75bp: high error rates on read2

[bioinfosm]

Is this common? Or there are ways around getting high error rates after cycle 50 for long reads? Especially read2 looks pretty bad!

[basickler]

I need a little more info on this. What are you aligning to? What chemistry are you using? Is this PhiX? If you're using the v2 kits and this is PhiX than yes these are very high, it should be below 5% at cycle 75 for both reads on a decent run.

The reason for this could range from degraded reagents, to out of focus images to a bad turn around on the PE module, etc. etc.

Brad

[bioinfosm]

Thanks Brad.
Its PE mRNA seq genomic DNA on human genome as reference. These are not PhiX, but human mRNA samples.

[basickler]

Well first off you might want to use the human cDNA library as a reference genome. The human genomic reference isn't going to work with ELAND very well. Eland doesn't support gaps (read introns). This still doesn't explain the higher error rates on the paired end read, that's probably a chemistry or technician issue.

Brad

[cnicolet]

This is typically what I see regardless of the run length we do, ie, similar curves to what you posted for our 60 x 60 and 85 x 85 runs. I don't think it's a technician issue. Because the curve does look proportionally similar (ie, starts to rise at about the same percentage into the read, not at a particular base) I think it may be a phasing or base calling issue. I'm also suspecting reagent longevity--our next long run will use freshly made topped off reagents like the old days just to see if that helps.

[bioinfosm]

Well first off you might want to use the human cDNA library as a reference genome. The human genomic reference isn't going to work with ELAND very well. Eland doesn't support gaps (read introns). This still doesn't explain the higher error rates on the paired end read, that's probably a chemistry or technician issue.

Brad

Sorry for the confusion, these are actually genomic DNA samples, not mRNA. I corrected that in my post above

[bioinfosm]

paired end genomic human dna - 2nd read!

Hi,

Do people have comments on this sort of error graph from GA pipeline for a particular tile, running paired 75bp genomic human dna?

Thanks,

[der_eiskern]

Is this common? Or there are ways around getting high error rates after cycle 50 for long reads? Especially read2 looks pretty bad!

if i recall correctly we've noticed higher error rates in the second read and attributed it to the lower signal-to-background ratio. i hear that PE constraints on mapping help make up for this problem but i've never seen the data to support that directly.

[GenoMax]

Is this common? Or there are ways around getting high error rates after cycle 50 for long reads? Especially read2 looks pretty bad!

Did this turn out to be a rare observation? These error rates seem rather high. We generally see error rates below 1.5% for 2x76 runs with human samples.

[bioinfosm]

yes it was just one of those tiles. We are also getting around 1-2% error rates for 2x76 runs