Jul 22, 2025
CoCo-ST detects global and local biological structures in spatial transcriptomics datasets
- Muhammad Aminu1,
- Bo Zhu1,
- Natalie Vokes1,
- Jia Wu1,
- Jianjun Zhang1
- 1The University of Texas MD Anderson Cancer Center
- The University of Texas MD Anderson Cancer Center
External link: https://doi.org/10.1038/s41556-025-01781-z
Protocol Citation: Muhammad Aminu, Bo Zhu, Natalie Vokes, Jia Wu, Jianjun Zhang 2025. CoCo-ST detects global and local biological structures in spatial transcriptomics datasets. protocols.io https://dx.doi.org/10.17504/protocols.io.x54v95m6pl3e/v1
Manuscript citation:
Muhammad Aminu, Bo Zhu, Natalie Vokes, Hong Chen, Lingzhi Hong, Jianrong Li, Junya Fujimoto, Mehdi Chaib, Yuqiu Yang, Bo Wang, Alissa Poteete, Monique B. Nilsson, Xiuning Le, Tina Cascone, David Jaffray, Nicholas Navin, Tao Wang, Lauren A. Byers, Don L. Gibbons, John Heymach, Ken Chen, Chao Cheng, Jianjun Zhang, Jia Wu (2025) CoCo-ST detects global and local biological structures in spatial transcriptomics datasets.Nature Cell Biology doi: 10.1038/s41556-025-01781-z
License: This is an open access protocol distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Protocol status: Working
We use this protocol and it's working
Created: July 21, 2025
Last Modified: July 22, 2025
Protocol Integer ID: 222846
Keywords: local biological structures in spatial transcriptomics dataset, spatial transcriptomics dataset, st algorithm on spatial transcriptomics dataset, st algorithm for spatial domain detection, spatial domain detection, complexity of spatial domain detection, low variance spatial structure, high variance in global spatial pattern, global spatial pattern, local biological structure, relevant spatial structure, variance spatial structure, diverse tissue sample, spatial structure, main challenges with precancer analysis, precancer analysis, application of the coco, coco, contrastive learning approach, varying spatial resolution, driving tumor progression, spatial resolution, variance domain, st detect, proposed coco
Funders Acknowledgements:
National Institute of Health (NIH)
Grant ID: R01CA262425
National Institute of Health (NIH)
Grant ID: R00CA218667
National Institute of Health (NIH)
Grant ID: R01CA276178
National Institute of Health (NIH)
Grant ID: 5R50CA265307
Cancer Prevention and Research Institute of Texas (CPRI)
Grant ID: RP240117
Cancer Prevention and Research Institute of Texas (CPRI)
Grant ID: RP250399
Abstract
A protocol describing the application of the CoCo-ST algorithm on spatial transcriptomics datasets to detect biologically relevant spatial structures. We demonstrate CoCo-ST's ability to detect low variance spatial structures driving tumor progression.
Two main challenges with precancer analysis are:
- The high variance in global spatial patterns which might lead to the overlook of local, low-variance domains that hold significant biological relevance.
- The complexity of spatial domain detection across diverse tissue samples with varying spatial resolutions, necessitating a robust and scalable approach.
Here, we show CoCo-ST’s contrastive learning approach for identifying both high and low-variance spatial structures, leveraging multi-sample data.
This protocol is associated with our proposed CoCo-ST algorithm for spatial domain detection in Nature Cell Biology
Troubleshooting
CoCoST Code: The CoCoST algorithm can be downloaded from the following GitHub repository:
Download CoCoST from GitHub
source("path/to/CoCoST.R")
Load Background Data (Normal Tissue)
normalTissue <- Load10X_Spatial(
data.dir = "X:/maminu/Processed 129S4 Urethane model/53430/outs",
filename = "filtered_feature_bc_matrix.h5",
assay = "Spatial",
filter.matrix = TRUE,
to.upper = FALSE)
Preprocess the data
normalTissue <- SCTransform(normalTissue, assay = "Spatial", verbose = FALSE)
Load Foreground Data (Abnormal Tissue)
abnormalTissue <- Load10X_Spatial(
data.dir = "X:/maminu/Processed 129S4 Urethane model/53433/outs",
filename = "filtered_feature_bc_matrix.h5",
assay = "Spatial",
filter.matrix = TRUE,
to.upper = FALSE)
Preprocess the data
abnormalTissue <- SCTransform(abnormalTissue, assay = "Spatial", verbose = FALSE)
Graph Construction
fdata <- abnormalTissue@assays[["SCT"]]@scale.data
flocation <- GetTissueCoordinates(abnormalTissue)
bdata <- normalTissue@assays[["SCT"]]@scale.data
blocation <- GetTissueCoordinates(normalTissue)
Construct affinity matrices
rbf <- laplacedot(sigma = 0.50)
fKernel <- kernelMatrix(rbf, t(fdata))
Wf <- [email protected]
bKernel <- kernelMatrix(rbf, t(bdata))
Wb <- [email protected]
Run CoCoST
para <- 0.10
Dim <- 20
gCPCA <- CoCoST(t(fdata), Wf, t(bdata), Wb, para, Dim)
Create Dimensionality Reduction Object in Seurat.
CoCo <- CreateDimReducObject(
embeddings = gCPCA[["fgComponents"]],
loadings = gCPCA[["projMatrix"]],
stdev = numeric(),
key = "CoCoST_",
assay = "SCT")
rownames([email protected]) <- rownames(fdata)
abnormalTissue@reductions[["CoCoST"]] <- CoCo
Run UMAP on CoCoST representations and perform clustering.
abnormalTissue <- RunUMAP(abnormalTissue, reduction = "CoCoST", dims = 1:10,
n.components = 5, reduction.name = "CoCo_UMAP")
abnormalTissue <- FindNeighbors(abnormalTissue, reduction = "CoCo_UMAP", dims = 1:5)
abnormalTissue <- FindClusters(abnormalTissue, verbose = TRUE, resolution = 0.05)
[email protected][["coco_clusters"]] <- [email protected][["SCT_snn_res.0.05"]]
Visualize lower dimensional representations
cols <- c("#E1BD6D", "deepskyblue1", "#7A3A9A", "#ED0000FF", "#0B775E", "#ff00ff", "#7B556C", "#44B05B")
DimPlot(abnormalTissue, reduction = "CoCo_UMAP", label = FALSE, cols = cols)
Generate the spatial domain cluster plot:
SpatialDimPlot(abnormalTissue, label = FALSE, label.size = 3, group.by = "coco_clusters", pt.size.factor = 4) +
scale_fill_manual(values = c("#E1BD6D", "deepskyblue1", "#7A3A9A", "#ED0000FF", "#0B775E", "#ff00ff", "#7B556C", "#44B05B"),
labels = c("Normal lung", "Fibrotic/Scar tissue", "Adjacent normal", "Bronchus/Alveoli",
"Adenoma", "Membrane"))