Hi folks,
I'm trying to determine suitable HKGs for my qPCR experiments. I have two groups with 13 biological samples in each group and so far I measured 11 HKGs. To find the most stable ones I used NormFinder and GeNorm. They both gave me similar results.
However, when I test the recommended HKGs with a ttest between the two groups it shows that some are significantly differentially expressed. That is especially confusing since the HKG which is deemed the most stable by both algorithms has a ttest p-value of 0.002.
So my question is which result is more reliable? The Ttest or the algorithms of NormFinder and GeNorm.
Many thanks in advance
Soner
Hi ! To choose a proper data normalization/house-keeping genes a priori to analysis can be quite difficult. Therefore, we suggest to employ analysis tools, which are independent of the chosen normalization, which means they don't need any data normalization at all. Have a look at our zero-sum regression publications.
Hi Helena,
thanks for your answer.
However, I would still like to know whether t-test p-value or normfinder/genorm algorithms are right.
AFAIK tools like NormFinder and GeNorm do not assume that there are no systematic differences between the groups. Instead, the pattern of expression (distribution of Ct values of a sample across all the tested genes) is analyzed. This pattern should be similar across samples (and thus, also across groups). The tools estimate a general (average) pattern and then remove those genes that diverge most from the average pattern, until any further removal does not anymore change the average pattern of the remaining genes significantly. Eventually, you will have a set of "suitable" reference genes, and you should not select one of them but use all of them together.
Hi Jochen,
thank you for your answer. I will use a set of reference genes that are recommended by the tools. However, at least for NormFinder I thought that it calculates intergroup variations because you can specify group identifiers. Hence, significant expression differences between groups should lead to their exclution. At least, this is what I though the group identifiers are being used for.
I don't know these tools very well. One can include the group in the model to estimate a pattern. This should make sense, I think, because a part of the variance can be explained by (systematic) differences between the groups. This should make the pattern more clear, so that it is easier to identify genes that deviate from that pattern more easily. But as I said: I don't know these tools very well...
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Probability
- Probability Theory
- Reliability