ChIP and qPCR calculation ? - (Aug/10/2007 )

Hi all,

I am establishing ChIP for a transcription factor and I perform quantitative PCR to ensure ChIP give good results.
The experiment is done on 400 µl of sonicated DNA. I take 20 µl for Input, 180 µl for ChIP and the last 180 µl for the ChIP negative control (immunoprecipitated with normal Rabbit IgG). After DNA isolation, I perform quantitative PCR on 20 ng for the input, the ChIP and the negative control. The results give 1 or 2 Ct between input and ChIP and 5 or 8 Ct between ChIP and negative control, so it seems good ( ).
Now, I don't know exactly how analyze these results because I've read that “we can admit that there is a factor 1/100 between input and ChIP” but I don't know why and in which case? (what represent this factor?). In the same way, there are many ways to calculate the enrichment in the ChIP. Some use interest gene and one negative gene in PCR, others use just the Ct between input, ChIP and IgG… Can you explain me the best calculation, please?
Thanks a lot !

-Dukon-

Hi Dukon:

We are having similar problems in quantitative PCR calculations, but for a MeDIP experiment. Can someone please explain how to calculate enrichment between bound and input DNA?

As far as we understand one can calculate the fold enrichment relative of the MeDIP relative to an unmethylated control (delta-delta Ct method?!)or one can calculate the fold enrichment relative to the Input (Pfaffl method?!). But we don´t understand how the equations would work or which is the appropiate method to use.

If someone could give us a hint we would be very greatful.

Diana and Nina

-dianasan-

Hello,

You can go to here
http://www.superarray.com/chipqpcrresource.php

Good luck

QUOTE (dianasan @ Aug 10 2007, 08:16 AM)
Hi Dukon:

We are having similar problems in quantitative PCR calculations, but for a MeDIP experiment. Can someone please explain how to calculate enrichment between bound and input DNA?

As far as we understand one can calculate the fold enrichment relative of the MeDIP relative to an unmethylated control (delta-delta Ct method?!)or one can calculate the fold enrichment relative to the Input (Pfaffl method?!). But we don´t understand how the equations would work or which is the appropiate method to use.

If someone could give us a hint we would be very greatful.

Diana and Nina

-sonixchip-

QUOTE (Dukon @ Aug 10 2007, 03:32 AM)
Hi all,

I am establishing ChIP for a transcription factor and I perform quantitative PCR to ensure ChIP give good results.
The experiment is done on 400 µl of sonicated DNA. I take 20 µl for Input, 180 µl for ChIP and the last 180 µl for the ChIP negative control (immunoprecipitated with normal Rabbit IgG). After DNA isolation, I perform quantitative PCR on 20 ng for the input, the ChIP and the negative control. The results give 1 or 2 Ct between input and ChIP and 5 or 8 Ct between ChIP and negative control, so it seems good ( ).
Now, I don't know exactly how analyze these results because I've read that “we can admit that there is a factor 1/100 between input and ChIP” but I don't know why and in which case? (what represent this factor?). In the same way, there are many ways to calculate the enrichment in the ChIP. Some use interest gene and one negative gene in PCR, others use just the Ct between input, ChIP and IgG… Can you explain me the best calculation, please?
Thanks a lot !

You won't be the first to run into the problem there there isn't a perfect way to express ChIP data.

Quite a number of people look at enrichment over input, as you did. If you do quantitate this way you might want to check and make sure that your background (signal in your negative control IP) is the same for all the regions you are looking at. I say this because we've noticed that background can differ quite a bit between one region and another (even after accounting for differences in the efficiencies of the different primers used). Even better would be to eliminate background all together. However, if you can't eliminate background and if you find that the background differs between the different regions that you are looking at then a good equation to use is:

(Ct IP/ Ct input) - (Ct negative control/ Ct input)

As for expressing data as fold over a negative control region, this has to do with the fact that ChIP data is all relative (i.e. the enrichment you get is not just a function of the amount of protein bound on the DNA but also the efficiency of the IP). Even if you don't express your data as fold over the signal for your negative control primer it's still good to at least show enrichment for your negative control primer.

Joel

-KPDE-

Hi,

Thanks a lot for yours responses.
KPDE, thanks for yours explanations. I think I will use the formula you cite. In another way, I advise everyone to look at the sonixchip’s link (http://www.superarray.com/chipqpcrresource.php) it’s very usefull. There are lots of explanations too.

