RNAseq analysis with DEGseq and no replicates

Question

[rsafavi]

Hi everyone, I preformed DEGseq analysis on my date which has no replicates, and I got the following MA plot.

I'm not sure if the graph makes sense or not? Most of the MA-graphs that I have seen so far merge around y =0. Also based on the DEGseq manual, the red points in the (MA-plot) are the genes identified as deferentially expressed.

Any help and clarification is really appriciated!

[ATpoint]

The y-axis of the MA is calculated as log2(sample1) - log2(sample2). Given that a) samples were properly normalized for sequencing depth and b) most genes did not change between the conditions (which you typically expect), the bulk of the data points should be along y=0. From this plot, it looks like trx-exp has been sequenced deeper than trx-control. I have no experience on DEGseq, is this a plot from the tool on the normalized counts or did you do it yourself, and on which counts (raw or normalized)?

[Kevin Blighe]

Good to have you back

[rsafavi]

Thanks for your prompt answer. You are absolutely right the trx has been sequenced deeper that control (almost twice as deep). This plot is generated by the tool itself, and the input was raw counts(from featurecounts) as they said in their manual. I guess base on what you said the reason I'm not seeing the bulk of the data point along y = 0 is because the sequence depth is very different

[ATpoint]

You can explore the normalized data yourself, e.g. after TPM normalization in R:

## Normalize by gene length:
x <- counts.matrix / gene.length

## Normalize for depth:
tpm.mat <- t( t(x) * 1e6 / colSums(x) )

## Prepare MA Plot (assuming column and and 2 of tmp.mat are the normalized counts):
M = log2(tpm.mat[,1]) - log2(tpm.mat[,2])
A = 0.5*(log2(tpm.mat[,1]*tpm.mat[,2]))

## Plot:
plot(A, M, xlab="A", ylab="M"); abline(h=0, lty=2)

[rsafavi]

Thank, While I'm trying that, do you have any insight on why the red points which based on the manual are "differentially expressed" are on line y = 0?

[grant.hovhannisyan]

Just to make you aware about further scientific decisions based on this data, DE analysis without replicates does not make sense in general.