11. 对第10步得到的表达矩阵进行探索，先画第一个样本的所有基因的表达量的boxplot,hist,density ， 然后画所有样本的 这些图
- 参考：http://bio-info-trainee.com/tmp/basic_visualization_for_expression_matrix.html
- 理解ggplot2的绘图语法，数据和图形元素的映射关系
### ggplot2
library(ggplot2)
p=ggplot(exprSet_L,aes(x=sample,y=value,fill=group))+geom_boxplot()
print(p)
p=ggplot(exprSet_L,aes(x=sample,y=value,fill=group))+geom_violin()
print(p)
p=ggplot(exprSet_L,aes(value,fill=group))+geom_histogram(bins = 200)+facet_wrap(~sample, nrow = 4)
print(p)
p=ggplot(exprSet_L,aes(value,col=group))+geom_density()+facet_wrap(~sample, nrow = 4)
print(p)
p=ggplot(exprSet_L,aes(value,col=group))+geom_density()
print(p)
p=ggplot(exprSet_L,aes(x=sample,y=value,fill=group))+geom_boxplot()
p=p+stat_summary(fun.y="mean",geom="point",shape=23,size=3,fill="red")
p=p+theme_set(theme_set(theme_bw(base_size=20)))
p=p+theme(text=element_text(face='bold'),axis.text.x=element_text(angle=30,hjust=1),axis.title=element_blank())
print(p)
image
image
image
image
image
image
12.理解统计学指标mean,median,max,min,sd,var,mad并计算出每个基因在所有样本的这些统计学指标，最后按照mad值排序，取top 50 mad值的基因，得到列表。
注意：这个题目出的并不合规，请仔细看。
g_mean <- tail(sort(apply(exprSet,1,mean)),50)
g_median <- tail(sort(apply(exprSet,1,median)),50)
g_max <- tail(sort(apply(exprSet,1,max)),50)
g_min <- tail(sort(apply(exprSet,1,min)),50)
g_sd <- tail(sort(apply(exprSet,1,sd)),50)
g_var <- tail(sort(apply(exprSet,1,var)),50)
g_mad <- tail(sort(apply(exprSet,1,mad)),50)
g_mad
names(g_mad)
[1] "DUSP5" "IGFBP4" "H1FX" "ENPP2" "FLNA" "CLEC2B" "TSPYL2" "ZNF266" "S100A9" "NR4A2" "TGFBI" "ARF6" "APBB2" "VCAN" "RBM38"
[16] "CAPG" "PLXNC1" "RGS2" "RNASE6" "VAMP5" "CYBB" "GNLY" "CCL3" "OAS1" "ENPP2" "TRIB2" "ZNF804A" "H1FX" "IGH" "JUND"
[31] "SLC25A1" "PCDH9" "VIPR1" "COBLL1" "GUSBP11" "S100A8" "HBB" "FOS" "LHFPL2" "FCN1" "ZAP70" "IGLC1" "LGALS1" "HBB" "FOS"
[46] "SLAMF1" "TCF7" "DMD" "IGF2BP3" "FAM30A"
13.根据第12步骤得到top 50 mad值的基因列表来取表达矩阵的子集，并且热图可视化子表达矩阵。试试看其它5种热图的包的不同效果。
## heatmap
library(pheatmap)
choose_gene=names(tail(sort(apply(exprSet,1,mad)),50))
choose_matrix=exprSet[choose_gene,]
choose_matrix=t(scale(t(choose_matrix)))
pheatmap(choose_matrix)
image
14.取不同统计学指标mean,median,max,mean,sd,var,mad的各top50基因列表，使用UpSetR包来看他们之间的overlap情况。
## UpSetR
# https://cran.r-project.org/web/packages/UpSetR/README.html
library(UpSetR)
g_all <- unique(c(names(g_mean),names(g_median),names(g_max),names(g_min),
names(g_sd),names(g_var),names(g_mad) ))
dat=data.frame(g_all=g_all,
g_mean=ifelse(g_all %in% names(g_mean) ,1,0),
g_median=ifelse(g_all %in% names(g_median) ,1,0),
g_max=ifelse(g_all %in% names(g_max) ,1,0),
g_min=ifelse(g_all %in% names(g_min) ,1,0),
g_sd=ifelse(g_all %in% names(g_sd) ,1,0),
g_var=ifelse(g_all %in% names(g_var) ,1,0),
g_mad=ifelse(g_all %in% names(g_mad) ,1,0)
)
upset(dat,nsets = 7)
image
15.在第二步的基础上面提取CLL包里面的data(sCLLex) 数据对象的样本的表型数据。
pdata=pData(sCLLex)
