Size-selection problem with ChIPseq-prep on Beckman SPRIworks

[mgravina]

Our core lab has been seeing unusual failures of size-selection when we prep ChIP samples for Illumina sequencing on our Beckman SPRI-TE robot. We generally size-select for 200-400bp. However, since the problem started, we've seen large post-enrichment bands at 500-700bp and little material in the 200-400bp range.
The problem appears semi-random (a few samples from each batch enrich normally) and does not seem to correlate with reagent lots. It is also confined to ChIP samples.
Our best hypothesis so far is that the problem is caused by undersonication- we ran a test of whole-cell extract with different sonication times and it suggests that more sonication ameliorates the problem.
Has anyone else seen this phenomenon, and do you agree with this hypothesis? Thanks!

[pmiguel]

There have been a few other previous discussion of spriworks library size abnormalities:

http://seqanswers.com/forums/showthread.php?t=12855

http://seqanswers.com/forums/showthread.php?t=10529

Basically I think it comes down to one of a few possibilities:

(1) All TruSeq libraries after amplification are subject to a phenomenon where double, or even multiple peaks appear. Sometimes, but not always, the larger peak is double the size of the smaller peak. Various explanations for these, but they all involve "ectopic" annealing of library strands to one another. These resolve if you repeat PCR with an excess of PCR primer. You can also strand denature the library and run it on an RNA bioanalzyer chip or denaturing gel to see the true sizes of your amplicons.

Basically, if this is you issue, you can ignore it and proceed. Sequencing should be fine because the samples are denatured prior to cluster PCR.

(2) There may be some issue with the SPRIworks instrument. You would have to troubleshoot this with Beckman. Probably an issue with the concentration of Ampure beads:sample being used.

(3) As you write, "undersonication". Seems like you could assay this prior to library construction with a DNA High Sensitivity chip.

--
Phillip

[mgravina]

Thanks for the links, they're very informative! In response to your suggestions of possible causes:

1. We've seen the ectopic annealing peak plenty of times before. These peaks seem to be different. They are slightly larger in MW than the ectopic annealing peaks, consistently larger in area than the main peaks, and further cycles of amplification with extra PCR primer do not cause them to resolve.

2. Some of our associate labs have also seen this phenotype on their machines, and it only affects ChIP samples on our machine, not all samples. So we're thinking it's not a problem specific to our instrument.

3. We run all of our input samples through a DNA High Sensitivity chip, but this does not help much with determining success of fragmentation because the target band is very low in the input, and tends to get lost in high levels of salt and noise even when the sample is diluted.