Related Experiment Video
Updated: Jul 8, 2025
Spatial Profiling of Protein and RNA Expression in Tissue: An Approach to Fine-Tune Virtual Microdissection
Published on: July 6, 2022
A comprehensive overview of graph neural network-based approaches to clustering for spatial transcriptomics
Teng Liu1,2, Zhao-Yu Fang3, Zongbo Zhang1
1Clinical Research Center (CRC), Clinical Pathology Center (CPC), Cancer Early Detection and Treatment Center (CEDTC) and Translational Medicine Research Center (TMRC), Chongqing University Three Gorges Hospital, Chongqing University, Wanzhou, Chongqing, China.
This study surveys graph neural network (GNN) methods for spatial clustering in spatial transcriptomics. It evaluates GNN performance for identifying spatial domains, offering insights into limitations and future research directions.
Area of Science:
- Computational Biology
- Bioinformatics
- Genomics
Background:
- Spatial transcriptomics provides high-resolution spatial context for messenger ribonucleic acid (mRNA) quantification.
- Identifying spatial domains (spatial clustering) is vital for analyzing spatial transcriptomics data.
- Graph neural networks (GNNs) show promise for spatial domain classification by integrating gene expression, location, and histology.
Purpose of the Study:
- To provide a comprehensive overview of recent GNN-based spatial clustering methods for spatial transcriptomics.
- To evaluate the performance of current GNN methods on prevalent spatial transcriptomics datasets.
- To identify limitations and suggest future research directions in GNN-based spatial clustering.
Main Methods:
- Systematic review and evaluation of GNN-based spatial clustering methods.
- Performance assessment across 60 clustering scenarios, varying GNNs, clustering algorithms, PCA reduction, and correction methods.
- Comparative analysis of accuracy, robustness, data stabilization, requirements, computational efficiency, and memory usage.
Main Results:
- Comprehensive performance evaluation of existing GNN-based spatial clustering techniques.
- Identification of strengths and weaknesses of different methods across various datasets and scenarios.
- Insights into the practical considerations for applying GNNs in spatial transcriptomics analysis.
Conclusions:
- GNNs are a powerful tool for spatial domain identification in spatial transcriptomics.
- Performance varies based on method choice, data characteristics, and parameter settings.
- Further research is needed to refine GNNs for enhanced accuracy, efficiency, and broader applicability in spatial transcriptomics.
