RNA-seq: read counts in single- vs paired-end sequences

[fbarreto]

Hi,

I've read several threads with very related topics, but none have addressed the one simple question I have (which stems from my being new to this).

We already have an assembled transcriptome (from 454 data), and would now like to do an experiment to assess differential gene expression with Illumina.

My question is (assume one Illumina lane is being sequenced): will a 2 x 100 bp paired-end lane generate approximately HALF the number of reads of a 1 x 100 bp single end lane? (By read I mean a copy of a transcript, i.e. the two ends of a paired-end sequence count as only ONE read, for my purposes)

I would assume so, since the sequencing effort (in terms of base pairs) for such a paired-end is twice that of the single end.

Since I'm interested only in maximizing read counts, the extra information that comes in the paired-end sequence is not of interest to me at this point.

Any insight regarding the comparison of read counts per sequencing effort between the two methods would be greatly appreciated.

Thanks!

[chadn737]

You produce the same number of reads. In a paired end run, you sequence BOTH ends of the same fragment, so both reads are coming from the same cluster on the flow cell. So with a paired-end run, you are getting exactly the same number, so this has no effect on your total read count (if you count the paired end reads as one read).

The advantage of a paired end run is that you can align the reads with greater confidence.

[fbarreto]

Thanks for insight! But wouldn't a a 2 x 100bp paired-end run be sequencing a total of 200bp from each sequence (i.e. transcript copy), 100bp from each end?

So if a run sequences 1,000,000 bp (I know it's more, but just to keep it simple), I would get 5000 paired-end and 10,000 single end reads. Am I thinking about this wrong?

Thanks again!

[chadn737]

I think you are confusing base pairs sequenced with reads count. The number of reads you get is a factor of the number of clusters on a flow cell. I think the recommended cluster density on an Illumina Hi-seq typically gives you ~65 millions per lane (somebody correct me if I am wrong). From these 65 million reads, doing a paired end sequencing run vs a single end run means that you will get 65 million pairs (or 130 million reads) vs 65 million single end reads. However, since you sequenced 200 bps versus 100 bps, you will get double the number of base pairs sequenced.

If all you are doing is differential gene expression, then you are not really concerned with base pairs sequenced, just number of reads. In this, paired end versus single end doesn't affect the number of reads.

[fbarreto]

That's great to know! Thanks a lot!
My confusion was that, in my simple mind, the sequencing effort (in bp) versus number of reads had to balance out such that getting longer sequences would mean getting fewer. But if not, like you said, then paired-end seems like the way to go.

Much appreciated!