A multi-scale graph frequency network for structural and functional region analysis in spatial transcriptomics

Clinical Snapshot

Key Findings

• P Value: Not reported

• Effect Size: Not reported in abstract; no quantitative performance metrics (e.g., Adjusted Rand Index, Normalised Mutual Information, clustering accuracy) provided

• Primary Outcome: SGFN achieved improved or competitive spatial domain identification performance relative to representative baseline methods across nine diverse biological datasets in reference-based benchmarks, and identified biologically coherent spatial and functional regions in unlabelled datasets

• Nnt Or Sensitivity: Not applicable (computational benchmarking study); no sensitivity, specificity, or equivalent diagnostic performance metrics reported in abstract

• Confidence Interval: Not reported

Clinical Application

Implementation requires access to spatial transcriptomics data (Visium, MERFISH, osmFISH, STARmap platforms), substantial computational infrastructure for deep learning (GPU resources), and advanced bioinformatics expertise. Not directly deployable in routine clinical settings without significant technical support. Feasibility in resource-limited settings is low. Spatial transcriptomics is an emerging research tool in Australian academic medical centres and cancer research institutes (e.g., WEHI, Garvan Institute, QIMR Berghofer). SGFN could support Australian translational research in oncology and neuroscience. It is not a therapeutic or diagnostic device subject to TGA regulation in its current form, and has no PBS implications. RACGP guidelines do not currently address spatial transcriptomics analytical tools. Australian researchers adopting SGFN should note the absence of independent external validation and should conduct local benchmarking before applying findings to clinical research conclusions. The National Health and Medical Research Council (NHMRC) framework for research translation would require prospective validation before any clinical decision-support application. Researchers and translational scientists working with spatial transcriptomics data in oncology (breast cancer, angiosarcoma, brain metastasis), neuroscience (cortical architecture, brain mapping), and developmental biology (embryogenesis). Indirect clinical relevance for pathologists and oncologists seeking spatially resolved molecular characterisation of tumour microenvironments.

Abstract

Spatial transcriptomics enables the systematic exploration of how gene expression patterns are organized within intact tissues, yet effective analysis remains difficult due to the complexity of spatial dependencies and multi-scale tissue architectures. Here, we present the Spatial Graph Frequency Network (SGFN), a deep learning framework that integrates graph signal processing, graph attention, and contrastive learning to jointly model spatial topology and molecular features. Central to SGFN is a frequency-domain enhancement module that decomposes spatial graphs into multi-scale spectral components using the Laplacian eigenbasis, complemented by adaptive wavelet denoising when the retained graph-frequency sequence length permits valid decomposition. Evaluation across diverse biological systems-including the human dorsolateral prefrontal cortex, mouse brain, human breast cancer, osmFISH, MERFISH, STARmap, mouse embryonic development, head and neck angiosarcoma, and brain metastasis-shows that SGFN achieves improved or competitive performance relative to representative baseline methods in reference-based benchmarks, and identifies biologically coherent spatial or functional regions in unlabeled datasets supported by marker-gene, spatial-autocorrelation, cell-type-colocalization, and pathway-enrichment evidence. SGFN accurately reconstructed cortical layer architecture in the human brain, delineated immune and metabolic modules in tumors, and revealed spatiotemporal trajectories during embryogenesis. By combining interpretable frequency-domain representations with data-driven learning, SGFN provides a unified computational framework for decoding tissue organization and molecular heterogeneity, advancing the understanding of developmental, physiological, and pathological spatial systems.

References

1. 1.Dai, R., Wang, Z., Zhang, Z., Lei, L., Wang, M., Li, Z., Liu, X., & Guo, Q. (2026). A multi-scale graph frequency network for structural and functional region analysis in spatial transcriptomics. Functional & Integrative Genomics. https://doi.org/10.1007/s10142-025-01600-w