Hi,

this question relates to post Time-course, comparison of two groups. I have different question related to the data presented there so I've made new question.

I'd like to plot observed and fitted log-CPMs for a selected gene or genes.

I've consulted edgeR user guide (p. 113, the bottom block of code). But there is plot for one time series without replications. I'd like to make a plot for experiment linked above, there are two levels of "line".

I wonder if there is conveient way to plot such data computed by edgeR.

For now I assumed tahat averaging log-CPMs values is fine (???)

Here is my code (continuation of the one in the linked post)

# data arrangement

# id of the top gene
> tab = as.data.frame(topTags(fit2, n=30)) # here I used "fit2" object generated in the linked post
> ids=row.names(tab)[1]

# data for lines for fitted CPMs
> logCPM.fit <- cpm(fit2, log=TRUE) # according to edgeR user guide
> mf1=logCPM.fit[ids,1:12] # data for mutant
> mf2=logCPM.fit[ids,13:24] # data for wt
> mf12=rbind(mf1,mf2) # binding  rows for mutant and wt
> x=as.data.frame(mf12)
# averages or biological reps
> x$sr15=rowMeans(x[,1:3])
> x$sr22=rowMeans(x[,4:6])
> x$sr29=rowMeans(x[,7:9])
> x$sr36=rowMeans(x[,10:12])
> x2=x[,13:16] # selecting only averages
> mf12av=x2

# data for scatterplot of observed CPMs
> logCPM.obs <- cpm(cts.tmm, log=TRUE, prior.count=fit2$prior.count) # according to edgeR user guide
> m1=logCPM.obs[ids,1:12] # data for mutant
> m2=logCPM.obs[ids,13:24] # data for wt
> m12=rbind(m1,m2) # binding data for mutant and wt
> x=m12 # new object, just in case if I corrupt the original
# averages for biological reps
> x$sr15=rowMeans(x[,1:3])
> x$sr22=rowMeans(x[,4:6])
> x$sr29=rowMeans(x[,7:9])
> x$sr36=rowMeans(x[,10:12])
> m12av=x[,13:16] # selecting only averages
> x=t(m12av) # transposing to get data for matplot

# plot
> matplot(c(15, 22, 29, 36), x)
> lines(c(15, 22, 29, 36), mf12av[1,], col="black", lwd=2)
> lines(c(15, 22, 29, 36), mf12av[2,], col="red", lwd=2)

And here is my plot. It is obviously bad way of making such plots. Could anyone suggest me the better method?