Follow Nat Med. Learn to map | GSVA + limma differential pathway analysis + divergent bar chart

Follow Nat Med. Learn to map | GSVA + limma differential pathway analysis + divergent bar chart

Lambrechts D, Wauters E, Boeckx B, et al. Phenotype molding of stromal cells in the lung tumor microenvironment[J]. Nat Med,  2018, 24(8): p. 1277-1289.

Recently, many students have said in the background that they want to talk about this picture, so I will write it briefly today.

Gene Set Variation Analysis(GSVA)Known as Gene Set Variation Analysis, it is a non-parametric, unsupervised analysis method primarily used to evaluate gene set enrichment results for microarrays and transcriptomes. It is mainly to evaluate whether different metabolic pathways are enriched in different samples by converting 基因the matrix between different samples 表达量into 基因集the expression matrix between samples. It is often used to compare the differential gene sets of different cell types in single-cell transcriptomes .

Schematic diagram of the GSVA process. doi: 10.1186/1471-2105-14-7

Sample data and code collection

• Appreciate this tweet and forward it to the circle 10￥of friends .点赞在看集赞5个

• Canoe notes 2022年VIP会员can be obtained for free.

  About Canoe Notes 2022 VIP Membership Program

  rights and interests:

  1. 2022年度木舟笔记所有推文示例数据及代码(含21年,除Cell合集外资源。
  2. 木舟笔记科研交流群。
  3. 半价购买`跟着Cell学作图系列合集`[(免费教程+代码领取)|跟着Cell学作图系列合集](http://mp.weixin.qq.com/s?__biz=MzIxMDExNDE0OQ==&mid=2247486072&idx=1&sn=d579a62420f722285eac318c4114d199&chksm=9768c972a01f4064a84f30fd48b8c2107af543c31b6adb5f36234b1377ac42c8af54f45efda9#rd)。

  toll:

  99￥/person. You can add WeChat: mzbj0002transfer money, or give a reward directly at the end of the article.

GSVA

Import sample data

## 为正常和肿瘤的内皮细胞的基因表达矩阵
library(readxl)
library(dplyr)
dat <- read_excel("data_test.xlsx") 
dat <- dat %>% data.frame()
row.names(dat) <- dat$gene
dat <- dat[,-1]
head(dat)

> head(dat)
              EC1_T EC2_T EC3_T EC0_N EC1_N EC3_N EC4_N
RP11-34P13.3      0     0     0     0     0     0     0
FAM138A           0     0     0     0     0     0     0
OR4F5             0     0     0     0     0     0     0
RP11-34P13.7      0     0     0     0     0     0     0
RP11-34P13.8      0     0     0     0     0     0     0
RP11-34P13.14     0     0     0     0     0     0     0

Reference gene set database

For details on MSigDB database, see: Q&A | How to use clusterProfiler to perform enrichment analysis on MSigDB database

Here we use: hallmark gene sets

start analysis

#BiocManager::install("GSEABase")
BiocManager::install("GSVA", version = "3.14") # R 4.1.2 注意版本w
library('GSEABase')
library(GSVA)
geneSets <- getGmt('h.all.v7.5.1.symbols.gmt')    ###下载的基因集
GSVA_hall <- gsva(expr=as.matrix(dat), 
                  gset.idx.list=geneSets, 
                  mx.diff=T, # 数据为正态分布则T，双峰则F
                  kcdf="Gaussian", #CPM, RPKM, TPM数据就用默认值"Gaussian"， read count数据则为"Poisson"，
                  parallel.sz=4) # 并行线程数目
head(GSVA_hall)

> head(GSVA_hall)
                                          EC1_T       EC2_T      EC3_T       EC0_N
HALLMARK_TNFA_SIGNALING_VIA_NFKB    -0.43490528 -0.36929145 -0.4694267  0.57790329
HALLMARK_HYPOXIA                    -0.19830871 -0.15040810 -0.1237437  0.31595670
HALLMARK_CHOLESTEROL_HOMEOSTASIS    -0.15223695 -0.15160770 -0.0505465  0.24273366
HALLMARK_MITOTIC_SPINDLE            -0.29757233  0.20140000  0.3185932 -0.09284047
HALLMARK_WNT_BETA_CATENIN_SIGNALING -0.08827878  0.03086390  0.2383694  0.24045204
HALLMARK_TGF_BETA_SIGNALING         -0.26357054 -0.01068719  0.3041132  0.28761931
                                         EC1_N      EC3_N       EC4_N
HALLMARK_TNFA_SIGNALING_VIA_NFKB    -0.1673606  0.2198982  0.33940521
HALLMARK_HYPOXIA                    -0.3292568  0.0906295  0.01407149
HALLMARK_CHOLESTEROL_HOMEOSTASIS    -0.3566404  0.1532081  0.05151732
HALLMARK_MITOTIC_SPINDLE            -0.3673806  0.1091566 -0.15295077
HALLMARK_WNT_BETA_CATENIN_SIGNALING -0.2352282 -0.1027337 -0.15598311
HALLMARK_TGF_BETA_SIGNALING         -0.4387025  0.2336080 -0.14090342

limma differential pathway analysis

## limma
#BiocManager::install('limma')
library(limma)
# 设置或导入分组
group <- factor(c(rep("Tumor", 3), rep("Normal", 4)), levels = c('Tumor', 'Normal'))
design <- model.matrix(~0+group)
colnames(design) = levels(factor(group))
rownames(design) = colnames(GSVA_hall)
design
# Tunor VS Normal
compare <- makeContrasts(Tumor - Normal, levels=design)
fit <- lmFit(GSVA_hall, design)
fit2 <- contrasts.fit(fit, compare)
fit3 <- eBayes(fit2)
Diff <- topTable(fit3, coef=1, number=200)
head(Diff)

