RT-PCR Data analysis of a non-longitudinal study on animals

[isabelle]

Helllo,
We are currently running a gene expression profiling analysis on animal (rat) samples. Our aim is to monitor the expression level of specific markers according to the animal age. This is a non-longitudinal study : the samples come from different animals. We have about 10 animals / group of age. Since we only did gene expression profiling analysis on cell culture samples, we are very surprised by the differences in the expression level between animals from the same group of age. This is reflected in the SEM that we measure for each animal group. Is it possible to exclude one sample from a group when the difference to the other samples of this group is very big (10 times) ? If this is possible, according to which criteria can we do it ?
I have another general and naïve question about GENex : is the formula used for the t-test in GENex the same as the one used in Excel ?
I thank you for your help,
Isabelle

[Mikael Kubista]

The spread of data collected on animals is larger than on cell culture samples due to the contribution of individual variation in the mice. See:

Ales Tichopad, Rob Kitchen, Irmgard Riedmaier, Christiane Becker, Anders Ståhlberg, and Mikael Kubista. Design and Optimization of Reverse-Transcription Quantitative PCR Experiments. Clin. Chem., Jul 2009; doi:10.1373/clinchem.2009.126201

Using the Nested Anova available in GenEx (ver. 5) you can estimate the contribution from the individual variation and then design your study adequately. The experimental design tool will guide you.

You can test for outlier’s with GenEx (available in data pre-processing) to see if you can motivate excluding a suspicious sample based on anomalous Cq value, Be careful though if you have many sets of groups, since the rate of false positives increases if multiple testing is performed.

Good luck!

[MultiDAdmin]

GenEx uses the following formulas to calculate the t-statistics,

Unpaired: [mean(group1) - mean(group2)] / [stdev / sqrt(1/n1 + 1/n2)]
Paired: mean(difference) / [stdev(difference) / sqrt(n)]

To calculate the p-value that corresponds to a t-statistic (and degrees of freedom), the Lohninger MathPack is used. I assume that Excel uses the same standard formulas for the t-statistics, but I find no information of what method they use to calculate the p-value. Since Student's t-test is a standard test that is widely used, I find it hard to believe that Excel gives results that differ from GenEx'.

/Anna