I think, now, I will be able to analysis my (good ) ChIP results.

But, one thing I don’t understand is the meaning of the factor 1/100 between input and IP. Is-it only because we use less volume of input than IP (20µL against 200µL, for me)? In my case should I use à 1/10 factor? Can someone help me about this factor, please?

-Dukon-

Enrichment of Signal vs. Enrichment of Noise.

This is (IMHO) the best method for ChIP analysis using qPCR. Basically this is signal to noise normalized to percent input. Th ebig problem with ChIP data from qPCR is that, unlike mRNA expression studies where you have an easily use an internal control or standard curve, you usually don't have this option for ChIP. The qPCR technique I use for my analysis goes something like this:

Setup -

Samples of Input (i), TargetRegion (tr), and Neg. Control (nc) are all from the same sonication. 3 qPCR reactons (wells) per sample. We use SYBR green for our ChIP, but TaqMan probes also work as well.

For each sample you end up with an average Ct value and a standard deviation. We'll call them CT.i, CT.tr, and CT.nc with standard deviations of SD.i, SD.tr, and SD.nc. It is extremely important to keep track of your standard deviations becuase otherwise you might propagate your error incorrectly.

Next, calculate the delta Ct values and error associated for your target region and your neg. control samples relative to your input sample. This is basically a log-transformed representation of your "percent input" sometimes seen in papers with ChIP data. Lets call these dCT.tr & dCT.nc and the propagated error values of these dCTs will be called dSD.tr and dSD.nc . These are all calculated using the following formulas:

dCT.tr = CT.i - CT.tr
dCT.nc = CT.i - CT.nc

dSD.tr = sqrt( (SD.i)^2 + (SD.tr)^2 ) / sqrt(n)
dSD.nc = sqrt( (SD.i)^2 + (SD.nc)^2 ) / sqrt(n)
where n=number of replicate qPCR wells per sample (usually n=3)

Once you have your delta CT values and the associate error values in hand, you can calculate your signal to noise (i.e. fold change over your negative control). This done by finding the "Delta-delta CT" of your target region (ddCT); essentially the difference in CT values between your dCT.tr and dCT.nc. Click the thumbnails below to see the ddCT formula.

ddCT = dCT.tr - ddCT.nc
ddSD = sqrt( (dSD.tr)^2 + (dSD.nc)^2 )

Once you have your ddCT and ddSD for your target region, the transformation to linear "fold-change" values (aka signal to noise) is done using these formulas

FC = 2^(ddCT)
FC.error = ln(2) * ddSD * FC

And with that, you have your fold-change and standard deviation of fold change with all the error properly propagated and linearly transformed.

Hope this helps.

-jonathanjacobs-

OH god... sorry for the HUGE images... next time I'll upload smaller ones. LOL.. they are from a powerpoint slide of mine. heh

Also.. the link to the SuperArray biosciences site above provides an excel worksheet that is IMHO kinda odd; can someone remind me why we care what the dilution factor is for the input? it's all relative so it shouldnt matter in the end... plus there's absolutely no inclusion of how to handle the error using SuperArray's methods. Maybe they should include that as well? just my 2 cents.

-jonathanjacobs-

QUOTE (Dukon @ Aug 13 2007, 02:23 AM)
Hi,

Thanks a lot for yours responses.
KPDE, thanks for yours explanations. I think I will use the formula you cite. In another way, I advise everyone to look at the sonixchip’s link (http://www.superarray.com/chipqpcrresource.php) it’s very usefull. There are lots of explanations too.

I think, now, I will be able to analysis my (good ) ChIP results.

But, one thing I don’t understand is the meaning of the factor 1/100 between input and IP. Is-it only because we use less volume of input than IP (20µL against 200µL, for me)? In my case should I use à 1/10 factor? Can someone help me about this factor, please?

Yes, the factor refers to the differences in the amount of input and amount of IP you load for your PCR. To truly be "enrichment" you have to load the same amount of DNA from both your input and IP (or at least correct for any differences in the amount you load; this is the 1/10 factor for you). If you don't load the same amount (and don't correct for the difference) you can still express your data as the ratio of IP/input it's just not truly enrichment. You can just say that you normalized to your input or something similar.

-KPDE-

Yep! Maybe I missed that in my previous post...

-jonathanjacobs-