bm2-lab / spado

Multi-slice Spatial Transcriptome Domain Analysis.

SpaDo

Multi-slice Spatial Transcriptome Domain Analysis.

Fig 1. Workflow of SpaDo. a. Calculating the SPACE for both single-cell resolution and spot resolution spatial transcriptomic data. SPACE: SPatially Adjacent Cell type Embedding. b. Three functions involved in multi-slice spatial domain analysis: multi-slice domain detection, reference-based spatial domain annotation, and multi-slice clustering analysis by consideration of spatial domain composition. JSD, Jensen Shannon Divergence.

Overview

With the rapid advancements in spatial transcriptome sequencing, multiple tissue slices are now available, enabling the integration and interpretation of spatial cellular landscapes. Herein, we introduce SpaDo, a tool for multi-slice spatial domain analysis, including modules for multi-slice spatial domain detection, reference-based annotation, and multiple slice clustering at both single-cell and spot resolutions. We demonstrate SpaDo's effectiveness with over 40 multi-slice spatial transcriptome datasets from 7 sequencing platforms. Our findings highlight SpaDo's potential to reveal novel biological insights in multi-slice spatial transcriptomes.

Installation

• SpaDo package can be installed from Github using devtools packages with R>=4.0.5.

• Attention： SpaDo was originally built on Seurat 4.0.4 and SeuratObject 4.0.4. To avoid problems due to Seurat updates, please make sure you are using the same version of Seurat and SeuratObject packages （Seurat 4.0.4 and SeuratObject 4.0.4）!

  library(devtools)
  install_github("bm2-lab/SpaDo")

Getting started

See Tutorials with different styles and Demo datasets

Citation

Duan, B., Chen, S., Cheng, X. et al. Multi-slice spatial transcriptome domain analysis with SpaDo. Genome Biol 25, 73 (2024). https://doi.org/10.1186/s13059-024-03213-x

spado's Issues

Bugs when integrated a large 10X Visium spatial RNA-seq data (total samples = 24, spots ≈ 60k) by using `DomainPlot_multiple`

Hi,
Thank you very much for developing such useful tools. When integrating a large spatial RNA-seq data (10X Visium, total samples = 24, spots ≈ 60k), I encountered some bugs with DomainPlot_multiple function:

> DomainPlot_multiple(domain_hclust = domain_hclust,
+                     distribution_distance = DD)
Loading required package: umap


 *** caught segfault ***
address 0x7f5874b53ec0, cause 'memory not mapped'

Traceback:
 1: metric.function(dT, i, neighbors)
 2: FUN(X[[i]], ...)
 3: lapply(seq_len(V), function(i) {    neighbors <- pick.random.k(i)    neighbor.distances <- metric.function(dT, i, neighbors)    result <- matrix(c(i, neighbors, -1, neighbor.distances),         ncol = 2)    result[order(result[, 2]), ]})
 4: knn.from.data(dT, k, metric.function, subsample.k)
 5: knn.from.data.reps(d, k, distfun, reps = config$knn_repeats)
 6: knn.info(d, config)
 7: implementations[[method]](d, config)
 8: umap(t(data))
 9: DrawCluster(t(distribution_distance), method = "umap", label = cluster_hc[row.names(distribution_distance)])
10: DomainPlot_multiple(domain_hclust = domain_hclust, distribution_distance = DD)

Possible actions:
1: abort (with core dump, if enabled)
2: normal R exit
3: exit R without saving workspace
4: exit R saving workspace
Selection: dev.off()
Selection: 1
R is aborting now ...
Segmentation fault (core dumped)

Any suggestions would be greatly appreciated!

Error: No normalization

Hello,

Trying to run either InitialClustering or SpatialCellTypeDistribution with Seurat v5:

Error in ScaleData(): ! No layer matching pattern 'data' found. Please run NormalizeData and retry

The problem can be fixed by adding NormalizeData:

sce_seurat <- FindVariableFeatures(sce_seurat, nfeatures = nfeatures)  
sce_seurat <- NormalizeData(sce_seurat) ## Added  
sce_seurat <- ScaleData(sce_seurat)