group_list=as.character(pdata[,2])
group_list
# [1] "progres." "stable" "progres." "progres." "progres." "progres." "stable" "stable" "progres." "stable" "progres." "stable" "progres." "stable"
# [15] "stable" "progres." "progres." "progres." "progres." "progres." "progres." "stable"
dim(exprSet)
# [1] 8585 22
exprSet[1:5,1:5]
# CLL11.CEL CLL12.CEL CLL13.CEL CLL14.CEL CLL15.CEL
MAPK3 5.743132 6.219412 5.523328 5.340477 5.229904
TIE1 2.285143 2.291229 2.287986 2.295313 2.662170
CYP2C19 3.309294 3.318466 3.354423 3.327130 3.365113
CXCR5 7.544884 7.671801 7.474025 7.152482 6.902932
DUSP1 5.083793 7.610593 7.631311 6.518594 5.059087
16.对所有样本的表达矩阵进行聚类并且绘图，然后添加样本的临床表型数据信息(更改样本名)
## hclust
colnames(exprSet)=paste(group_list,1:22,sep='')
# Define nodePar
nodePar <- list(lab.cex = 0.6, pch = c(NA, 19),
cex = 0.7, col = "blue")
hc=hclust(dist(t(exprSet)))
par(mar=c(5,5,5,10))
plot(as.dendrogram(hc), nodePar = nodePar, horiz = TRUE)
image
17.对所有样本的表达矩阵进行PCA分析并且绘图，同样要添加表型信息。
# install.packages("ggfortify")
library(ggfortify)
exprSet <- exprs(sCLLex)
df <- as.data.frame(t(exprSet))
df$group <- group_list
# autoplot uses ggplot2 to draw a particular plot for an object of a particular class in a single command.
autoplot(prcomp(df[,1:(ncol(df)-1)]), data=df, colour = 'group')
image
18.根据表达矩阵及样本分组信息进行批量T检验，得到检验结果表格
## t.test
dat = exprSet
group_list=as.factor(group_list)
group1 = which(group_list == levels(group_list)[1])
group2 = which(group_list == levels(group_list)[2])
dat1 = dat[, group1]
dat2 = dat[, group2]
dat = cbind(dat1, dat2)
pvals = apply(exprSet, 1, function(x){
t.test(as.numeric(x)~group_list)$p.value
})
p.adj = p.adjust(pvals, method = "BH")
avg_1 = rowMeans(dat1)
avg_2 = rowMeans(dat2)
log2FC = avg_2-avg_1
DEG_t.test = cbind(avg_1, avg_2, log2FC, pvals, p.adj)
DEG_t.test=DEG_t.test[order(DEG_t.test[,4]),]
DEG_t.test=as.data.frame(DEG_t.test)
head(DEG_t.test)
# avg_1 avg_2 log2FC pvals p.adj
36129_at 7.875615 8.791753 0.9161377 1.629755e-05 0.2057566
37676_at 6.622749 7.965007 1.3422581 4.058944e-05 0.2436177
33791_at 7.616197 5.786041 -1.8301554 6.965416e-05 0.2436177
39967_at 4.456446 2.152471 -2.3039752 8.993339e-05 0.2436177
34594_at 5.988866 7.058738 1.0698718 9.648226e-05 0.2436177
32198_at 4.157971 3.407405 -0.7505660 2.454557e-04 0.3516678
19.使用limma包对表达矩阵及样本分组信息进行差异分析，得到差异分析表格，重点看logFC和P值，画个火山图(就是logFC和-log10(P值)的散点图)。
# DEG by limma
suppressMessages(library(limma))
design <- model.matrix(~0+factor(group_list))
colnames(design)=levels(factor(group_list))
rownames(design)=colnames(exprSet)
design
contrast.matrix<-makeContrasts(paste0(unique(group_list),collapse = "-"),levels = design)
contrast.matrix
##这个矩阵声明，我们要把progres.组跟stable进行差异分析比较
##step1
fit <- lmFit(exprSet,design)
##step2
fit2 <- contrasts.fit(fit, contrast.matrix) ##这一步很重要，大家可以自行看看效果
fit2 <- eBayes(fit2) ## default no trend !!!
