复现文章信息: 文章题目 :Single-cell analysis reveals prognostic fibroblast subpopulations linked to molecular and immunological subtypes of lung cancer 期刊 :Nature Communications日期 :2023年1月31日DOI : 10.1038/s41467-023-35832-6
复现图——Figure 5 基于机器学习的scRNA-seq数据和mxIHC分类显示外膜和肌成纤维细胞在胰腺癌、结直肠癌和口腔癌中是保守的,而肺泡成纤维细胞是肺特异性的 R包载入与数据准备 library (Seurat)library (sctransform)library (ggplot2)library (WGCNA)library (tidyverse)library (ggpubr)library (ggsci) data_directory <- "H:\\文献复现\\6\\" source (paste0(data_directory, "0_NewFunctions.R" )) load(paste0(data_directory, "IntegratedFibs_Zenodo.Rdata" )) load(paste0(data_directory, "CrossTissueAnalysis_Zenodo.Rdata" )) load(paste0(data_directory, "MxIHC_TMAdata_Zenodo.Rdata" ))
Figure 5 A-C 分离自不同癌症类型并通过scRNA-seq分析的成纤维细胞的UMAP降维。检测这些成纤维细胞表型是否在不同癌症类型中是保守的,分析了PDAC49、HNSCC29和结肠直肠癌(CRC)。在每种情况下,成纤维细胞都是通过无监督聚类和壁细胞排除法鉴定
Sample_UMAP <- Merged_MetaData %>% filter(Group %in % c("Pancreas" , "Oral" , "Colon" )) %>% ggplot(aes(x = UMAP_1, y = UMAP_2, colour = Sample.type)) + geom_point(size = 0.1 ) + facet_wrap(~Group, scales = "free" , nrow = 1 ) + theme_pubr(base_size = 15 ) + scale_color_npg(name = "Sample type" ) + theme(legend.position = "right" ,legend.key.size = unit(10 , "pt" ))+ guides(colour = guide_legend(override.aes = list(size = 2 )))#突出显示机器学习分类器预测的与每个细胞相关的成纤维细胞亚群 Class_UMAP <- Merged_MetaData %>% filter(Group %in % c("Pancreas" , "Oral" , "Colon" )) %>% ggplot(aes(x = UMAP_1, y = UMAP_2, colour = predicted.id)) + geom_point(size = 0.1 ) + facet_wrap(~Group, scales = "free" , nrow = 1 ) + theme_pubr(base_size = 15 ) + scale_colour_manual(values = Fibs_col.palette, name = "Predicted class" ) + theme(legend.position = "right" ,legend.key.size = unit(10 , "pt" )) + guides(colour = guide_legend(override.aes = list(size = 2 )))#小提琴图显示了按亚群分组的机器学习分类器模型预测的概率 Prob_VlnPlot <- Merged_MetaData %>% filter(Group %in % c("Pancreas" , "Oral" , "Colon" )) %>% ggplot(aes(x = predicted.id, y = prediction.score.max, fill = predicted.id)) + geom_violin(scale = "width" ) + geom_boxplot(width = 0.1 , outlier.shape = NA , fill = "white" ) + facet_wrap(~Group, scales = "free" , nrow = 1 ) + theme_pubr(base_size = 15 ) + scale_fill_manual(values = Fibs_col.palette, name = "Predicted class" ) + rotate_x_text(angle = 45 ) + theme(legend.position = "right" , axis.title.x = element_blank(), legend.key.size = unit(10 , "pt" )) + ylab("Classification Probability" ) + ylim(c(0 ,1 )) Fig_5ABC <- ggarrange(Sample_UMAP, Class_UMAP, Prob_VlnPlot, nrow = 3 , align = "v" )
Fig_5ABC 这表明,在分析的所有癌症类型中,外膜细胞和肌成纤维细胞群都是高度保守的,而分配给肺泡亚群的成纤维细胞的概率得分一直较低,表明与肺的表型差异程度更大。
Figure 5D 来自组织微阵列(TMA)的mxIHC分析的代表性图像,所述组织微阵列由胰腺癌、口腔癌和结肠癌组织块构建。可视化显示成纤维细胞亚群的空间分布
