?現(xiàn)在10x Visium數(shù)據(jù)基礎(chǔ)的分析思路是將每個spot看作一個細(xì)胞，然后參考單細(xì)胞轉(zhuǎn)錄組的分析思路進(jìn)行分析。但是現(xiàn)在的實驗中，單個spot中包含不僅是一個細(xì)胞。如何確定每個spot中包含的細(xì)胞，對于空間轉(zhuǎn)錄組的分析是有幫助的。SPOTlight可以結(jié)合單細(xì)胞RNA測序信息反卷積空間數(shù)據(jù)，識別每個spot中的細(xì)胞類型和比例。
安裝
##安裝穩(wěn)定版
devtools::install_github(“https://github.com/MarcElosua/SPOTlight”)
##安裝開發(fā)版
devtools::install_github(“https://github.com/MarcElosua/SPOTlight”, ref = “devel”)
測試數(shù)據(jù)的獲取
可以從https://github.com/MarcElosua/SPOTlight和https://satijalab.org/seurat/v3.2/spatial_vignette.html，獲取10x空間轉(zhuǎn)錄組數(shù)據(jù)和配套的單細(xì)胞數(shù)據(jù)，作為測試數(shù)據(jù)集。
數(shù)據(jù)的加載與分析
單細(xì)胞數(shù)據(jù)：
##單細(xì)胞數(shù)據(jù)處理
sc_data <- readRDS(“../data/allen_cortex_dwn.rds”)
##常規(guī)處理
sc_data <- Seurat::SCTransform(sc_data , verbose = FALSE)
sc_data <-sc_data %>% Seurat::RunPCA(verbose = FALSE) %>%
Seurat::RunUMAP(dims = 1:30, verbose = FALSE) %>%
Seurat::FindNeighbors(dims = 1:30, verbose = FALSE) %>%
Seurat::FindClusters(verbose = FALSE)
空間數(shù)據(jù)：
#InstallData(“stxBrain”)
anterior <- LoadData(“stxBrain”, type = “anterior1”)
anterior <- Seurat::SCTransform(anterior, assay = “Spatial”, verbose = FALSE)
anterior <- anterior %>% Seurat::RunPCA(verbose = FALSE) %>%
Seurat::RunUMAP(dims = 1:30, verbose = FALSE) %>%
Seurat::FindNeighbors(dims = 1:30, verbose = FALSE) %>%
Seurat::FindClusters(verbose = FALSE)
簡單的可視化
sc_p <- DimPlot(sc_data,reduction = “umap”,label = T,group.by = “subclass”)*NoLegend()
st_p <- Seurat::SpatialDimPlot(anterior,label = T)*NoLegend()
sc_p+st_p
SPOTlight的使用
1.獲得單細(xì)胞不同亞型細(xì)胞的marker基因：
###find marker gene
Idents(object = sc_data)
Seurat::Idents(object = sc_data) <- sc_data@meta.data$subclass
cluster_markers_all <- Seurat::FindAllMarkers(object = sc_data,
assay = “SCT”,
slot = “data”,
verbose = TRUE,
only.pos = TRUE,
logfc.threshold = 1,
min.pct = 0.9)
參數(shù)設(shè)置：
1）從SCT的data數(shù)據(jù)中獲取差異基因，為了獲得所有可能的差異基因而不僅是高突變基因中的差異基因；
2）只選擇在cluster表達(dá)的基因，only.pos=T；
3）選擇在同類細(xì)胞中絕大多數(shù)細(xì)胞都表達(dá)的基因，min.pct=0.9。
2. 利用單細(xì)胞數(shù)據(jù)對空間表達(dá)數(shù)據(jù)進(jìn)行反卷積：
set.seed(123)
spotlight_ls <- spotlight_deconvolution(se_sc = sc_data,
counts_spatial = anterior@assays$Spatial@counts,
clust_vr = “subclass”,
cluster_markers = cluster_markers_all,
cl_n = 50,
hvg = 3000,
ntop = NULL,
transf = “uv”,
method = “nsNMF”,
min_cont = 0.09)
?spotlight_deconvolution查看參數(shù)的詳細(xì)說明
3. 獲取每個Spot中細(xì)胞的組成比例
###獲取每個spot中的細(xì)胞比例矩陣
decon_mtrx <- spotlight_ls[[2]]
cell_types_all <- colnames(decon_mtrx)[which(colnames(decon_mtrx) != “res_ss”)]