##eBayes() with trend=TRUE
##step3
tempOutput = topTable(fit2, coef=1, n=Inf)
nrDEG = na.omit(tempOutput)
#write.csv(nrDEG2,"limma_notrend.results.csv",quote = F)
head(nrDEG)
## volcano plot
DEG=nrDEG
logFC_cutoff <- with(DEG,mean(abs( logFC)) + 2*sd(abs( logFC)) )
DEG$change = as.factor(ifelse(DEG$P.Value < 0.05 & abs(DEG$logFC) > logFC_cutoff,
ifelse(DEG$logFC > logFC_cutoff ,'UP','DOWN'),'NOT')
)
this_tile <- paste0('Cutoff for logFC is ',round(logFC_cutoff,3),
'\nThe number of up gene is ',nrow(DEG[DEG$change =='UP',]) ,
'\nThe number of down gene is ',nrow(DEG[DEG$change =='DOWN',])
)
g = ggplot(data=DEG, aes(x=logFC, y=-log10(P.Value), color=change)) +
geom_point(alpha=0.4, size=1.75) +
theme_set(theme_set(theme_bw(base_size=20)))+
xlab("log2 fold change") + ylab("-log10 p-value") +
ggtitle( this_tile ) + theme(plot.title = element_text(size=15,hjust = 0.5))+
scale_colour_manual(values = c('blue','black','red')) ## corresponding to the levels(res$change)
print(g)
image
image
20.对T检验结果的P值和limma包差异分析的P值画散点图，看看哪些基因相差很大。
### different P values
head(nrDEG)
head(DEG_t.test)
DEG_t.test=DEG_t.test[rownames(nrDEG),]
plot(DEG_t.test[,3],nrDEG[,1])
plot(DEG_t.test[,4],nrDEG[,4])
plot(-log10(DEG_t.test[,4]),-log10(nrDEG[,4]))
image
image
image
rownames(exprSet)=ids[match(rownames(exprSet),ids$probe_id),2]
exprSet[1:5,1:5]
exprSet['GAPDH',]
exprSet['ACTB',]
exprSet['DLEU1',]
library(ggplot2)
library(ggpubr)
my_comparisons <- list(
c("stable", "progres.")
)
dat=data.frame(group=group_list,
sampleID= names(exprSet['DLEU1',]),
values= as.numeric(exprSet['DLEU1',]))
ggboxplot(
dat, x = "group", y = "values",
color = "group",
add = "jitter"
)+
stat_compare_means(comparisons = my_comparisons, method = "t.test")
image
## heatmap
library(pheatmap)
choose_gene=head(rownames(nrDEG),25)
choose_matrix=exprSet[choose_gene,]
choose_matrix=t(scale(t(choose_matrix)))
pheatmap(choose_matrix)
image
生信技能树公益视频合辑：学习顺序是linux，r，软件安装，geo，小技巧，ngs组学！
B站链接：https://m.bilibili.com/space/338686099
YouTube链接：https://m.youtube.com/channel/UC67sImqK7V8tSWHMG8azIVA/playlists
生信工程师入门最佳指南：https://mp.weixin.qq.com/s/vaX4ttaLIa19MefD86WfUA
学徒培养：https://mp.weixin.qq.com/s/3jw3_PgZXYd7FomxEMxFmw
生信技能树 - 简书 https://www.jianshu.com/u/d645f768d2d5
