## Run R script using $R CMD BATCH Mya_differential_gene_expression_analysis.R, best to run on high-spec workstation!
## Data file "vs_ex_11_matrix.counts_outliers_removed" available in Supplementary File 2_countmatrix
## Data description: 
# Sample ID	Treatment	Library	Time point
# Vs ex11 1.1C	1 DAY CONTROL	LIB20596	1
# Vs ex11 1.3C	1 DAY CONTROL	LIB20598	1
# Vs ex11 1.1D	1 DAY DAMAGED	LIB20599	1
# Vs ex11 1.2D	1 DAY DAMAGED	LIB20600	1
# Vs ex11 1.4D	1 DAY DAMAGED	LIB20601	1
# Vs ex11 2.1C	3 DAYS CONTROL	LIB20602	2
# Vs ex11 2.2C	3 DAYS CONTROL	LIB20603	2
# Vs ex11 2.3C	3 DAYS CONTROL	LIB20604	2
# Vs ex11 2.1D	3 DAYS DAMAGED	LIB20605	2
# Vs ex11 2.2D	3 DAYS DAMAGED	LIB20606	2
# Vs ex11 2.4D	3 DAYS DAMAGED	LIB20607	2
# Vs ex11 3.1C	5 DAYS CONTROL	LIB20608	3
# Vs ex11 3.4C	5 DAYS CONTROL	LIB20610	3
# Vs ex11 3.1D	5 DAYS DAMAGED	LIB20611	3
# Vs ex11 3.4D	5 DAYS DAMAGED	LIB20613	3
# Vs ex11 4.1C	1 WEEK CONTROL	LIB20614	4
# Vs ex11 4.2C	1 WEEK CONTROL	LIB20615	4
# Vs ex11 4.3C	1 WEEK CONTROL	LIB20616	4
# Vs ex11 4.1D	1 WEEK DAMAGED	LIB20617	4
# Vs ex11 4.2D	1 WEEK DAMAGED	LIB20618	4
# Vs ex11 4.3D	1 WEEK DAMAGED	LIB20619	4
# Vs ex11 5.2C	2 WEEKS CONTROL	LIB20620	5
# Vs ex11 5.3C	2 WEEKS CONTROL	LIB20621	5
# Vs ex11 5.4C	2 WEEKS CONTROL	LIB20622	5
# Vs ex11 5.2D	2 WEEKS DAMAGED	LIB20623	5
# Vs ex11 5.3D	2 WEEKS DAMAGED	LIB20624	5
# Vs ex11 5.4D	2 WEEKS DAMAGED	LIB20625	5


library(edgeR)
library(locfit)

#READ IN THE DATA WHICH IS A FILE IN THE CURRENT WORKING DIRECTORY, IT IS RAW GENE COUNTS CALCULATED BY RSEM WITH 3 OUTLYING LIBRARIES REMOVED (LIB20597 [1 d control], LIB20608 [5 d control] + LIB20612 [5 d damaged])
mya_adult_rawdata <-read.delim("vs_ex_11_matrix.counts_outliers_removed", check.names=FALSE, stringsAsFactors=FALSE)

#MAKE DGE LIST AND TELL IT THE NAMES OF THE COLLUMNS
y <- DGEList(counts=mya_adult_rawdata[,2:28], genes=mya_adult_rawdata[,1])

#NORMALISE THE DATA
y$samples$lib.size <- colSums(y$counts)
y <- calcNormFactors(y)
y$samples

#FILTER THE DATA TO REMOVE VERY LOW COUNTS 
keep <- rowSums(cpm(y)>2) >=3
y <- y[keep, , keep.lib.sizes=FALSE]

#RE-NORMALISE FOR NEW FILTERED LIBRARY (PROBABLY NEGLIGIBLE BUT GOOD PRACTISE)
y$samples$lib.size <- colSums(y$counts)
y <- calcNormFactors(y)

#SET THE BASIC DESIGN, NB absolute outliers as per outlier analysis have been removed, but note time step 3 which = 5 days is still included. Although the replication of n=2 is now no longer enough to carry out statistical analysis on this time point, we have kept it in the dataframe to inform the estimate of disperion only
Treatment <- factor(c("Control","Control","Damaged","Damaged","Damaged","Control","Control","Control","Damaged","Damaged","Damaged","Control","Control","Damaged","Damaged","Control","Control","Control","Damaged","Damaged","Damaged","Control","Control","Control","Damaged","Damaged","Damaged"))
Time <- factor(c(1,1,1,1,1,2,2,2,2,2,2,3,3,3,3,4,4,4,4,4,4,5,5,5,5,5,5))

# SET THE DATAFRAME
data.frame(Sample=colnames(y),Treatment,Time)

#MAKE THE DATAFRAME A SET VARIABLE JUST SO CAN PRINT IT AT THE END
DATAFRAME1 = data.frame(Sample=colnames(y),Treatment,Time)

#Build Groups Collumn
Group <- factor(paste(DATAFRAME1$Treatment,DATAFRAME1$Time,sep="."))
cbind(DATAFRAME1,Group=Group)

#DEFINE THE MODEL DESIGN
Design <- model.matrix(~0+Group, data=DATAFRAME1)
colnames(Design) <- levels(Group)

#ESTIMATE "ROBUST" DISPERSION
RobustDesigny <- estimateGLMRobustDisp(y, Design) 

#PLOT DISPERSION
pdf("Mya_filtered_Robust_BCV.pdf",width=5,height=5)
plotBCV(RobustDesigny)
dev.off()

