Skip to contents

Interaction analysis

Document generated: 2025-08-27 15:59:01 UTC+0000

Source: vignettes/interaction_analysis.Rmd
interaction_analysis.Rmd

Overwiev

For advanced spatial transcriptomics analysis, we recommend exploring intercellular pattern identification using the Python package LIANA+ in combination with MISTY.

Below we show the easiest way to export count matrices, spatial coordinates, and associated metadata from R, so they can be loaded into Python for further analysis.

library(SingleCellExperiment)
sce <- drake::loadd(sce_final_norm_clustering)
# Save count matrices
counts_matrix <- counts(sce)
writeMM(counts_matrix,file = "counts_matrix.mtx")

logcounts_matrix <- logcounts(sce)
writeMM(logcounts_matrix, file = "logcounts_matrix.mts")

# Save gene names, preferebly gene symbols
genes <- rownames(counts_matrix)
write.table(genes,file="genes.tsv", quote = FALSE, sep = "\t", row.names = FALSE, col.names = FALSE)

# Save cell barcodes
barcodes <- colnames(counts_matrix)
write.table(barcodes,"barcodes.tsv",quote = FALSE, sep = "\t", row.names = FALSE, col.names = FALSE)

# Save metadata - make sure that spatial coordinates are in sce metadata
write.csv(as.data.frame(colData(sce)), "metadata.csv", row.names = TRUE)

The following example shows how to load the exported data into an AnnData object. We recommend usage of Jupyter notebook.

from scipy.io import mmread
import anndata
import numpy as np
import pandas as pd

X_raw = mmread("counts_matrix.mtx").T.tocsr()

# Then: normalized data
X_norm = mmread("logcounts_matrix.mtx").T.tocsr()

# Load genes and barcodes again
var_names = pd.read_csv("genes", header=None)[0].tolist()
obs_names = pd.read_csv("barcodes", header=None)[0].tolist()

# Create the base AnnData object with normalized data
adata = anndata.AnnData(X=X_norm)
adata.var_names = var_names
adata.obs_names = obs_names

# Add raw counts as a layer or raw slot
adata.layers["counts"] = X_raw
adata.raw = adata.copy()  # optional, depends if you want .raw as raw counts

# Add metadata if available
metadata = pd.read_csv("metadata.csv", index_col=0)
adata.obs = metadata.loc[adata.obs_names]

adata.layers["logcounts"] = adata.X

#check spatial coord names, might differ
spatial_coords = metadata.loc[adata.obs_names, ["sdimx", "sdimy"]].to_numpy()

# Add to obsm
adata.obsm["spatial"] = spatial_coords

# Save created anndata object
adata.write("adata.h5ad")

Once the data is prepared, you can proceed with the LIANA tutorial