> head(Diff)
                                        logFC      AveExpr         t      P.Value
HALLMARK_INTERFERON_GAMMA_RESPONSE -0.7080628 -0.021269395 -4.393867 0.0002713325
HALLMARK_INTERFERON_ALPHA_RESPONSE -0.7419587 -0.044864170 -4.299840 0.0003385128
HALLMARK_INFLAMMATORY_RESPONSE     -0.5940473 -0.003525139 -4.193096 0.0004352397
HALLMARK_IL6_JAK_STAT3_SIGNALING   -0.6117943 -0.038379008 -4.157977 0.0004727716
HALLMARK_TNFA_SIGNALING_VIA_NFKB   -0.6670027 -0.043396767 -4.149307 0.0004825257
HALLMARK_ALLOGRAFT_REJECTION       -0.4645747  0.015248663 -3.278981 0.0036971108
                                     adj.P.Val           B
HALLMARK_INTERFERON_GAMMA_RESPONSE 0.004825257  0.43891329
HALLMARK_INTERFERON_ALPHA_RESPONSE 0.004825257  0.22763343
HALLMARK_INFLAMMATORY_RESPONSE     0.004825257 -0.01221232
HALLMARK_IL6_JAK_STAT3_SIGNALING   0.004825257 -0.09109393
HALLMARK_TNFA_SIGNALING_VIA_NFKB   0.004825257 -0.11056468
HALLMARK_ALLOGRAFT_REJECTION       0.030809257 -2.03863951

Divergent bar graph drawing

## barplot
dat_plot <- data.frame(id = row.names(Diff),
                       t = Diff$t)
# 去掉"HALLMARK_"
library(stringr)
dat_plot$id <- str_replace(dat_plot$id , "HALLMARK_","")
# 新增一列 根据t阈值分类
dat_plot$threshold = factor(ifelse(dat_plot$t  >-2, ifelse(dat_plot$t >= 2 ,'Up','NoSignifi'),'Down'),levels=c('Up','Down','NoSignifi'))
# 排序
dat_plot <- dat_plot %>% arrange(t)
# 变成因子类型
dat_plot$id <- factor(dat_plot$id,levels = dat_plot$id)
# 绘制
library(ggplot2)
library(ggtheme)
# install.packages("ggprism")
library(ggprism)
p <- ggplot(data = dat_plot,aes(x = id,y = t,fill = threshold)) +
  geom_col()+
  coord_flip() +
  scale_fill_manual(values = c('Up'= '#36638a','NoSignifi'='#cccccc','Down'='#7bcd7b')) +
  geom_hline(yintercept = c(-2,2),color = 'white',size = 0.5,lty='dashed') +
  xlab('') + 
  ylab('t value of GSVA score, tumour versus non-malignant') + #注意坐标轴旋转了
  guides(fill=F)+ # 不显示图例
  theme_prism(border = T) +
  theme(
    axis.text.y = element_blank(),
    axis.ticks.y = element_blank()
    )
p
# 添加标签
# 此处参考了：https://mp.weixin.qq.com/s/eCMwWCnjTyQvNX2wNaDYXg
# 小于-2的数量
low1 <- dat_plot %>% filter(t < -2) %>% nrow()
# 小于0总数量
low0 <- dat_plot %>% filter( t < 0) %>% nrow()
# 小于2总数量
high0 <- dat_plot %>% filter(t < 2) %>% nrow()
# 总的柱子数量
high1 <- nrow(dat_plot)

# 依次从下到上添加标签
p <- p + geom_text(data = dat_plot[1:low1,],aes(x = id,y = 0.1,label = id),
              hjust = 0,color = 'black') + # 小于-1的为黑色标签
  geom_text(data = dat_plot[(low1 +1):low0,],aes(x = id,y = 0.1,label = id),
            hjust = 0,color = 'grey') + # 灰色标签
  geom_text(data = dat_plot[(low0 + 1):high0,],aes(x = id,y = -0.1,label = id),
            hjust = 1,color = 'grey') + # 灰色标签
  geom_text(data = dat_plot[(high0 +1):high1,],aes(x = id,y = -0.1,label = id),
            hjust = 1,color = 'black') # 大于1的为黑色标签
ggsave("gsva_bar.pdf",p,width = 8,height  = 8)

Results display

Past

Mapping with Cell Learning | Proteomaps Diagram

---

- END -