qPCR analysis with error bars

[Siefker, Ed B.]

I'm sorry if this isn't the right list for this. Hopefully data analysis
counts as a "method". If there's a better list for this, please let
me know. I was surprised to not find a biostatistics list.

I'm trying to plot some qPCR data in a bar chart as fold change
with error bars. I understand the ddCt calculation just fine, but
I can't find any information on how to carry standard deviations
through the calculation.

I have two conditions I'm looking at, control and mutant. I have
three biological replicates for each. For each of those, I look at
a control and reference gene. For each gene I do 3 qPCR reactions.

>From that I get a mean Ct and SD for the triplicate PCR reactions.
To get the dCt, I subtract the reference from the control. What do
I do with the standard deviations?

I see a lot of info out there about qPCR, but no one answers this question.
Am I missing something simple here?

[TR]

Have a look at this manual:
http://www3.appliedbiosystems.com/cms/groups/mcb_support/documents/generaldocuments/cms_042380.pdf

Step 6 (pag. 58) has the following title:
Incorporating the standard deviation of the ΔΔCT values into the fold
difference.

[Siefker, Ed B.]

________________________________________
From: methods...@oat.bio.indiana.edu [methods...@oat.bio.indiana.edu] on behalf of TR [tas...@gmail.com]
Sent: Thursday, December 01, 2011 2:59 AM
To: met...@magpie.bio.indiana.edu
Subject: RE: qPCR analysis with error bars

_______________________________________________

Yes, I've seen that. Their example problem doesn't
contain any biological replicates. Taking biological variation
through the calculations is what is confusing me.

I also find their "step 5" to be very suspect.

"The calculation of ΔΔCT involves subtraction of the ΔCT calibrator value. This is
subtraction of an arbitrary constant, so the standard deviation of the ΔΔCT value is the
same as the standard deviation of the ΔCT value.

Therefore, ΔΔCT Drug Treatment A sample =
ΔΔCT = 4.37±0.10 – 6.86±0.17 = –2.5±0.10"

6.86±0.17 sure doesn't look like an arbitrary constant. It looks like a measured
value with uncertainty that will affect the outcome of the calculation.

[Peter Ellis]

>
> 6.86±0.17 sure doesn't look like an arbitrary constant. It looks like a measured
> value with uncertainty that will affect the outcome of the calculation.

Yes. The ±0.17 is the uncertainty in your reference (untreated) sample
though, not in the test sample. You should be plotting both in your
final figure, with the appropriate uncertainty on each.

So, in page 58 Step 5, after calculating ddCt, i.e. calibrating to the
untreated sample, for you have:

Drug A: -2.5 ±0.10
untreated: 0.0 ±0.17 <-- NOTE THAT THIS ZERO STILL HAS AN UNCERTAINTY

This is all clearly shown in Table 11, along with the figures for the
other two drug treatments. One place confusion may be coming in is that
for some reason Table 11 and the subsequent workings on pages 57-59 use
different figures for the s.d. for untreated/GAPDH. Specifically, in
table 11 this is given as ±0.09, whereas on page 57 (final line) they
use ±0.022. Carrying this difference forward is why page 59 line 3
gives the untreated dCt value as 6.86±0.17, compared to the value of
6.86±0.19 shown in table 11.

In all other respects the workings are accurate. The uncertainty in the
untreated sample stays with the untreated sample final value, the
uncertainty in the Drug A sample stays with the Drug A final value, and
so on.

Peter

[TR]

> Have a look at this manual:
> http://www3.appliedbiosystems.com/cms/groups/mcb_support/documents/generaldocuments/cms_042380.pdf
>
> Step 6 (pag. 58) has the following title:
> Incorporating the standard deviation of the ΔΔCT values into the fold
> difference.
> _______________________________________________
>
> Yes, I've seen that. Their example problem doesn't
> contain any biological replicates. Taking biological variation
> through the calculations is what is confusing me.
>
> I also find their "step 5" to be very suspect.
>
> "The calculation of ΔΔCT involves subtraction of the ΔCT calibrator value. This is
> subtraction of an arbitrary constant, so the standard deviation of the ΔΔCT value is the
> same as the standard deviation of the ΔCT value.
>
> Therefore, ΔΔCT Drug Treatment A sample =

> ΔΔCT = 4.37±0.10 - 6.86±0.17 = -2.5±0.10"

>
> 6.86±0.17 sure doesn't look like an arbitrary constant. It looks like a measured
> value with uncertainty that will affect the outcome of the calculation.

-------------------------------------------------------------

Well,
I am not statistician, we probably need one here. However, the
subtraction in step5, as I see it, serves only to give an arbitrary
value of 1 to the expression of your gene in the calibrator. Actually
you do not need to do it, the bars of relative expression for your
samples would look just the same, only that the scale in the axis
would be different: instead of "expression relative to the calibrator"
you would have "expression in arbitrary units": Imagine a set of
samples, from which you consider the sample A to be the calibrator:

If you do the subtraction, the relative expression could be:
A: 1
B: 2
C: 4
D: 8

If you do not do it the relative expression in arbitrary units would be:
A: 300
B: 600
C: 1200
D: 2400

The biological interpretation of the experiment would be the same in
both cases, since you do not know, in any of them, any absolute value
for the amount of the specific mRNA. So the standard deviation which
is relevant is the one that is used in the second case, that is, that
of ΔCT.

Hope this helps,
TR.

[strish...@gmail.com]

can some one please attach the document from applied biosystems, as the link is no longer working.