##將信息添加到空間數(shù)據(jù)中
anterior@meta.data <- cbind(anterior@meta.data, decon_mtrx)
##可視化
SPOTlight::spatial_scatterpie(se_obj = anterior,
cell_types_all = cell_types_all,
img_path = “../data/spatial/tissue_lowres_image.png”,
pie_scale = 0.4)
4. 特定細(xì)胞的展示
###展示特定細(xì)胞所在位置
p1 <- SPOTlight::spatial_scatterpie(se_obj = anterior,
cell_types_all = cell_types_all,
img_path = “../data/spatial/tissue_lowres_image.png”,
cell_types_interest = “L6b”,
pie_scale = 0.5)
###展示特定細(xì)胞的表達(dá)情況
p2 <- SpatialFeaturePlot(anterior,
features = “L6b”,
pt.size.factor = 1,
alpha = c(0, 1)) +theme(legend.position = “right”)+
ggplot2::scale_fill_gradientn(
colours = heat.colors(10, rev = TRUE),
limits = c(0, 1))
p1+p2
5. 空間交互信息的獲取展示
在獲取了每個spot的細(xì)胞類型后，可以利用get_spatial_interaction_graph獲得細(xì)胞在空間中的相互作用的信息。
###空間交互信息
graph_ntw <- get_spatial_interaction_graph(decon_mtrx = decon_mtrx[, cell_types_all])
graph_ntw
[1] Astro–Endo Astro–L2.3.IT Astro–L4
[4] Astro–L5.IT Astro–L5.PT Astro–L6.CT
[7] Astro–L6.IT Astro–L6b Astro–Lamp5
[10] Astro–Macrophage Astro–Meis2 Astro–NP
[13] Astro–Oligo Astro–Peri Astro–Pvalb
[16] Astro–Serpinf1 Astro–SMC Astro–Sncg
[19] Astro–Sst Astro–Vip Astro–VLMC
[22] Endo –L2.3.IT Endo –L4 Endo –L5.IT
###繪圖
library(igraph)
library(RColorBrewer)
deg <- degree(graph_ntw, mode=”all”)
# Get color palette for difusion
edge_importance <- E(graph_ntw)$importance
# Select a continuous palette
qual_col_pals <- brewer.pal.info[brewer.pal.info$category == ‘seq’,]
# Create a color palette
getPalette <- colorRampPalette(brewer.pal(9, “YlOrRd”))
# Get how many values we need
grad_edge <- seq(0, max(edge_importance), 0.1)
# Generate extended gradient palette dataframe
graph_col_df <- data.frame(value = as.character(grad_edge),
color = getPalette(length(grad_edge)),
stringsAsFactors = FALSE)
# Assign color to each edge
color_edge <- data.frame(value = as.character(round(edge_importance, 1)), stringsAsFactors = FALSE) %>%
dplyr::left_join(graph_col_df, by = “value”) %>%
dplyr::pull(color)
# Open a pdf file
plot(graph_ntw,
# Size of the edge
edge.width = edge_importance,
edge.color = color_edge,
# Size of the buble
vertex.size = deg,
vertex.color = “#cde394”,
vertex.frame.color = “white”,
vertex.label.color = “black”,
vertex.label.family = “Ubuntu”, # Font family of the label (e.g.“Times”, “Helvetica”)
layout = layout.circle)
以上，便是SPOTlight的簡單使用的展示，讀者有興趣還可以將SPOTlight獲得結(jié)果與Seurat中的單細(xì)胞和空間的聯(lián)合分析的結(jié)果進(jìn)行比較。
京公網(wǎng)安備 11011302003368號