s = "PANCREAS" PDAC_CTR_C01 <- All_TMA.data.df %>% filter(TvN == "Normal" & Core == "C01" ) %>% ggplot(aes(x = X.Center..Pxl., y = Y.Center..Pxl., colour = Cell.type2)) + geom_point(size = 1.5 ) + theme_void(base_size = 15 ) + theme(legend.position = "none" ) + guides(colour = guide_legend(override.aes = list(size = 4 ))) + scale_colour_manual(values = Fibs_col.palette, na.value = "grey80" ) PDAC_Tumour_CO2 <- All_TMA.data.df %>% filter(TvN == "Tumour" & Core == "C02" ) %>% ggplot(aes(x = X.Center..Pxl., y = Y.Center..Pxl., colour = Cell.type2)) + geom_point(size = 1.5 ) + theme_void(base_size = 15 ) + theme(legend.position = "none" ) + guides(colour = guide_legend(override.aes = list(size = 4 ))) + scale_colour_manual(values = Fibs_col.palette, na.value = "grey80" ) HNSCC_CTR_E07 <- All_TMA.data.df %>% filter(TvN == "Normal" & Core == "E07" ) %>% ggplot(aes(x = X.Center..Pxl., y = Y.Center..Pxl., colour = Cell.type2)) + geom_point(size = 1.5 ) + theme_void(base_size = 15 ) + theme(legend.position = "none" ) + guides(colour = guide_legend(override.aes = list(size = 4 ))) + scale_colour_manual(values = Fibs_col.palette, na.value = "grey80" ) HNSCC_Tumour_E07 <- All_TMA.data.df %>% filter(TvN == "Tumour" & Core == "E07" ) %>% ggplot(aes(x = X.Center..Pxl., y = Y.Center..Pxl., colour = Cell.type2)) + geom_point(size = 1.5 ) + theme_void(base_size = 15 ) + theme(legend.position = "none" ) + guides(colour = guide_legend(override.aes = list(size = 4 ))) + scale_colour_manual(values = Fibs_col.palette, na.value = "grey80" ) COLON_CTR_B03 <- All_TMA.data.df %>% filter(TvN == "Normal" & Core == "B03" ) %>% ggplot(aes(x = X.Center..Pxl., y = Y.Center..Pxl., colour = Cell.type2)) + geom_point(size = 1.5 ) + theme_void(base_size = 15 ) + theme(legend.position = "none" ) + guides(colour = guide_legend(override.aes = list(size = 4 ))) + scale_colour_manual(values = Fibs_col.palette, na.value = "grey80" ) COLON_Tumour_B05 <- All_TMA.data.df %>% filter(TvN == "Tumour" & Core == "B05" ) %>% ggplot(aes(x = X.Center..Pxl., y = Y.Center..Pxl., colour = Cell.type2)) + geom_point(size = 1.5 ) + theme_void(base_size = 15 ) + theme(legend.position = "none" ) + guides(colour = guide_legend(override.aes = list(size = 4 ))) + scale_colour_manual(values = Fibs_col.palette, na.value = "grey80" ) Fig_5D <- ggarrange(PDAC_CTR_C01, PDAC_Tumour_CO2, HNSCC_CTR_E07, HNSCC_Tumour_E07, COLON_CTR_B03, COLON_Tumour_B05, nrow = 6 , ncol = 1 )
Fig_5D 通过将多重免疫组化面板应用于由来自PDAC、HNSCC和CRC的肿瘤和对照组织核心组成的组织微阵列来验证这些结果。与scRNA-seq结果一致,这表明在每种癌症类型中,外膜和肌成纤维细胞是主要的亚群
