Dear all,
we have been running multiple ddRAD libraries prepared by following the protocol of Petersen et al 2014 in our lab on the HiSeq2500 (SR 100) in the past couple of years. We pooled either 48 individuals (using 48 inline-barcodes) or 96 individuals (48 inline-barcodes x 2 indices) and used 10% PhiX on the sequencing runs. This always produced nice data for us.
For our latest work, he had to move our sequencing runs to the NovaSeq6000 (SR 100, SP flowcell) as our sequencing provider stopped operating the HiSeq2500. For this we pooled 288 individuals (48 inline-barcodes x 6 indices). According to the Sequencing facility, again 10%PhiX was added.
When we received our data, approximately 30% of the reads got filtered out by the chastity filter due to poor read quality and another 30% of the data consisted only of polyG reads.
After some read-up we learnt that the polyG could be sequencing artefacts based on poor clustering which might be caused from low complexity areas in our libraries (the enzyme cut sites, since they are identical in all reads and potentially also the barcodes).
A peculiar detail of the output was that only 3% PhiX could be detected in the sequencing data.
The sequencing facility assured us that they added 10% PhiX and recommended to use 25% or 30% PhiX in our next run.
So my questions are:
1. Does anyone have experience with running ddRADseq or RADseq data on a NovaSeq? If so how much PhiX did you use for your run?
2. Did anyone experience this polyG problem in their sequencing runs?
3. If 10% PhiX really was added to the run, what phenomenon could cause a reduction down to 3% only in the output data?
Many thanks in advance!
Tamara
We have recently experienced almost the identical problem - but we were never sure if the company really added 10% Phi-X - and we argued (politely) a bit about it. They were not convinced our libraries were OK, consultation with the Illumina led to the suggestion of adding 25 or 30%. In the end, we made a deal that if our next set of libraries worked ok (using 15%, because I was not willing to try 25 or 30% right away), they would re-sequence our library batch that produced such bad results (including many poly-Gs as you describe). We did that, our 2nd batch worked fine, and so they resequenced our old batch and all is fine again. However, there is still a bit of unclarity as all of our orders were on the NovaSeq6000 (SP)(previously we had used HiSeq2500 with no such problems), but the company has announced that some orders on the SP will be run on the SP1,2 or 4 flow cell set up, as the company chooses, without asking. We think this has occurred, and the strange thing is that the total read numbers now are way above what the minimum guarentee is - we are getting 1.1 Billion reads per flow cell, instead of 600-800 Million as promised on the SP set up. So, not sure what is going on here - but the bottom line is that the company admitted that something technically was not right on their end, and a re-running of the libraries that failed worked perfectly. It is clear that NovoSeq has more of a problem with low complexity reads than the HiSeq, but I also think that companies are only learning themselves how to deal with this, and this % Phi-X is a bit of a black box. Above all, you have no guarentee what the sequencing service is actually adding - in our original case, we are convinced (also based on control checks with the company), that much less than 10% was added - they never admitted this outright, but no more than 3% was found on the original run that did not work properly.
- Badges
- Science method