu
AU - Yilei Meng
AU - Sangyu Hu
AU - Yuping Zhu
AU - Yizhuo Sun
AU - Jianing Gu
AU - Zhaoxuan Tian
AU - Fei Yang
AU - Jun Wang
AU - Li Li
PY - 2026
DA - 2026/11/05
JO - Briefings in Bioinformatics
DO - 10.1093/bib/bbag520
UR - https://doi.org/10.1093/bib/bbag520
AB - Spatial multi-omics data provide complementary molecular views of tissue organization. However, their integration remains challenging due to data sparsity, technical noise, and modality-specific distributions. Here, we present Spatial Masked autoEncoder and Diffusion-based Model (SpaMEDM), an end-to-end graph representation learning framework that augments a dual-attention graph-integration backbone with masked reconstruction and diffusion-based latent noise-prediction regularization. In the masked track, randomly masked spot features are recovered by leveraging modality- and graph-specific neighborhood information. In parallel, Gaussian perturbations are applied to modality-specific latent representations at randomly sampled noise levels, and a spatial-graph-informed network is trained to predict the injected noise. This diffusion-based objective is used exclusively during training to encourage robustness to latent perturbations, rather than to perform iterative reverse diffusion or sampling during inference. Furthermore, cosine-based paired-spot directional alignment and adjacency reconstruction encourage compatible cross-modality representations and preservation of the selected graph structure. SpaMEDM flexibly supports purely spatial, dual feature–spatial, and modality-specific feature-pruned spatial graph configurations. Evaluations across diverse spatial multi-omics datasets demonstrate competitive spatial-domain identification performance and produce integrated embeddings that facilitate downstream exploratory analyses.
SN - 1467-5463
SN - 1477-4054
ER -