Paired end vs Single end Differential gene expression analysis

Question

[Sreeraj Thamban]

Hi, I have two conditions untreated and treated with two biological replicates each(RNASeq). But the samples are from different sequencing runs. The untreated library is PAIRED END and the treated library is SINGLE END. Is it possible to do a differential gene expression analysis using the above libraries? is it legitimate do so? I am really looking for your answers. Thanks in advance

[WouterDeCoster]

Was only the sequencing performed differently, or the whole library prep?

[Sreeraj Thamban]

Hello Coster, Only the sequencing performed differently,library prep was same for both.

[WouterDeCoster]

In that case, it would probably be most valid to take only the R1 of the paired-end data.

Answer 1

[Ido Tamir]

The best thing you can do is to disregard the 2. read, trimm them to the same length from the 3' end and align only the 1. read. Look at technical parameters errors/cycle, coverage across genes, etc ... per group to ensure that differences are probably mostly biological. Most of the times the runs have very little technical variation compared to the biological variance. I hope the libraries were prepared in the same batch? This really makes a huge difference. Much bigger than different flowcells or sequencing centers and it can be larger than biological variance (depending on the biological system of course).

[Sreeraj Thamban]

Hello, Tamir thank you for the answer, Is it possible to generate a counts table for the paired end and single end libraries separately and use them for DESEq2?

[venu]

@Sreeraj, of course it is possible to generate counts table and use it with DESeq2. But the problem is, the results you observe might not be the ones due to biological variation. What @Ido suggested seems reasonable for your case. Better collect some information about library prep also.

[Ido Tamir]

I don't understand the question. Please explain in steps what you want to do and how it differs from what I described above.

[Sreeraj Thamban]

My idea is to get read counts from featurecounts, followed by DESeq2. feature counts paired-end data as read pairs, therefore an aligned paired-end reads will only be counted as one. I already aligned the PE reads using tophat.

[Ido Tamir]

No. align again only read 1.