ChIPseq on tissue - eliminating background

[JHU-ChIPmaniac]

Definitely do qPCR on your libraries to make sure you still have good enrichment following library construction. I have made libraries from ChiP samples that were good (high fold enrichment of several targets) that no longer showed good enrichment so I didn't proceed with sequencing.

[mudshark]

what is your model organism and how many mapped reads do you get? what's the mappability fraction of your reads?

i would also try to keep the duplicates as in my opinion the duplicate removal can potentially generate a HUGE problem if your IP works very well, there are very few target sites and you have tons of mapped reads. then you will elminate all your signals due to true, i.e. non-artificial, read duplication.

and do the qPCR on your library!

[Theorbe27]

Hi mudshark,

I am not entirely sure about all aligner parameters since I have only received aligned reads from our seq facility. I would have to ask them for more details if this turns out to be the crucial point. This is as much as I know: they use the first 25nt of a read as a seed and allow 2 mismatches. If multiple matches for a seed are reported, quality scores for the entire read are taken into account and the read will be called as 'multiple' or 'unique' based on parameters that I am not aware of.
Multiple matches and duplicates were excluded in the analysis.
I have not yet done qPCR on the library.

Theorbe

[mudshark]

what are your aligner parameters? did you eliminate multiple match or duplicate reads berfore analysis?
did you do qPCR on your sequencing library?

[Theorbe27]

Hi,

I am trying to do ChIPseq on tissue for a transcription factor that is expressed in a single cell type within a tissue. Therefore, specific signal is only going to be derived from about 3% of the input chromatin.
By ChIP-qPCR, I have managed to consistently get 15 to 100x enrichment (calculated in both ways, foldchange of specific IP vs IgG IP, and %input candidate region/%input negative control region) for six different candidate genes, while negative control regions were not enriched.
So I went forward to ChIP-seq which turned out somewhat disappointing as there is a lot of non-specific signal in the IP sample that looks similar to an input signal. However, I do get peaks that reflect the enrichment at my candidate genes but these peaks don't look great. I don't know yet how to summarize peak statistics in an appropriate fashion, but when you plot the read count there is like one or two reads in almost any 100bp window, and the count just goes up to like 6-7 reads within a peak. My feeling is that I would be able to get much better peaks if I was not having all those unspecific fragments lingering around and interfering with library amplification.
Has anybody come across a similar problem and would be able to give some suggestions?

Thanks

Theorbe

Here is what I did for this sample:
Dissect tissue, xlink.
Lyse and sonicate to ~300bp on average (>75% of signal smaller than 500bp).
IP (antibody is good and has been giving us excellent results in any application it has been used for including ChIPchip and ChIP-qpCR). Capture Ab with magnetic beads (blocked with BSA and tRNA). Multiple washes, including several high-salt washes with both NaCl and LiCl containing buffers. Elute DNA and clean up with PCI x3.
Pooled six of such ChIP-DNAs and checked for enrichment of best candidate gene by PCR(100fold). Library prep done on a SPRIworks system followed by Illumina sequencing.