Figure 5E-F All_TMA.data.df.Fibroblasts <- All_TMA.data.df %>% filter(Cell.type2 %in % c("Alveolar" , "Adventitial" , "Myo" )) table(All_TMA.data.df.Fibroblasts$Cell.type2) All_TMA.data.df.Fibroblasts$Cell.type2 <- factor( as.character(All_TMA.data.df.Fibroblasts$Cell.type2), levels = c("Adventitial" , "Alveolar" , "Myo" ) ) dt <- as.table(as.matrix(table(All_TMA.data.df.Fibroblasts$Core_ID, All_TMA.data.df.Fibroblasts$Cell.type2))) Sample.pct_long <- as.data.frame(dt/rowSums(dt)*100 ) CoreData_long <- merge(MxIHC_TMA_metaData, Sample.pct_long, by.x = "Core_ID" , by.y = "Var1" ) names(CoreData_long)[names(CoreData_long) == "Freq" ] <- "Core.pct" names(CoreData_long)[names(CoreData_long) == "Var2" ] <- "Fibs_SubPop" CoreData_long$Group <- factor(CoreData_long$Structure, levels = unique(CoreData_long$Structure)[c(3 ,5 ,2 ,7 ,6 ,4 ,1 )], labels = c("Pancreas" , "Oral" , "Colon" , "Lung" , "Skin" , "Breast" , "Kidney" )) names(CoreData_long)#箱形图显示肿瘤或对照组织中外膜成纤维细胞的相对丰度,通过TMA细胞核的mxIHC分析测定 Fig_5E <- CoreData_long[] %>% drop_na(Structure.filtered) %>% filter(Fibs_SubPop == "Adventitial" ) %>% filter(Structure %in % c("COLON" , "PANCREAS" , "HNSCC" )) %>% ggplot(aes(x = TvN, y = Core.pct)) + theme_pubr(base_size = 15 ) + facet_wrap(~Group) + geom_boxplot(outlier.shape = NA , aes(fill = Fibs_SubPop)) + geom_jitter(alpha = 0.5 , width = 0.2 ) + scale_fill_manual(values = Fibs_col.palette) + rotate_x_text(angle = 45 ) + scale_y_continuous(breaks = c(0 ,25 ,50 , 75 , 100 ), limits = c(0 ,125 )) + stat_compare_means(comparisons = list(c("Normal" , "Tumour" )), size = 3 , label.y = 110 , size = 2.5 ) + ylab("% of all fibroblast per core\n(MxIHC)" ) + theme(axis.title.x = element_blank(), legend.position = "none" ) Fig_5E#箱形图显示肿瘤或对照组织中肌成纤维细胞的相对丰度,通过TMA细胞核的mxIHC分析测定 Fig_5F <- CoreData_long[] %>% drop_na(Structure.filtered) %>% filter(Fibs_SubPop == "Myo" ) %>% filter(Structure %in % c("COLON" , "PANCREAS" , "HNSCC" )) %>% ggplot(aes(x = TvN, y = Core.pct)) + theme_pubr(base_size = 15 ) + facet_wrap(~Group) + geom_boxplot(outlier.shape = NA , aes(fill = Fibs_SubPop)) + geom_jitter(alpha = 0.5 , width = 0.2 ) + scale_fill_manual(values = Fibs_col.palette) + rotate_x_text(angle = 45 ) + scale_y_continuous(breaks = c(0 ,25 ,50 , 75 , 100 ), limits = c(0 ,125 )) + stat_compare_means(comparisons = list(c("Normal" , "Tumour" )), size = 3 , #label = "p.signif", method = "wilcox", label.y = 110 , size = 2.5 ) + ylab("% of all fibroblast per core\n(MxIHC)" ) + theme(axis.title.x = element_blank(), legend.position = "none" ) Fig_5F Fig_5EF <- ggarrange(Fig_5E, Fig_5F, nrow = 2 , align = "v" )
Fig_5EF Figure 5 正如在非小细胞肺癌中发现的那样,与所有三种肿瘤类型的肿瘤组织相比,对照组中上皮成纤维细胞的丰度明显更高,而肿瘤组织中肌成纤维细胞的丰度更高。为了测试肺泡表型是否对肺纤维化具有特异性,对特发性肺纤维化(IPF)样本中产生的scRNA-seq数据进行了类似的分析,结果表明所有三个亚群都具有高概率得分,值得注意的是,该分析还显示,IPF中与肌成纤维细胞分类相关的概率低于癌症数据集,这表明癌症和纤维化中发现的肌成纤维细胞之间可能存在细微差异。