A Multi-view Graph Contrastive Learning Framework for Deciphering Spatially Resolved Transcriptomics Data.
MuCoST is a multi-view graph contrastive learning framework for deciphering SRT data. a. Workflow of MuCoST. MuCoST uses gene expression profile and spatial location information of SRT data as input data. Spatially adjacent graph, co-expression graph and shuffled graph are constructed in MuCoST. MuCoST adopts a multi-view graph contrastive learning framework to learn latent representations of three views by the InfoNCE loss function. MuCoST extracts the compact latent representation through the reconstruction loss of GCN autoencoder. b. Downstream Analysis. MuCoST performs clustering on latent representation of spatial view and realizes downstream analysis tasks, such as spatial cluster visualization, spatial domain identification, PAGA trajectory inference, subtle biological texture recognition and functional analysis of spatial domain.
See Documentation and Tutorials.
All datasets are open access. The DLPFC datasets are available online at http://research.libd.org/spatialLIBD/. The coronal mouse brain of 10X Visium dataset is available online at https://squidpy.readthedocs.io. Mouse olfactory bulb of Stereo-seq dataset is available online at https://github.com/JinmiaoChenLab/SEDR_analyses/. Human breast cancer of 10X Visium dataset is available online at https://support.10xgenomics.com/spatial-gene-expression/datasets/1.1.0/. Tips: We provide a collection of part of the above data at https://zenodo.org/record/8303947/.
In order to use the structural information of histological images, we also provide a histological mode, and the preprocessing of histological images is basis on the tutorial of SpaGCN. Set mode_his to his to open this mode. We do not enable this mode in the evaluation.
- Scanpy 1.9.3
- PyG 2.3.0
- Pytorch 2.0.1
- scikit-learn 1.2.2
- numpy 1.23.4
- tqdm 4.65.0
- numba 0.56.4
- pot 0.9.0
- argparse 1.1
- r-base 3.6.1
Lei Zhang, Shu Liang, Lin Wan, A multi-view graph contrastive learning framework for deciphering spatially resolved transcriptomics data, Briefings in Bioinformatics, Volume 25, Issue 4, July 2024, bbae255, https://doi.org/10.1093/bib/bbae255