復現PCA原圖之蛋白組學數據

NGS系列文章包括NGS基礎、轉錄組分析?（Nature重磅綜述|關于RNA-seq你想知道的全在這）、ChIP-seq分析?（ChIP-seq基本分析流程）、單細胞測序分析?(重磅綜述：三萬字長文讀懂單細胞RNA測序分析的最佳實踐教程 （原理、代碼和評述）)、DNA甲基化分析、重測序分析、GEO數據挖掘（典型醫學設計實驗GEO數據分析 (step-by-step) - Limma差異分析、火山圖、功能富集）等內容。

2020年4月14日，Sanger研究團隊于nature communication在線發表了題為Single-cell transcriptomics identifies an effectorness gradient shaping the response of CD4+ T cells to cytokines的研究內容，作者使用蛋白質組學、bulk RNA-seq和單細胞轉錄組測序對人體40,000個以上的na?ve and memory CD4+ T cells進行分析，發現細胞類型之間的細胞因子反應差異很大。memory T細胞不能分化為Th2表型，但可以響應iTreg極化獲得類似Th17的表型。單細胞分析表明，T細胞構成了一個轉錄連續體（transcriptional continuum），從幼稚到中樞和效應記憶T細胞，形成了一種效應梯度，并伴隨著趨化因子和細胞因子表達的增加。最后，作者表明，T細胞活化和細胞因子反應受效應梯度的影響。

該文獻通過蛋白質組學（(液相色譜-串聯質譜法，LC-MS/MS）進行了探索性分析，樣品對應于從健康個體的外周血中分離的幼稚和記憶T細胞，并用多種細胞因子刺激5天，每個條件平均3個生物學重復。

這次復現Fig1cPCA圖和Fig2aPCA圖的另一部分，這次作者是通過蛋白組學數據進行PCA的展現：

以上是Fig1c原圖，圖注為“PCA plots from the whole transcriptome of TN and TM cells. Different colors correspond to cell types and different shades to stimulation time points. PCA plots were derived using 21 naive and 19 memory T cell samples for proteomics”

以上為Fig 2a原圖，圖注為“PCA plot from the full transcriptome of TN and TM cells following five days of cytokine stimulations. Only stimulated cells were included in this analysis. PCA plots were derived using 18 naive and 17 memory T cells samples ”

我們需要復現該圖之前，先需要下載數據，可以點擊https://www.opentargets.org/projects/effectorness對proteomics的abundances數據和metadata數據進行下載，然后進行以下步驟：

library(SummarizedExperiment) library(annotables) library(rafalib) library(ggplot2) library(ggrepel) library(limma)

加載數據

加載標準化后的豐度：

MassSpec_data <- read.table("NCOMMS-19-7936188_MassSpec_scaled_abundances.txt", header = T, stringsAsFactors = F) View(MassSpec_data) #從以上可以看出，每列除了代表每個樣本外，前三列分別為Protein_id，Gene_id和Gene_name，每行代表一個蛋白

建立SummarizedExperiment object

創建帶有蛋白質注釋的dataframe

protein_annotations <- data.frame(MassSpec_data[,c("Protein_id","Gene_id","Gene_name")], row.names = MassSpec_data$Gene_name) rownames(MassSpec_data) <- MassSpec_data$Gene_name#構成一個由"Protein_id","Gene_id","Gene_name"的數據框 MassSpec_data <- MassSpec_data[,-c(1:3)]

創建帶有sample注釋的dataframe

sample_ids <- colnames(MassSpec_data) sample_annotations <- data.frame(row.names = sample_ids, donor_id = sapply(sample_ids, function(x){strsplit(x, split = "_")[[1]][1]}), cell_type = paste("CD4", sapply(sample_ids, function(x){strsplit(x, split = "_")[[1]][3]}), sep="_"), cytokine_condition = sapply(sample_ids, function(x){strsplit(x, split = "_")[[1]][4]}), stringsAsFactors = T) sample_annotations$activation_status <- ifelse(sample_annotations$cytokine_condition == "resting", "Resting", "Activated") View(sample_annotations)

