qRT-PCR reproducibility of triplicates? - (Jul/02/2008 )
I am doing some quantitative RT-PCR and having some difficulty getting the changes I expect. Unfortunately, I only expect to see changes of about 25-30% in my gene after manipulation and I seem to get Ct variance of 0.2-0.3 within my triplicates.
For example, I am performing 25ul SYBR reactions on an ABI machine (7000 or 7900). I aliquot 22.5ul of master mix into all the wells of my plate and then I aliquot 2.5ul of cDNA #1 into 3 wells, 2.5ul of cDNA #2 into 3 wells, etc. The problem is that each well within a triplicate can vary from the others by 0.2-0.3 Ct, which is close to the change I expect to see between cDNAs.
Is this the normal variance people observe? Am I setting up my reactions incorrectly? Should I make a sub-master mix for each cDNA and then just aliquot it into three wells? This would seem to provide a less stringent error control to me. Would it help to do 20ul reactions and measure out 2ul of cDNA with a P2 rather than 2.5ul with a P20 pipetman?
Thanks!
Kurt
For example, I am performing 25ul SYBR reactions on an ABI machine (7000 or 7900). I aliquot 22.5ul of master mix into all the wells of my plate and then I aliquot 2.5ul of cDNA #1 into 3 wells, 2.5ul of cDNA #2 into 3 wells, etc. The problem is that each well within a triplicate can vary from the others by 0.2-0.3 Ct, which is close to the change I expect to see between cDNAs.
Is this the normal variance people observe? Am I setting up my reactions incorrectly? Should I make a sub-master mix for each cDNA and then just aliquot it into three wells? This would seem to provide a less stringent error control to me. Would it help to do 20ul reactions and measure out 2ul of cDNA with a P2 rather than 2.5ul with a P20 pipetman?
Thanks!
Kurt
Hi,
In my experience making a sub-mastermix with cDNA before aliquoting and then adding primers is better than adding cDNA at the end. This procedure reduces the differences between your repilcates' Ct, but i'm not sure you can consider a 0.2-0.3 Ct difference as significant. I usually count my replicates as "significant" if they differ between them less than 1 Ct.
What about the protein? making a western blot is more time consuming but could be the only way to see a 30% difference.
By the way, if you have to measure 2.5 ul, why don't you try a P10 instead of a p20?
Bye
Fizban
Wouldn't you expect a sample replicate to be much closer than 1 Ct? That's a two-fold difference in copy number.
You're right, I should use something other than a P20, but we don't have a P10. I'm planning to move to a 20 ul rxn and using a P2.
Unfortunately, we don't have an antibody that works, so performing Westerns is out. Good idea though.
What is the average Ct that you would expect?
Generally, with high Ct numbers (> 30) one should expect notable variability between the replicates.
Generally, with high Ct numbers (> 30) one should expect notable variability between the replicates.
I get Ct values around 20 for my samples.
So I'm still interested in answers to my original question. If you get variability of 1 Ct, that equates to 2x gene expression. 0.5 Ct equates to a 40% change in gene expression. So if you get standard deviations for three replicates of 0.5, then there is no way to measure gene expression differences in this range. I'm just curious to know if I am trying to do something that isn't really done very often. Is it practical to measure a 30% gene expression difference? Do I just need to increase my sample replicate to 6 from 3 to get the standard deviation down?
On a related note, what is the minimum volume people use for qRT-PCR. I'm using SYBR green Power Mix on an ABI 7900 and I've been doing 25ul samples. Is it possible to go down to 10 or 15ul samples?
Generally, with high Ct numbers (> 30) one should expect notable variability between the replicates.
I get Ct values around 20 for my samples.
So I'm still interested in answers to my original question. If you get variability of 1 Ct, that equates to 2x gene expression. 0.5 Ct equates to a 40% change in gene expression. So if you get standard deviations for three replicates of 0.5, then there is no way to measure gene expression differences in this range. I'm just curious to know if I am trying to do something that isn't really done very often. Is it practical to measure a 30% gene expression difference? Do I just need to increase my sample replicate to 6 from 3 to get the standard deviation down?
On a related note, what is the minimum volume people use for qRT-PCR. I'm using SYBR green Power Mix on an ABI 7900 and I've been doing 25ul samples. Is it possible to go down to 10 or 15ul samples?
Huh, lotsa questions!
With Ct of 20 your replicates should really be neat. Have you maybe considered a stepper (multi-step pipette)? I've seen ppl getting perfect results with them.
To reflect on your thoughts on expression and variability, you are, unfortunatelly right. If your between-the-replicates variability is higher or comparable to the expected expression change, then you have no way of detecting it.
I have seen ppl reporting even smaller differences as "significant" (because of the P-values) but then, you have to think about the biological significance of 30% change in your system. IMHO, differences lower then at least 2-fold (100%), I find difficult to find of relevance.The biochemistry will work fine if you scale down to 15 uL. What you should worry about is the optics (i.e. where the focus of the laser is). If we talk about 96-well 7900 standard, then the minimal validate volume is 20 uL. With 7900 Fast or 384-plate you could go down to 10 uL. Try to downscale, but validate it carefully, and I would not do it if I am not happy with my assay performance in the "standard" volume.
Pallas is absolutely right. A variation of 0.2 - 0.3 on the ABI 7900 is relatively typical and reducing the volume will increase your deviations - so it doesn't make sense to reduce your volume to half and then making twice as much replicates. Doing a master and then pipetting your sample on top is the usual and correct method to create your replicates (in my opinion)!
However if you want to decrease your variation I would go for a newer qPCR instrument (the ABI 7900 is almost a dinosaur in the qPCR field). We had one in our lab before, but changed to LightCycler 480 last year. They have really a lower variation because of the better temp. homogenity on their block and the better optic which does not need a reference dye which can not compensate for an old optic and an insensitive laser. In combination with their FastStart Master optimized for this instrument (and not for a whole group of systems) and the white plates (not so nice to handle, but more sensitive) I think they are more precise.
Maybe you find a collegue or working group who is already working with the LC 480.
Whish you much success to quantify your challenging gene expression difference...