空间转录组（1）—输入文件说明及数据读取

原创

sheldor没耳朵

发布于 2026-01-28 15:12:34

1570

空间转录组（1）—输入文件说明及数据读取

对于空转的数据分析，一直没有时间整理笔记，今天抽空写一篇，就从读取数据的开始吧。

1.文件说明

空转和单细胞一样，对于数据的读取有着严格的限制，如原始数据的文件名，文件位置等。对于单个样本（以样本名Sample2为例）的读取，需要以下的文件
文件介绍
filtered_feature_bc_matrix.h5：存储经过初步过滤后的基因表达矩阵，记录每个空间 spot 上各基因的转录本计数信息，是后续分析的基础数据
scalefactors_json.json：存储空间坐标与组织图像之间的缩放因子信息，用于保证空间坐标与图像的准确对齐

tissue_positions_list.csv：记录每个 spot 的空间坐标信息，包括其在阵列中的位置及在组织图像中的像素坐标，用于将基因表达与空间位置对应（无列名，若加上列名，后续代码可能报错）

tissue_lowres_image.png：组织切片的低分辨率图像，主要用于快速空间可视化展示（分析用，一定需要）

tissue_hires_image.png：组织切片的高分辨率图像，用于精细的空间可视化和形态学展示（展示用，不一定需要）
目录层级

2.数据读取

Seurat对象构建

rm(list = ls())
library(Seurat)        # 单细胞分析主包
library(ggplot2)       # 绘图
library(patchwork)     # 多图拼接
library(dplyr)         # 数据处理
library(Rfast2)        # 快速数据处理
library(hdf5r)         # 读写h5文件
library(viridis)
library(glmGamPoi)
library(SingleR)
library(celldex)
library(RColorBrewer)
# 用户自定义参数
project_id <- "Sample2"    # 项目/样本名称
output_dir <- "../SpatialAnalysisResults"  # 输出文件夹
raw_h5_file <- "../data_fixed/filtered_feature_bc_matrix.h5"  # 原始表达矩阵文件
#设置工作目录和输出目录
project_id <- "Sample2"
output_dir <- "../SpatialAnalysisResults"
# h5 文件和 spatial 文件路径（相对于工作目录 code/）
h5_file <- "../data_fixed/h5/filtered_feature_bc_matrix.h5"
spatial_dir <- "../data_fixed/spatial"
if (!dir.exists(output_dir)) { dir.create(output_dir) } 
# 使用 Load10X_Spatial 读取整个 10x Visium 空间数据
sp_obj <- Load10X_Spatial(
  data.dir = "../data_fixed",             # data_fixed 目录
  filename = "h5/filtered_feature_bc_matrix.h5",  # h5 文件相对于 data.dir 的路径
  assay = "Spatial",
  slice = project_id
)

ncount、nfeature展示（小提琴图和空间图，只是看一下）

# 获取分组数量
group_count <- length(unique(Idents(sp_obj)))
my_colors <- viridis::viridis(group_count, option = "C")  # 你可以换成"A"/"B"/"D"等

# 绘制nCount_Spatial的小提琴图
vln_plot <- VlnPlot(
  sp_obj,
  features = "nCount_Spatial",
  pt.size = 0,                        # 不显示点
  cols = my_colors                    # 自动分配颜色
) +
  geom_boxplot(width = 0.08, outlier.shape = NA, fill = "white", color = "black", alpha = 0.3) +  # 窄箱线
  theme_minimal(base_size = 16) +
  theme(
    panel.grid = element_blank(),
    axis.text = element_text(color = "black", size = 14),
    axis.title = element_text(size = 16, face = "bold"),
    plot.title = element_text(hjust = 0.5, size = 18, face = "bold"),
    legend.position = "none"
  ) +
  labs(
    title = "nCount_Spatial Distribution",
    x = "Group",
    y = "nCount_Spatial"
  )
# 绘制nCount_Spatial的空间分布图
spatial_plot <- SpatialFeaturePlot(sp_obj, features = "nCount_Spatial") + 
  theme(legend.position = "right")  # 空间分布

# 图片保存
pdf(file = file.path(output_dir, "FeatureDistribution_violin.pdf"), width = 5, height = 6)
print(vln_plot)
dev.off()

pdf(file = file.path(output_dir, "FeatureDistribution_spatial.pdf"), width = 9, height = 6)
print(spatial_plot)
dev.off()

进一步处理，因为我这次处理的数据，是需要复现别人的结果，因此，我是没有自己再聚类（FindClusters）和注释的，直接用的别人注释好的文件

sp_obj <- SCTransform(sp_obj, assay = "Spatial", verbose = FALSE)  # 标准化
sp_obj <- RunPCA(sp_obj, assay = "SCT", verbose = FALSE)  # PCA
sp_obj <- FindNeighbors(sp_obj, reduction = "pca", dims = 1:30)  # 邻居查找
#sp_obj <- FindClusters(sp_obj, verbose = FALSE)  # 聚类，这里直接注释掉
sp_obj <- RunUMAP(sp_obj, reduction = "pca", dims = 1:30)  # UMAP降维