創建relevant metadata的變量

meta <- list( Study="Mapping cytokine induced gene expression changes in human CD4+ T cells", Experiment="Quantitative proteomics (LC-MS/MS) panel of cytokine induced T cell polarisations", Laboratory="Trynka Group, Wellcome Sanger Institute", Experimenter=c("Eddie Cano-Gamez", "Blagoje Soskic", "Deborah Plowman"), Description="To study cytokine-induced cell polarisation, we isolated human naive and memory CD4+ T cells in triplicate from peripheral blood of healthy individuals. Next, we polarised the cells with different cytokine combinations linked to autoimmunity and performed LC-MS/MS.", Methdology="LC-MS/MS with isobaric labelling", Characteristics="Data type: Normalised, scaled protein abundances", Date="September, 2019", URL="https://doi.org/10.1101/753731" )

建立SummarizedExperiment object

proteomics_data <- SummarizedExperiment(assays=list(counts=as.matrix(MassSpec_data)), colData=sample_annotations, rowData=protein_annotations, metadata=meta) saveRDS(proteomics_data, file="proteinAbundances_summarizedExperiment.rds")

數據可視化

將NA值設置為零
注意：此操作僅出于可視化目的。執行統計測試時，NA不會設置為零。

assay(proteomics_data)[is.na(assay(proteomics_data))] <- 0

定義函數：

提取蛋白質表達值;

進行主成分分析;

返回一個矩陣，其中包含每個樣品和樣品注釋的PC坐標;

返回每個主要成分解釋的方差百分比。

getPCs <- function(exp){ pcs <- prcomp(t(assay(exp))) pVar <- pcs$sdev^2/sum(pcs$sdev^2) pca.mat <- data.frame(pcs$x) pca.mat$donor_id <- colData(exp)$donor_id pca.mat$cell_type <- colData(exp)$cell_type pca.mat$cytokine_condition <- colData(exp)$cytokine_condition pca.mat$activation_status <- colData(exp)$activation_status res <- list(pcs = pca.mat, pVar=pVar) return(res) }

對所有樣本執行PCA

pcs <- getPCs(proteomics_data)ggplot(data=pcs$pcs, aes(x=PC1, y=PC2, color=cell_type, shape=activation_status)) + geom_point(size = 8) + xlab(paste0("PC1:", round(pcs$pVar[1]*100), "% variance")) + ylab(paste0("PC2: ", round(pcs$pVar[2]*100), "% variance")) + scale_colour_manual(values = c("#5AB4AC","#AF8DC3")) + scale_alpha_discrete(range = c(0.5,1)) + coord_fixed() + theme_bw() + theme(panel.grid = element_blank())

去掉個體間變異性：

proteomics_data_regressed <- proteomics_data assay(proteomics_data_regressed) <- removeBatchEffect(assay(proteomics_data_regressed), batch = factor(as.vector(colData(proteomics_data_regressed)$donor_id)) )

重新計算PCA：

pcs <- getPCs(proteomics_data_regressed)ggplot(data=pcs$pcs, aes(x=PC1, y=PC2, color=cell_type, shape=activation_status)) + geom_point(size = 8) + xlab(paste0("PC1:", round(pcs$pVar[1]*100), "% variance")) + ylab(paste0("PC2: ", round(pcs$pVar[2]*100), "% variance")) + scale_colour_manual(values = c("#5AB4AC","#AF8DC3")) + scale_alpha_discrete(range = c(0.5,1)) + coord_fixed() + theme_bw() + theme(panel.grid = element_blank())

原圖

細胞類型特異性分析

將naive和memory T細胞樣本分為僅包含受刺激細胞的兩個不同數據集。

proteomics_data_naive <- proteomics_data[,(proteomics_data$cell_type=="CD4_naive") & (proteomics_data$activation_status=="Activated")] proteomics_data_memory <- proteomics_data[,(proteomics_data$cell_type=="CD4_memory") & (proteomics_data$activation_status=="Activated")]

Naive T cells

對 5 days-stimulated naive T cells進行PCA:

pcs_naive <- getPCs(proteomics_data_naive) ggplot(data=pcs_naive$pcs, aes(x=PC1, y=PC2)) + geom_point(aes(color=donor_id), size=8) + xlab(paste0("PC1:", round(pcs_naive$pVar[1]*100), "% variance")) + ylab(paste0("PC2: ", round(pcs_naive$pVar[2]*100), "% variance")) + coord_fixed() + theme_bw() + theme(plot.title=element_text(size=20, hjust=0.5), axis.title=element_text(size=14), panel.grid = element_blank(), axis.text=element_text(size=12),legend.text=element_text(size=12), legend.title=element_text(size=12), legend.key.size = unit(1.5,"lines"))

