【一步到位】统一空转visium以及scRNA单细胞seurat转h5ad

#R中演示
#----------------------------------------------------------------------------------
#我们这里演示的visium数据包含多个样本，提取单个的进行演示
#多个的数据其实也是需要先分开在进行转化的
library(Seurat)
CTR3 <- subset(merged,orig.ident=='CTR3')
#查看整合的降维UMAP
color_pal<-c("#11A579","#F2B701","#66C5CC","#80BA5A","#F6CF71","#7F3C8D","#CF1C90","#3969AC","#f97b72","#E73F74","#4b4b8f","#ACA4E2","#31C53F","#B95FBB","#D4D915","#28CECA")
DimPlot(CTR3, reduction = "umap", label = F, pt.size = 0.5,cols = color_pal)
#可视化slice
SpatialDimPlot(CTR3, stroke=0.1)& #如果spots不需要边框，设置stroke=NA
  scale_fill_manual(values = color_pal) 
#----------------------------------------------------------------------------------

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import anndata as ad
import scanpy as sc
import squidpy as sq

#这里我是直接使用转化好的演示，转化的数据只有矩阵信息，没有image
CTR3 = sc.read_h5ad('./CTR3.h5ad')
CTR3
# AnnData object with n_obs × n_vars = 1801 × 18274
# obs: 'orig.ident', 'nCount_Spatial', 'nFeature_Spatial', 'group', 'percent.mt', 'nCount_SCT', 'nFeature_SCT', 'nCount_RNA', 'nFeature_RNA', 'integrated_snn_res.0.6', 'seurat_clusters'
# obsm: 'pca', 'umap'

sc.pl.umap(CTR3,color='seurat_clusters')

#因为之前的数据分析是整合的，所以原有的barcode加了后缀，为了与原始的一致，需要处理一下，把多余添加的后缀去除
#这里的处理是直接使用replace将“_1“替换掉就可以了
CTR3.obs_names = CTR3.obs_names.str.replace(r'_1$', '', regex=True)

#读入tissue_positions文件并过滤在组织上的spots
pos = pd.read_csv('./visium/CTR3_processed_data/spatial/tissue_positions.csv')
pos = pos.set_index("barcode")
pos = pos.loc[pos.index.isin(CTR3.obs_names)]

#添加坐标
CTR3.obs = CTR3.obs.merge(pos, left_index=True, right_index=True, how="left")
pxr, pxc = "pxl_row_in_fullres", "pxl_col_in_fullres"
CTR3.obsm["spatial"] = CTR3.obs[[pxc, pxr]].to_numpy()

#添加image
import json
import os
from PIL import Image
with open(os.path.join('./visium/CTR3_processed_data/spatial/scalefactors_json.json'), "r") as fh:
    scales = json.load(fh)
img_hires = np.array(Image.open(os.path.join('./visium/CTR3_processed_data/spatial/tissue_hires_image.png')))
img_low   = np.array(Image.open(os.path.join('./visium/CTR3_processed_data/spatial/tissue_lowres_image.png')))
#加入adata
lib_id = 'CTR3'
CTR3.uns["spatial"] = {
    lib_id: {"images": {"hires": img_hires, "lowres": img_low}, "scalefactors": scales, "metadata": {}}
}

CTR3

AnnData object with n_obs × n_vars = 1801 × 18274
    obs: 'orig.ident', 'nCount_Spatial', 'nFeature_Spatial', 'group', 'percent.mt', 'nCount_SCT', 'nFeature_SCT', 'nCount_RNA', 'nFeature_RNA', 'integrated_snn_res.0.6', 'seurat_clusters', 'in_tissue', 'array_row', 'array_col', 'pxl_row_in_fullres', 'pxl_col_in_fullres'
    uns: 'seurat_clusters_colors', 'spatial'
    obsm: 'pca', 'umap', 'spatial'

sq.pl.spatial_scatter(CTR3)

sc.pl.spatial(CTR3, img_key="lowres", color="seurat_clusters", size=1.5)

#转化演示
setwd("/home/ks_ts/data_analysis/Spatial_analysis/")
ks_convert_seurat_to_h5ad(seurat_obj = CTR3,h5ad_file_path = './',
                       image_path = "./visium/CTR3_processed_data/spatial/",
                       use_counts = FALSE,use_data = TRUE,
                       assay = 'SCT',include_spatial = T)

#读入转化号的数据
CTR3 = sc.read_h5ad('./CTR3.h5ad')
CTR3
# AnnData object with n_obs × n_vars = 1801 × 18274
# obs: 'orig.ident', 'nCount_Spatial', 'nFeature_Spatial', 'group', 'percent.mt', 'nCount_SCT', 'nFeature_SCT', 'nCount_RNA', 'nFeature_RNA', 'integrated_snn_res.0.6', 'seurat_clusters'
# uns: 'spatial'
# obsm: 'pca', 'spatial', 'umap'

sq.pl.spatial_scatter(CTR3)

sq.pl.spatial_scatter(CTR3,color="seurat_clusters", size=1.5)