Quality testing of ChIP-seq library prior to sequencing.

[The Bric]

Hi all,

I'm new in the ChIP-seq world. I have a good antibody that I know works well in ChIP and I also know some good target regions that I can use to validate the ChIP. After performing and validating a ChIP I used the ChIP'd material to construct a library to use for Solexa sequencing. To test if the library construction went fine I tried to perform qPCR validation using the same primer pair as before assuming that there would be less signal in the IgG control than in the one using the specific antibody. However, the results looked really strange - no signal at all in both samples. Does anybody have a good method to test the quality of the library prior to sequencing?

Da Bric

[ssmemo]

Hello,
I'm having a similar experience. Did you figure out what was going on?

Thanks,
Scott

[frozenlyse]

You could try a few serial dilutions of your library as too much DNA can inhibit qPCR

[ssmemo]

In my case at least, I've measured the libraries with pico Green and diluted them to a concentration at which I've shown the qPCR primers work well.

(think I should start my own thread; I don't mean to hijack this one)

[Chipper]

What is the library insert size and size of your PCR? You need a short amplicon that is centered around the binding site to amplify it from the library since you exclude the longer fragments from the sample in the library prep.

[ssmemo]

We've size selected around 175bp (bioanalyzer peak spans 125-225), and our amplicons are <=100bp. I believe that after subtracting for the adapters our library fragments are large enough to support the qPCR amplicons. Maybe though we're on the cusp of the amplicons being too large and there's fewer of those long molecules available for PCR? I'll have to look into this and get back to you.

[james hadfield]

We recommend all ChIP-Seq users to test their IP beofre library prep with some positive controls by real-time PCR. Then once the library is made to repeat this on the library and see if the enrichment is still there. Exactly as it appears you have done.

We can get libraries that give exactly the same result as you appear to have got and our advice is to repeat the library prep. A pain in the neck I know but it generally resolves the problem.

Real-time PCR is the best way to validate libraries other than running a lane of seqeuncing. Is your service provider still using PhiX? If so you could ask that they spike your library into this lane at 10% and see what it looks like. They may even do this for free!

I would recommend: ChIP-seq: using high-throughput sequencing to discover protein-DNA interactions. PMID: 19275939 (but I am an author so I guess I am biased!)

[ssmemo]

That's the problem: we don't have enough material to remake the libraries.
Using more readily-available material, I've done some more digging into the problem and there does seem to be large variation in library quality. I've read both your paper, James, as well as the Sanger Nature Methods one with recommendations on library prep, and are generally following the same protocols.

Something I have confirmed is that for reactions with similar template quantity, the Ct's using raw ChIP output are much (~8) cycles lower than for using libraries as template. I realize a sizable proportion (~37% in my case) of the mass of DNA from the library material is linker, and therefore leads to an over-estimation of the amount of template, but even so, I don't think that would lead to an 8 cycle difference.
One idea is to perform a re-amplification on the material, allowing me to use more material per reaction and thus bring the Cts down into a usable range (i.e. <30). But I'm concerned with yet another round of amplification altering the natural ratio of genomic regions.
I will likely do this experiment, but has anyone compared qPCR-based enrichments from raw ChIP output to those from libraries? Do you see such large differences, or any other patterns?

Relatedly, does anyone know of thorough data looking at the variation among libraries, preferrably via qPCR? We have had very successful sequencing runs, and our protocol is unchanged, so there has to be some variable we're failing to account for. Any other ideas at how to assess library quality other than reamplification and throwing on an array?

[theduke]

Well I'm new to this forum so apologies for bringing up old posts but I was interested in this post since I have just prepared my first library. Essentially I had no idea how much DNA was going in to the Illumina library prep but I seriously doubt it was close to 10 ng (this is relevant later). Nevertheless prior to preparing the library I had excellent enrichment at a positive control promoter (2% Input) when compared to a negative control promoter region (0.1% Input). This is quite significant given I am working with a TF which is always more challenging. IgG is always non-detectable in my ChIP experiments so is really a useless control for me.

After the Illumina prep I ended up with a very nice looking library around 250 bp. I actually size selected twice as I didnt like the primer/adapter dimers etc. I only have about 30 ng of library however so I diluted it 1:50 for qRT-PCR using SYBR and the same sets of primers as before. I got exactly the same enrichment pattern when compared to my negative control and even with a 50-fold dilution my positive control region came out at 30 Cts.

One point for me is that I think it is next to impossible to normalize the amount of DNA you put into the PCR pre- and post-library prep as the quantification is just not good enough, especially when you take into account adapters, primers, size selection etc. In my experience also, attempting to quantifying the DNA post-chip is pointless so we really do not know how much DNA really goes into the library prep.

Anyway, those are my thoughts.