#FIT THE MODEL USING makeContrasts
RobustFit <- glmFit(RobustDesigny, Design)
my.contrasts <- makeContrasts(CvD = (((Damaged.1)+(Damaged.2)+(Damaged.4)+(Damaged.5))-((Control.1)+(Control.2)+(Control.4)+(Control.5))),
Time1.CvsD = Damaged.1-Control.1, 
Time2.CvsD = Damaged.2-Control.2, 
Time4.CvsD = Damaged.4-Control.4, 
Time5.CvsD = Damaged.5-Control.5,  
levels=Design)

#Response to damage, time 1
CvD_time1_lrt <- glmLRT(RobustFit, contrast=my.contrasts[,"Time1.CvsD"])

#Response to damage, time 2
CvD_time2_lrt <- glmLRT(RobustFit, contrast=my.contrasts[,"Time2.CvsD"])

#Response to damage, time 4
CvD_time4_lrt <- glmLRT(RobustFit, contrast=my.contrasts[,"Time4.CvsD"])

#Response to damage, time 5
CvD_time5_lrt <- glmLRT(RobustFit, contrast=my.contrasts[,"Time5.CvsD"])

#Time independant Genes
CvD_time_independent_lrt <- glmLRT(RobustFit, contrast=my.contrasts[,"CvD"])

#Time dependant Genes
Time_dependant_lrt <- glmLRT(RobustFit, contrast=my.contrasts[,c("Time1.CvsD","Time2.CvsD","Time4.CvsD","Time5.CvsD")])

#make top tags
toptags_CvD_time1_lrt <- topTags(CvD_time1_lrt, n=NULL)
toptags_CvD_time2_lrt <- topTags(CvD_time2_lrt, n=NULL)
toptags_CvD_time4_lrt <- topTags(CvD_time4_lrt, n=NULL)
toptags_CvD_time5_lrt <- topTags(CvD_time5_lrt, n=NULL)
toptags_CvD_time_independent_lrt <- topTags(CvD_time_independent_lrt, n=NULL)
toptags_Time_dependant_lrt <- topTags(Time_dependant_lrt, n=NULL)

#Make to toptags files
write.table(toptags_CvD_time1_lrt, file="mya_edgeR_CvD_time1", row.names=FALSE, col.names=TRUE)
write.table(toptags_CvD_time2_lrt, file="mya_edgeR_CvD_time2", row.names=FALSE, col.names=TRUE)
write.table(toptags_CvD_time4_lrt, file="mya_edgeR_CvD_time4", row.names=FALSE, col.names=TRUE)
write.table(toptags_CvD_time5_lrt, file="mya_edgeR_CvD_time5", row.names=FALSE, col.names=TRUE)
write.table(toptags_CvD_time_independent_lrt, file="mya_edgeR_CvD_time_independant", row.names=FALSE, col.names=TRUE)
write.table(toptags_Time_dependant_lrt, file="mya_edgeR_CvD_time_dependant", row.names=FALSE, col.names=TRUE)

#DGE object for plotting and summary
#make the DE objects for all the comparisons

CvD_time1_0.05 <- decideTestsDGE(CvD_time1_lrt, adjust.method="BH", p.value=0.05)
CvD_time2_0.05 <- decideTestsDGE(CvD_time2_lrt, adjust.method="BH", p.value=0.05)
CvD_time4_0.05 <- decideTestsDGE(CvD_time4_lrt, adjust.method="BH", p.value=0.05)
CvD_time5_0.05 <- decideTestsDGE(CvD_time5_lrt, adjust.method="BH", p.value=0.05)
CvD_0.05 <- decideTestsDGE(CvD_time_independent_lrt, adjust.method="BH", p.value=0.05)

#print the summary for each DE object

summary(CvD_time1_0.05)
summary(CvD_time2_0.05)
summary(CvD_time4_0.05)
summary(CvD_time5_0.05)
summary(CvD_0.05)


#Make smear plots of control vs damaged for each time point
pdf("1DAY_smear.pdf",width=5,height=5)
plotSmear(CvD_time1_lrt, de.tags=detags_CvD_time1_0.05, pch=19, cex=0.15, panel.first=NULL)
abline(h=c(-2,2), col="lightskyblue4", lty=5, lwd=1)
dev.off()

pdf("3DAY_smear.pdf",width=5,height=5)
plotSmear(CvD_time2_lrt, de.tags=detags_CvD_time2_0.05, pch=19, cex=0.15, panel.first=NULL)
abline(h=c(-2,2), col="lightskyblue4", lty=5, lwd=1)
dev.off()

pdf("7DAY_smear.pdf",width=5,height=5)
plotSmear(CvD_time4_lrt, de.tags=detags_CvD_time4_0.05, pch=19, cex=0.15, panel.first=NULL)
abline(h=c(-2,2), col="lightskyblue4", lty=5, lwd=1)
dev.off()

pdf("DAY14_smear.pdf",width=5,height=5)
plotSmear(CvD_time5_lrt, de.tags=detags_CvD_time5_0.05, pch=19, cex=0.15, panel.first=NULL)
abline(h=c(-2,2), col="lightskyblue4", lty=5, lwd=1)
dev.off()


#CHECK IT FINISHED BY MAKING THE DUMMY FILE
write.table(my.contrasts, file="WORKED_yipee", row.names=FALSE, 
col.names=FALSE)
save.image()
y
#