0% duplicates in RNA-Seq/Drop-seq library

[GenoMax]

I suggest that you run clumpify.sh from BBMap to get an exact idea of the duplication. You can allow for errors when doing the sequence match. FastQC does not look at the entire dataset for some of the modules (only a % of data is sampled).

Even though there is a thread for clumpify here the one over at Biostars has the directions clearly defined on one page.

[nucacidhunter]

I think HS4000 reads contain lots of errors due to positional lower quality so the sequences of duplicates do not match and they are reported as unique reads. Also lots of reads seems to have very low quality over the whole length of read. If you trim or filter low quality reads you should get similar duplication rate for both runs.

[Jaeb]

Thanks for your fast reply. I did attach now the complete FastQC reports....

[nucacidhunter]

I agree with you that with deeper sequencing %duplicate should increase and read length differences are less likely to be the cause as FastQC uses initial 50 sequence of a subset of reads for duplicate calculation.

It would be helpful if you could post the whole FastQC report for both runs as other plots might give some clues about the cause.

[Jaeb]

Hi all,

I recently sequenced a cDNA Drop-Seq library on a HiSeq2500 and later to get more reads per cell on a HiSeq4000. I am a bit confused when looking at the "Sequence dublication level" plots of the FastQC reports:
Percent of seqs remaining if deduplicated: HiSeq2500 25% (about 25Mio reads in total) and HiSeq4000 100% (100Mio reads in total)

Could some one explain me this? So for my understanding, it means that the HiSeq4000 run somehow doesn't have sequencing duplicates - which is super-unlikely given the fact that a Drop-Seq library is generated by quite some PCR cycles. Also, if anything I would have expected more duplicates as the same library was just sequenced deeper.
I don't know whether it matters, but cycle length was 180 in HiSeq2500 and 70 in HiSeq4000. The Q30 mean seq quality was 85% in HiSeq2500 and 65% in HiSeq4000.
Any ideas on this issue are more than welcome. Thanks