Should we expect different cluster numbers when running SE vs. PE?

[Heisman]

Hey all,

Very basic question: I've been under the impression that if you prepped a sample with PE adapters and then ran it for SE or PE sequencing, you would get the same number of clusters/sequenced fragments, with a total of 2x as many PE reads as you are sequencing both ends.

Is this correct? Or is the chemistry somehow different for each set such that this is not true? I ask because somebody else told me that he's under the impression we typically get considerably less fragments sequenced with PE vs. SE sequencing, which makes no sense to me. Any thoughts?

[SNPsaurus]

I think a very small fraction of clusters don't switch over well for the second read. But I've seen, for example, where a set of samples were run SE for RAD-Seq, then a subset run again in part of a PE lane for RAD PE local assembly. The SE read counts and PE counts correlated well. Also, I've never noticed any trend in reduced read counts in PE lanes.

[Heisman]

Thanks for your response, that's what I expected. I've never thought about this but I'd be curious if clusters expand in size at all when they are reformed, but I imagine the increase in cluster diameter would be rather negligible relative to the original diameter.

[pmiguel]

Thanks for your response, that's what I expected. I've never thought about this but I'd be curious if clusters expand in size at all when they are reformed, but I imagine the increase in cluster diameter would be rather negligible relative to the original diameter.

Yes, the clusters increase slightly in size during turn-around. Here is a picture of PE run of a RADseq lib we did earlier this week (2x150 using Rapid chemistry)


--
Phillip

[SNPsaurus]

Neat! How sensitive is rapid mode to the lower complexity at the cut site? Any difference compared to high output? I thought I read clustering was higher density in rapid, and getting good basecalling on a restriction-enzyme based complexity reduction has been sensitive to dense clustering.

[pmiguel]

Neat! How sensitive is rapid mode to the lower complexity at the cut site? Any difference compared to high output? I thought I read clustering was higher density in rapid, and getting good basecalling on a restriction-enzyme based complexity reduction has been sensitive to dense clustering.

FWIW, I did not notice a difference between HO and Rapid in this matter. That said, RADseq libraries are not a major library type for us. Also, we did not push the density. Raw density was in the 700-800 Kclusters/mm^2.

Also, we spiked in ~0.5% phiX, which may have ameliorated issues. Since the RAD seq libs have 6 bases of bar code preceding the restriction site, most of the really critical calibration has already happened when the problematic bases arrive.

--
Phillip