ChIP-seq, Best control?

[jvais]

Dear all,
I plan to make ChIP-seq for a specific transcriptional factor (TF) in D. melanogaster.

In this context, we have developed an ectopic protein (TF) tag with Flag: TF-Flag
We can precipitate the chromatin with Flag antibody.
I've two genotypes available the wild-type- flies (no TF -lag) & the TF-Flag- flies.

So at this time, I've made several ChIP and validated the specificity by qPCR using different controls.
I tested four possible controls.
- Input in flies - in TF_Flag- flies
- No Antibody - in TF_Flag- flies
- Mouse - Antibody (Flag ab is made in mouse) - in TF_Flag- flies
- Flag ab in wild-type- flies
In all, I've a high enrichment for the target of my TF compared to these four controls.
Our results in qPCR showed the most aspecific binding was in wild-type- flies with Flag ab.

Do you think this is the best control?

[mudshark]

I'd say
1) Input control
2) FLAG in wt. Not preferred as it is a different chromatin prep and in the worst case the FLAG-TF and wt flies might have local copy number variations.

[jvais]

thank for your answer mudshark!
But I'm not totally understand your opinion
For the input, It's just total DNA coming from the same chromatin extraction. So it's probably very close to the total DNA of D. Melanogaster?

I'm thinking to a other possible control: TF (without flag) and precipited with my Flag-ab?
To summarize, two conditions:
tested condition=> da-GAL4>UAS-TF_Flag precipated with Flag-ab
Control=> da-GAL4>UAS-TF precipated with Flag-ab
comment: da-GAL4 ubiquist driver

Do you think in this case this control is better to the input?

[mudshark]

the input is not equal to genomic DNA sequencing as the release of DNA fragments in chromatin preps varies locally. depending on the activity status of chromatin, i.e. transcribed regions are released more efficiently and will therefore have a higher number of reads.

your choice of control also depends on your total experimental layout, whether you plan to look at other TFs or other features as well. to be on the safe side, input is the better control imho.

[jvais]

thank for your answer.
Regarding to your comments, we decided to use the input as control.
Additionnally, we have add two other control.
to summarize:
experimental condition:
da-GAL4>UAS-TF_Flag precipated with Flag-ab
control:
Input => normalization of data.
da-GAL4>+ precipated with Flag-ab => nonspecific binding
da-GAL4>UAS-TF_Flag precipated with IGM mouse (Flag ab is made in mouse) => nonspecific binding

All of these will be done two times in illumina with 20 millions reads (20 millions-5'-3' & 20 millions-3'-5').

That's make sens for you?

[mudshark]

sounds good. everything controlled. in case you want to save some bucks i would skip the da-GAL4>UAS-TF_Flag precipated with IGM mouse