去掉個體間變異性：

assay(proteomics_data_naive) <- removeBatchEffect(assay(proteomics_data_naive), batch = factor(as.vector(colData(proteomics_data_naive)$donor_id)) ) pcs_naive <- getPCs(proteomics_data_naive) ggplot(data=pcs_naive$pcs, aes(x=PC1, y=PC2, color=cytokine_condition)) + geom_point(size = 8) + geom_label_repel(aes(label=cytokine_condition, color=cytokine_condition)) + xlab(paste0("PC1: ", round(pcs_naive$pVar[1]*100), "% variance")) + ylab(paste0("PC2: ", round(pcs_naive$pVar[2]*100), "% variance")) + scale_colour_brewer(palette = "Dark2") + scale_fill_brewer(palette = "Dark2") + coord_fixed() + theme_bw() + theme(panel.grid = element_blank(), legend.position = "none")

刪除由PCA標識的異常樣本:

proteomics_data_naive <- proteomics_data_naive[, colnames(proteomics_data_naive) != "D257_CD4_naive_Th1"]pcs_naive <- getPCs(proteomics_data_naive)ggplot(data=pcs_naive$pcs, aes(x=PC1, y=PC2, color=cytokine_condition)) + geom_point(size = 8) + geom_label_repel(aes(label=cytokine_condition, color=cytokine_condition)) + xlab(paste0("PC1: ", round(pcs_naive$pVar[1]*100), "% variance")) + ylab(paste0("PC2: ", round(pcs_naive$pVar[2]*100), "% variance")) + scale_colour_brewer(palette = "Dark2") + scale_fill_brewer(palette = "Dark2") + coord_fixed() + theme_bw() + theme(panel.grid = element_blank(), legend.position = "none")

原圖

Memory T cells

again。。。

Performing PCA on 5 days-stimulated memory T cells only. ```{r compute_pca_naive, message=FALSE, warning=FALSE} pcs_memory <- getPCs(proteomics_data_memory)ggplot(data=pcs_memory$pcs, aes(x=PC1, y=PC2)) + geom_point(aes(color=donor_id), size=8) + xlab(paste0("PC1:", round(pcs_memory$pVar[1]*100), "% variance")) + ylab(paste0("PC2: ", round(pcs_memory$pVar[2]*100), "% variance")) + coord_fixed() + theme_bw() + theme(plot.title=element_text(size=20, hjust=0.5), axis.title=element_text(size=14), panel.grid = element_blank(), axis.text=element_text(size=12),legend.text=element_text(size=12), legend.title=element_text(size=12), legend.key.size = unit(1.5,"lines"))

Regressing out inter-individual variability

assay(proteomics_data_memory) <- removeBatchEffect(assay(proteomics_data_memory), batch = factor(as.vector(colData(proteomics_data_memory)$donor_id)) )

再次計算PCs

pcs_memory <- getPCs(proteomics_data_memory)ggplot(data=pcs_memory$pcs, aes(x=PC1, y=PC2, color=cytokine_condition)) + geom_point(size = 8) + geom_label_repel(aes(label=cytokine_condition, color=cytokine_condition)) + xlab(paste0("PC1: ", round(pcs_memory$pVar[1]*100), "% variance")) + ylab(paste0("PC2: ", round(pcs_memory$pVar[2]*100), "% variance")) + scale_colour_brewer(palette = "Dark2") + scale_fill_brewer(palette = "Dark2") + coord_fixed() + theme_bw() + theme(panel.grid = element_blank(), legend.position = "none")

原圖

基本分布還是差不多的，，，，

快去試一試呀！

你可能還想看

• 蛋白質組學研究概述

• PCA主成分分析實戰和可視化 附R代碼和測試數據

• 用了這么多年的PCA可視化竟然是錯的！！！

• 什么？你做的差異基因方法不合適？

• NBT：單細胞轉錄組新降維可視化方法PHATE

往期精品(點擊圖片直達文字對應教程)

后臺回復“生信寶典福利第一波”或點擊閱讀原文獲取教程合集

總結

以上是生活随笔為你收集整理的这篇Nature子刊文章的蛋白组学数据PCA分析竟花费了我两天时间来重现|附全过程代码...的全部內容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網站內容還不錯，歡迎將生活随笔推薦給好友。