作者，Evil Genius

今天我们分享一个简单的内容，Seurat V5 整合空间的visium数据和HD数据

脚本就不封装了，大家自己封装吧。

首先是visium,大家还记得如何判断空转的整合效果吧，忘记的赶紧复习一下。

未封装的脚本

suppressMessages({
library(Seurat)
library(argparse)
library(dplyr)
library(ggplot2)
library(Banksy)
library(SeuratWrappers)
library(harmony)
})

HC_1L =  Load10X_Spatial('/home/samples/DB/Spatial/visium_data/HC_1L',slice = "HC_1L")

HCC_1L =  Load10X_Spatial('/home/samples/DB/Spatial/visium_data/HCC_1L',slice = "HCC_1L")

HCC_1N =  Load10X_Spatial('/home/samples/DB/Spatial/visium_data/HCC_1N',slice = "HCC_1N")

HCC_1T =  Load10X_Spatial('/home/samples/DB/Spatial/visium_data/HCC_1T',slice = "HCC_1T")

sceList = list(HC_1L,HCC_1L,HCC_1N,HCC_1T)

sceMerged <- merge(x = sceList[[1]], y = sceList[-1], add.cell.ids = c("HC_1L","HCC_1L","HCC_1N","HCC_1T"))

DefaultAssay(sceMerged) = 'Spatial'

sceMerged <- NormalizeData(sceMerged)

sceMerged <- FindVariableFeatures(sceMerged)

sceMerged <- ScaleData(sceMerged)

sceMerged <- RunPCA(sceMerged)

obj <- IntegrateLayers(object = sceMerged, method = HarmonyIntegration,orig.reduction = "pca", new.reduction = "harmony",verbose = FALSE)

sce <- FindNeighbors(obj, reduction = "harmony",dims = 1:20)

sce <- FindClusters(sce,resolution = 0.5)

sce <- RunUMAP(sce, reduction = "harmony",dims = 1:20)

dual.plot <- function(fig, file.prefix, w=7, h=7, res=75){
pdf(paste(file.prefix,".pdf",sep=""), width = w, height = h)
print(fig)
dev.off()
png(paste(file.prefix,".png",sep=""), width = w*res, height = h*res, res = res, type="cairo-png")
print(fig)
dev.off()
}

low_res = 70
mid_res = 150
high_res = 300
mid_res = 300

p <- SpatialDimPlot(sce, group.by = "seurat_clusters", label = T, repel = T, label.size = 4,pt.size.factor = 2.5)

outdir = '/home/samples/DB/Spatial/visium_data'

prefix = 'combined'

dual.plot(p, file.path(outdir, paste(prefix,'.Spatial.cluster',sep="")), w=20, h=6, res=mid_res)

那么对于HD呢？因为涉及到banksy的用法(识别空间域)，所以复杂一点，需要大算力服务器

其中HD的数据质量大家要自己控制一下，8um的基因中位数过少是没办法进行下游分析的。

未封装的脚本