Lower quality in reverse reads only. Is it normal?

[drramki]

We did 100bp PE sequencing (RNA-Seq) on 8 samples and we see a distinct systematic drop in quality only for reverse reads (attached screenshots). We contacted the sequencing center and they said it "passed" their QC and also Illumina's spec of ~83% bases > q30. However, when we look at the % of sequences below q30, there is a remarkable difference (~7% in fwd reads vs ~24% in reverse reads).

I wonder if this is okay. Similar problem has been discussed before in this forum or elsewhere.

http://seqanswers.com/forums/showthread.php?t=34648

[nucacidhunter]

Could you please post FastQC plot and which chemistry was used.

[GenoMax]

It is "normal" (if one wants to call it that) to see reduction in Q-scores in later R2 cycles (and the difference can be similar to what you found). I assume this is MiSeq data. As the sequencing center pointed out the data technically satisfies published spec for MiSeq Q-scores.

One reason that can lead to rapid decrease in Q-scores is inserts shorter than the read length.

[drramki]

Here is the link for fastQC plots. And I suppose this is MiSeq. We have never seen this pattern in HiSeq data.

https://www.dropbox.com/s/rs5xlputqcmobbp/RProjectRnaSeqFastQC.pdf

[nucacidhunter]

If the read numbers are over 30M pairs/lane, it should be from HiSeq. In any case the quality drop is more than expected in any of Illumina systems and indicates an issue most likely with sequencing reagents. As GenoMax has pointed out short inserts also can be the cause where in later cycles adapter sequences are being read. It will be interesting to know output read number and % of reads which were demultiplexed.

[drramki]

Thanks GenoMax and nucacidhunter.

I was wrong when I said MiSeq. This is actually from HiSeq. We got the data twice from the sequencing center and both looked same.

[GenoMax]

This may be a "not so good" quality library, if the result has been the same two times in a row on two runs. You would want to do some additional QC/cleanup (trimmomatic, bbduk etc) to figure out what is causing this drop (adapters?).

Also ask the provider if they had seen something similar for other lanes on those runs to eliminate possibility of technical problems with the runs.

[nucacidhunter]

In addition to above suggestions you can have a look at Adapters plot in FastQC report to see the level of adapter sequences and if that correlates with drop in quality. Also you can check size distribution of the library by examining the Bioanalyser or TapeStation trace to work out insert sizes. If high percentage of library inserts is below 100 base (fragment sizes lower than 230 bp) in this case, the reads will run into adapters causing quality issues that can be fixed by trimming. If none of those seems to be likely cause it could be the chemistry specially if it was Rapid run.

[N311V]

With pretty much all of our HiSeq data we do see a drop-off in quality towards the end of the second read but nothing that extreme. I tend to agree with previous comments that something has gone wrong here. In my experience sequencing centres etc. have rather soft QC metrics.