##### Scripts used in Project "Differential expression of miRNAs in pancreatobiliary type of periampullary adenocarcinoma and its associated stroma"
##### Written by Vandana Sandhu


#### Script 1
#### Script for moderated t test


tum=read.table("Tumor.txt",sep="\t")    ## File having tumor expression data
str=read.table("Stroma.txt",sep="\t")   ## File having stroma expression data 

genes=read.table("miRNA-names.txt",sep="\t")


tum1=tum[2:nrow(tum),2:ncol(tum)]
str1=str[2:nrow(str),2:ncol(str)]

gp1=as.matrix(tum1)
gp2=as.matrix(str1)


dat=cbind(gp1,gp2)
dim(dat)
x1=ncol(gp1)
x2=ncol(gp2)

design <- model.matrix(~ 0+factor(c(rep(1,x1),rep(2,x2))))   ### design matrix for fitting in lmfit function
colnames(design)<-c("Tumor","Stroma")
library(limma)
fit<-lmFit(dat,design)
contrast.matrix<-makeContrasts(Tumor-Stroma,levels=design)
fit2<-contrasts.fit(fit,contrast.matrix)                        
fit2<-eBayes(fit2)
tp=topTable(fit2,coef=1,genelist=genes,adjust="fdr")     ### toptable function carries out moderated t-test and fdr correcteion using benjamini and hochberg correction is done
write.table(tp,"tumor-stroma.txt")


#########################################################################
############

###### Script 2
###### Script for hierarchical clustering of samples

xx<- read.table("miRNA_expression_data", sep="\t")
xmat<-xx[2:nrow(xx) ,2:ncol(xx)]
xmat <- sapply(xmat, function(xx) as.numeric(as.character(xx)))
xmat<-data.matrix(xmat)
c<-xx[1, 2:ncol(xx)]
cx=c
r<-xx[2:nrow(xx), 1]
rx=r
a.sample <- c(rep("Normal",8),rep("Carcinoma",18),rep("Stroma",18))

f.sample <- factor(a.sample)
vec.sample <- c("grey","red","blue")
a.color <- rep(0,length(f.sample))
for(i in 1:length(f.sample))
a.color[i] <- vec.sample[ f.sample[i]==levels(f.sample) ]

mydatamatrix <- data.matrix(xmat) 
mydatascale <- t(scale(t(mydatamatrix)))            ## Scaling the data

hr <- hclust(as.dist(1-cor(t(mydatascale), method="pearson")), method="complete")  # Calculating pearson score row-wise
hc <- hclust(as.dist(1-cor(mydatascale, method="pearson")), method="complete")      # Calculating pearson score column-wise
myclhr <- cutree(hr, h=max(hr$height)/2); mycolhr <- sample(rainbow(256)); 
myclhc <- cutree(hc, h=max(hc$height)/2); mycolhc <- sample(rainbow(256)); 
mycolhr <- mycolhr[as.vector(myclhr)]; 
mycolhc <- mycolhc[as.vector(myclhc)]; 

library("gplots") 
library("RSvgDevice") 
pdf("Plot.pdf")
heatmap.2(mydatamatrix , Rowv=as.dendrogram(hr),Colv=as.dendrogram(hc), dendrogram="col", 
scale="row", col=greenred, trace="none", density.info="none", ColSideColors=a.color,
key=TRUE, keysize=1.0, margin=c(5,8), sepcolor = "",
labCol=c ,labRow=r) 

dev.off()

#########################################################################
############

###### Script 3
###### Script for sparce PCA of samples

xx<- read.table("mirna_expression_data_input.txt", sep="\t")
xmat<-xx[2:nrow(xx)  ,2:ncol(xx)]
xmat <- sapply(xmat, function(xx) as.numeric(as.character(xx)))
dd=t(xmat)

cv.out=SPC.cv(dd, sumabsvs=seq(1, 13,len=13), nfolds=5, niter=5, v=NULL,     ### 5 fold Cross validaton for different sumabsv values range; it must be between 1 and square root of number of columns of data
       trace=TRUE, orth=FALSE, center=TRUE, vpos=FALSE, vneg=FALSE)

print(cv.out)
plot(cv.out)


#### Best sumabsv value (lowest CV error):  5 
##Smallest sumabsv value that has CV error within 1 SE of best CV error:  4 

library("PMA")
out <- SPC(dd,sumabsv=4, K=1)     ### Used sumabsv equlas to 4 because it gives Smallest sumabsv value that has CV error within 1 SE of best CV error

x=c(rep(15,8),rep(16,18),rep(17,18))    #### For giving different shapes to samples
cc=c(rep("grey",8),rep("red",18),rep("blue",18))

plot(out$u,col=cc,pch=x,cex=1)

#########################################################################
############

###### Script 4
###### Script for boxplots and one way ANNOVA 




dd1=read.table("anova-tumor.txt",sep="\t",header=TRUE)    ### Tumor expression data
dd2=read.table("anova-stroma.txt",sep="\t",header=TRUE)   ### Stroma expression data
dd3=read.table("anova-normal.txt",sep="\t",header=TRUE)   ### Normal expression data


cc<-c(dd1[,1])   #miRNA names
j=0
p <- vector()
for( i in 2:44){
dd_matrix<- data.frame(wt=c(dd1[,i],dd2[,i],dd3[,i]),type=factor(c(rep("2.Carcinoma",18),rep("1.Stroma",18),rep("3.Normal",8))))
sapply(split(dd_matrix$wt,dd_matrix$type),mean)
sapply(split(dd_matrix$wt,dd_matrix$type),var)
sqrt(sapply(split(dd_matrix$wt,dd_matrix$type),var))
j=j+1
file=cc[j];

pdf(paste("boxplot/",file,".pdf",sep=""))
#boxplot(split(dd_matrix$wt,dd_matrix$type),xlab="typement",ylab="Pod Weight",col="green")
fitdd_matrix <- lm(wt~type, data=dd_matrix)

zzz=anova(fitdd_matrix)                 ## Annova test
p[j]<-zzz[,5]

boxplot(split(dd_matrix$wt,dd_matrix$type),xlab="",ylab="miRNA expression",col=c("blue","red","grey"))   ### Plotting boxplots
points(rep(1,length(dd_matrix$wt)),dd_matrix$wt)
title(main=paste(cc[j]))
legend("topright",paste(p[j]))


dev.off()
 i=i+1
 dd_matrix <- vector()
 file=vector()
}


write.table(cbind(cc,p),"p_val.txt");
adjust_p=p.adjust(p)
write.table(cbind(cc,adjust_p),"adjust_p_val.txt");