挂载别人注释好的文件

> head(sp_obj@meta.data)
                      orig.ident nCount_Spatial nFeature_Spatial nCount_SCT nFeature_SCT
AACACCTACTATCGAA-1 SeuratProject          90225            11403      15803         6494
AACACGTGCATCGCAC-1 SeuratProject           2638             1898      12580         5813
AACACTTGGCAAGGAA-1 SeuratProject          63980            10160      15590         6199
AACAGGAAGAGCATAG-1 SeuratProject           2956             2310      12369         5787
AACAGGATTCATAGTT-1 SeuratProject           2274             1539      13026         6160
AACAGGCCAACGATTA-1 SeuratProject           2161             1655      13554         6184
> load("ano_2.Rdata")
> head(ano_2)
              Barcode Cluster                      CellType
1  AACACCTACTATCGAA-1      11 Neuroendocrine_carcinoma_cell
5  AACACGTGCATCGCAC-1       4                   Fibroblasts
8  AACACTTGGCAAGGAA-1       1 Neuroendocrine_carcinoma_cell
12 AACAGGAAGAGCATAG-1       7                   Fibroblasts
14 AACAGGATTCATAGTT-1       4                   Fibroblasts
17 AACAGGCCAACGATTA-1       3                 Necrotic_cell
> table(rownames(sp_obj@meta.data) %in% ano_2$Barcode)

FALSE  TRUE 
  101  4891 
> # FALSE  TRUE 
> # 101  4891
# 找出可以匹配的 spot
matched_barcodes <- intersect(rownames(sp_obj@meta.data), ano_2$Barcode)
# 过滤 Seurat 对象，只保留匹配的 spot
sp_obj <- subset(sp_obj, cells = matched_barcodes)

# 挂载 cluster 和 CellType 注释
matched_idx <- match(matched_barcodes, ano_2$Barcode)
sp_obj$Cluster <- ano_2$Cluster[matched_idx]
sp_obj$CellType <- ano_2$CellType[matched_idx]
# 检查
table(is.na(sp_obj$Cluster))  # 应该全部为 FALSE
table(is.na(sp_obj$CellType)) # 应该全部为 FALSE

metatada <- sp_obj@meta.data
metatada[1:2,]
table(metatada$CellType)

#一个 CellType 来自哪些 cluster
cluster_main_celltype <- table(metatada$Cluster, metatada$CellType)
write.csv(cluster_main_celltype,
          file = "../SpatialAnalysisResults/1.数据概览/Sample#2/cluster_main_celltype.csv")

可视化（因涉及到项目保密，只贴了空间聚类图）

# 可视化
#umap
pdf(file = file.path("../SpatialAnalysisResults/1.数据概览/Sample#2/umap_celltype.pdf"), width = 8.5, height = 6)
DimPlot(sp_obj, reduction = "umap", group.by = "CellType", label = TRUE,label.size = 4)
dev.off()
pdf(file = file.path("../SpatialAnalysisResults/1.数据概览/Sample#2/umap_cluster.pdf"), width = 7.5, height = 6)
DimPlot(sp_obj, reduction = "umap", group.by = "Cluster", label = TRUE)
dev.off()
#空间聚类CellType
pdf(file = file.path("../SpatialAnalysisResults/1.数据概览/Sample#2/Spatial_celltype_image.pdf"), width = 8.5, height = 6)
SpatialDimPlot(
  sp_obj,
  group.by = "CellType",
  pt.size.factor = 4.5,
  stroke = 0
) + theme(legend.position = "right")
dev.off()
pdf(file = file.path("../SpatialAnalysisResults/1.数据概览/Sample#2/Spatial_celltype.pdf"), width = 8.5, height = 6)
SpatialDimPlot(
  sp_obj,
  group.by = "CellType",
  pt.size.factor = 4.5,
  stroke = 0,
  image.alpha = 0#组织切片完全透明
) +
  theme(legend.position = "right")
dev.off()
#空间聚类Cluster
pdf(file = file.path("../SpatialAnalysisResults/1.数据概览/Sample#2/Spatial_cluster_image.pdf"), width = 7.5, height = 6)
SpatialDimPlot(
  sp_obj,
  group.by = "Cluster",
  pt.size.factor = 4.5,
  stroke = 0
  
) +
  theme(legend.position = "right")
dev.off()
pdf(file = file.path("../SpatialAnalysisResults/1.数据概览/Sample#2/Spatial_cluster.pdf"), width = 7.5, height = 6)
SpatialDimPlot(
  sp_obj,
  group.by = "Cluster",
  pt.size.factor = 4.5,
  stroke = 0,
  image.alpha = 0#组织切片完全透明
) +
  theme(legend.position = "right")
dev.off()

#只显示Adenocarcinoma_cell（腺癌细胞）与Adipocytes（脂肪细胞）
{
  
  sp_obj$CellType_highlight <- sp_obj$CellType
  sp_obj$CellType_highlight[
    !sp_obj$CellType %in% c("Adenocarcinoma_cell", "Adipocytes")
  ] <- "Other"
  
  cols <- c(
    "Adenocarcinoma_cell" = "#4575B4",
    "Adipocytes" = "#FDAE61",
    "Other" = "grey85"
  )
  
  pdf(file = file.path("../SpatialAnalysisResults/1.数据概览/Sample#2/Spatial_celltype_highlight.pdf"), width = 8.5, height = 6)
  SpatialDimPlot(
    sp_obj,
    group.by = "CellType_highlight",
    pt.size.factor = 4.5,
    stroke = 0,
    image.alpha = 0,
    cols = cols
  ) +
    theme(
      legend.position = "right",
      legend.title = element_blank()
    )
  dev.off()
  
  pdf(file = file.path("../SpatialAnalysisResults/1.数据概览/Sample#2/Spatial_celltype_highlight_image.pdf"), width = 8.5, height = 6)
  SpatialDimPlot(
    sp_obj,
    group.by = "CellType_highlight",
    pt.size.factor = 4.5,
    stroke = 0,
    cols = cols
  ) +
    theme(
      legend.position = "right",
      legend.title = element_blank()
    )
  dev.off()
}
sample2_sp_obj <- sp_obj
save(sample2_sp_obj,file = "../Rdata/step1_sample2_sp_obj.Rdata")