Very low read numbers from haloplex libraries

[henry.wood]

Hello all. We've recently done some HiSeq sequencing of pooled haloplex libraries. We have sequenced 6 lanes so far from 4 different captures. 4 of those lanes have behaved themselves perfectly, but 2 have very low read numbers, around 5% of the good lanes. We looked at the run QC, and for the poor lanes we only have around 14% of clusters passing filters. We initially thought the lanes were overloaded, so we reduced the amount of input library from 10pM to 8 but it didn't have much affect. Also, although the cluster density was a bit on the high side 8-900K/mm2, one of the good lanes was over 1000K/mm2.
We don't think it is an adapter issue, as some of the good lanes had more adapter than the failed lanes. We are investigating with the manufacturers whether we have been sent a faulty kit. I was wondering if anyone else had seen these issues. Are haloplex kits particularly sensitive to overloading (have we just been lucky with our successful lanes). Have people seen some designs being more sensitive to this than others.
Any thoughts from the collective experience?

[SEQquestions]

Our group is considering using haloplex for deep sequencing specific variants (identified from exome sequencing) so I would be very interested to know if anyone else has had the above issues or, indeed, to hear of any success stories when using this approach?

[mmah]

Hi;

We've recently tried using the Haloplex kit on our MiSeq machines (running the 150bp PE), first at 8pM and our cluster density was >1000K/mm2, the cluster passing filter was low (~30%) with the quality metrics decreasing in the second read. So we decreased the pM to 6pM yielding a lower cluster density (to around 600K/mm2) and the cluster passing filter was better (~60%) but not up to our expectations when we run the Illumina Amplicon Cancer Panel kit.

Have you resolved your problems since?

[henry.wood]

We've made improvements, but they're not completely solved. Our first go for the troublesome samples was loading 10pM and we got around 10% of reads passing filters with cluster density > 1000K/mm2. Our next go was at 8pM and the cluster density was still over 1000, and we got around 15-20% of reads passing filters. We then tried 6pM and density was still over 1000, but now we have 68% reads passing filters, so it seems to me that we are getting closer to proper clustering, but it is still too high. Other lanes have been run at 6pM and are around 600k/m2 and have 95% reads passing filters.
If it was the case that all lanes needed to be clustered at 5pM then I would be happy, but the fact that some run really well at 10pM and others are too dense at 6pM is a bit frustrating. Agilent are testing out our kits to see if anything odd is going on.

[mmah]

Thanks for responding. Your results are quite similar to the quality metrics we've been experiencing too. In fact, the very first run we did with the Haloplex library, we pooled 12 samples and ran it at 10pM and yielded a cluster density of ~1600K/mm2 with cluster passing filter of 75%, phasing/prephasing seemed fine and we haven't been able to recapitulate that run. We've tried pooling less samples and using much lower pM.

We've talked to Illumina and their suggestion was to decrease the pM to avoid over clustering (they think it is likely we've overloaded the flowcell) and or increase sample diversity so we've spiked in 20% to 30% PhiX.

We've considered running it at 5pm or 4pM (which would be fine if we get better run data) but I may just talk to Agilent first.

Thanks again!

[vally20]

Hi!
We had the same problem using Haloplex HS.
How did you resolve it?
Thanks

[henry.wood]

I don't know if you're asking me or mmah. We think the problem was that our libraries were not being quantified correctly, and their concentrations were underestimated. For some reason Haloplex libraries seemed particularly sensitive to over clustering. Whether that is because their concentrations are difficult to measure, or their qualities are affected more for over clustering I don't know. Maybe we were just having a bad day.
We were using an agilent chip to quantify, and there was so much library it overwhelmed the chip and gave a falsely low result. Once we used the correct chip for the quantity of library, our troubles stopped. Hope that helps.

[vally20]

ok,
thanks for your help!