A comprehensive survey of state-of-the-art spatial clustering algorithms in spatially resolved transcriptomics
Spatial transcriptomics techniques enable researchers to quantify and localize mRNA transcripts at high resolution while preserving their spatial context. Spatial domain identification, or spatial clustering, is crucial in investigating spatial transcriptome technologies. Graph neural networks (GNNs) have emerged as a promising approach to performing cell type classification by leveraging gene expression, spatial location, and histology images. This manuscript comprehensively overviews state-of-the-art GNN-based spatial clustering algorithms for spatially resolved transcriptomics. We evaluate their performance on prevalent spatial transcriptome data, including clustering accuracy, robustness, achievement, requirement, computational efficiency, memory usage, and other relevant metrics. In each method, we estimate a total of 60 clustering situations by extending the essential spatial clustering frameworks in GNN selections, downstream clustering algorithms, principal component analysis (PCA) reduction, and refined correction. By comparing and analyzing the spatial clustering performance, we identify the relevant limitations of current applications and specify future research directions in this field. This survey provides novel insights and additional motivations for investigating spatial transcriptomics.
- Python (>=3.8)
- Scanpy
- Squidpy
- Pytorch_pyG
- Pandas
- Numpy
- Sklearn
- Seaborn
- Matplotlib
- Torch_geometric
- Selecting the optimal GNN for spatial clustering task in GraphST
running GraphST.py to choose the suitable GNN for GraphST on human breast cancner data - Selecting the optimal GNN for spatial clustering task in STAGATE
running STAGATE.py to choose the suitable GNN for GraphST on human breast cancner data
Benchmarking methods used in this paper include:
Distributed under the MIT License. See LICENSE.txt for more information.
-
data files
Please down load the spatial transcriptomics data from the provided links. -
Porch_pyg
Please follow the instruction to install pyG and geometric packages.