## read the rds and integration ----
library(spacexr)
library(Seurat)
## build reference ----
rds_fn1 <- glue("ref/E9.5.rds")
rds_fn2 <- glue("ref/E11.5.rds")
rds_fn3 <- glue("ref/E13.5.rds")
e9.5 <- read_rds(rds_fn1)
e11.5 <- read_rds(rds_fn2)
e13.5 <- read_rds(rds_fn3)
ref_lst <- list(e9.5, e11.5, e13.5)
mrg_seu <- merge(ref_lst[[1]], ref_lst[-1])
samp_df <- mrg_seu@meta.data |> dplyr::slice_sample(by = major_trajectory, prop = 0.05) |>
dplyr::filter(!major_trajectory %in% c("Oligodendrocytes", "T_cells", "Testis_and_adrenal",
"Mast_cells", "B_cells", "Intestine")) # exclude those of minority
ref <- mrg_seu[, rownames(samp_df)]
ref[["RNA"]] <- JoinLayers(ref[["RNA"]])
counts <- ref[["RNA"]]$counts
cluster <- as.factor(ref$major_trajectory)
names(cluster) <- colnames(ref)
nUMI <- ref$nCount_RNA
names(nUMI) <- colnames(ref)
reference <- Reference(counts, cluster, nUMI, min_UMI = 35, n_max_cells = 1e5)
write_rds(reference, "reference_tissue.rds")
## get visium object lists ----
rds_fn <- "st_seu_obj.rds"
seu_lst <- read_rds(rds_fn)
rctd_lst2 <- lapply(1:4, \(idx) {
stg = names(seu_lst)[idx]
visium <- seu_lst[[stg]]
counts <- visium[["Spatial"]]$counts
coords <- GetTissueCoordinates(visium)
colnames(coords) <- c("x", "y")
coords[is.na(colnames(coords))] <- NULL
query <- SpatialRNA(coords, counts, colSums(counts))
write_rds(query, glue("rctd_query_obj_{stg}.rds"))
RCTD <- create.RCTD(query, reference, max_cores = 16)
RCTD <- run.RCTD(RCTD, doublet_mode = "doublet")
return(RCTD)
})
write_rds(rctd_lst2, "rctd_res_lst.rds")12 Deconvolution
12.1 Deconvolution of ST data
A single-nucleus RNA sequencing dataset for mice, GSE186069, was retrieved along with its associated annotations. Cells from embryonic stages E9.5, E11.5, and E13.5 were selected and merged. Deconvolution was performed using the RCTD function implemented in the spacexr package (v2.2.1).
For the whole embryo, due to the large number of cells (n = 691,652), 5% of the total cells were randomly sampled, and rare cell types (fewer than 35 cells) were excluded. The remaining data served as the reference, with the ‘major_trajectory’ column in the metadata used as labels. In the liver region, cell type frequencies were ranked in descending order, and minority cell types (proportion < 2%) were excluded from further analysis. Reference construction involved a 20% random sampling of the majority cell types with detailed annotations from the celltype_update column.