Ho to calculate significance?(p-value) - (Oct/23/2007 )
Is there any way of calculating significance of the difference in the "fold differences" between two genes? For example if gene A is 12 times up regulated and gene B is 3 times up regulated when compared to a negative control. How will 'students t-test(two tailed)' be helpful?
Thanks
If your data are not normally distributed, you have to try a transformation to make them so. Like log transformation. Then you can use t-test. If you dont have normally distributed data, then you'll have to use a non-parametric test.
A more important question is: Is there _any_ biological meaning to the question you ask. If the means/medians are significantly different, what does that mean? Different genes have different mRNA half-lifes, splicing, post-translational modifications, phosphorylation, etc. IMHO a significant difference in mRNA amounts of two different genes does not carry biological significance. All you can say is that there is a larger increase in case of gene A then gene B. And the reviewers can say: So what? 
I am sorry for putting the question completely wrong. To be straight:
I have a gene X for which I need to know the status in 3 types of cells. Cell type A is a positive control,(say 12folds up) a negative control (no upregulation) and the cells of my interest which indeed shows 3 folds up. Therefore when a positive control shows a 12 folds up is there any significance for the 3 folds up which I saw for my cells of interest.
Getting there. So there's only one gene.
Three groups: + control, - control, treatment
What do you compare them to? An earlier timepoint? Or are the + and treatment compared to -?
I have a gene called say CAL which is over expressed in cancer cells such as HeLa which is the +ve control. A normal primary cell line such as skin fibroblasts is -ve as the gene is known to be not expressed. We want to check the levels of the same gene expression in a cell (a kind of stem cells) where we do not know if gene is expressed and if yes is it very low, moderate or high. Considering the result that when compared to fibroblasts HeLa shows 12 folds up and my cells of interest show only 3 folds up I want to know if this increase or amount of expression is significant. Crux is parental cells my cells of interest also do not show upregulation(ie equal to -ve control). Sorry for making it too complicated!
Okay, so what we are dealing with here is comparison of three groups. First you have to do an ANOVA, to see if there is any difference between the three, and then do pairwise comparisons of the three groups (a-b, b-c and a-c) to see which groups are different from each other. Because the comparison of negative control and positive control does not make sense, you can skip that one test, with this you can increase the power of your analysis.
Things to remember: if the ANOVA is negative, you can not do the pairwise comparisons. If your samples can be transformed into normal distribution, an ANOVA and t-test is applicable. In any other case, you have to choose a non-parametric test for both ANOVA and pairwise comparisons. (e.g. Kruskal-Wallis test and pairwise Wilcoxon rank sum test) Don't forget to adjust the p value of your pairwise tests (google for bonferroni correction).
BTW: your fibroblasts must be expressing your gene, otherwise you would not have a Ct value, and you wouldnt be able to compare the other two groups to these ones.