150bp read?

[NGS_user]

Hi Folks,
I am in a situation where I need long reads but with as few sequencing errors as possible. I currently have 75bp reads but am looking to generate longer ones. Would you recommend 100bp or 150bp? Ideally the 150 bp reads would avoid assembling errors I am having but a high number of errors in the reads could also cause all sorts of bother! Any recommendations?

[tonybolger]

Hi Folks,
I am in a situation where I need long reads but with as few sequencing errors as possible. I currently have 75bp reads but am looking to generate longer ones. Would you recommend 100bp or 150bp? Ideally the 150 bp reads would avoid assembling errors I am having but a high number of errors in the reads could also cause all sorts of bother! Any recommendations?

Use 150bp if you have the choice.

It should be better out to 100bp, plus you should get decent quality (on some reads) out to 130-140 bases. Adaptive trimming is, as ever, a good idea to take out the poorer reads.

[HESmith]

An alternative is the one used by the Broad Institute for de novo genome assembly. Construct libraries with 180bp inserts, performed paired-end 100bp sequencing, and use the 20bp overlap in the center to generate 180bp reads. In addition to increased length, there's the added advantage that the bases at the ends of the reads (which typically have the lowest q-scores) are sequenced twice. The critical parameter is a tight size distribution for your libraries so that the reads overlap.

[Andrew_Slatter]

All depends on what quality you need at the end of the run. I've just been looking at the latest MySeq performance data and they report a mean qscore of >20 at 150 bases.

[chkuo]

We had a lane of 150bp PE reads for de novo assembly of a bacterial genome. Curiously, trimming the reads down to 100bp actually produced larger contigs. Not sure what's going on, I just posted this issue here:

http://seqanswers.com/forums/showthread.php?t=11784

[tonybolger]

An alternative is the one used by the Broad Institute for de novo genome assembly. Construct libraries with 180bp inserts, performed paired-end 100bp sequencing, and use the 20bp overlap in the center to generate 180bp reads. In addition to increased length, there's the added advantage that the bases at the ends of the reads (which typically have the lowest q-scores) are sequenced twice. The critical parameter is a tight size distribution for your libraries so that the reads overlap.

You can do the same trick with 150bp reads, using 250bp or so inserts. However, you have to make absolutely sure that the distribution of reads is tight, or you'll possibly end up collapsing a small repeat region.

[kwaraska]

it depends on what machine and chemistry version.

On the GA they can comnfortably go to 150 without too many errors.
On the current HiSeq chemistry(not the one released last week-v4) I wouldn't go beyond 100, maybe not even beyond 75
That has been our experience.
They claim the v4 chemitry will be accurate to 150 bases (each way) but I don't have first hand knowledge yet. We will not be switching until we use up our old stuff and we have 6-8 weeks worth.