[samoth]

yes thats right.
The gene i investigated is not the sole gene related to the phenotype. but my project is to determine the effect of the expression of this one gene on the phenotype.
gene expression was carried out using real time qPCR measuring relative expression (using TaqMan MGB probes)
What do you mean with "2) how about the ones with extremely low expression levels? "
thanks

[Simon Anders]

Interesting discussion, and good idea to occasionally go back to the basics here on the forum.

A few more points:

If a cell want to change its protein inventory in reaction to a stimulus, it can do so by changing the production, the translation or the degradation of mRNA. So, if you want to know how the cell reacts to a stimulus, one would want to measure all three processes genome-wide in control and treatment samples and check where one sees a statistically significant change. Of course, we can only measure mRNA level (i.e., the combined effect of production and degradation) in a convenient way, so we may miss the cell's main response mechanism. However, if we find something, we can be sure that it is in fact a genuine response of the cell to the stimulus (even though we should not jump to the conclusion that it is the functional aspect of the response rather than a byproduct).

Just to be clear: By statistical significance, I mean that we can be confident that the change in expression level when comparing treatment with control is large enough that it is unlikely to be due to the random variation in expression that we also see when comparing replicates. To test for this, we need a test that uses replicates to estimate this within-group noise, and Fisher's exact test (mentioned in the original post) does not fulfill this criterion and hence is not suitable.

By biological significance, we mean that the observed effect is likely to be a functional part of the cell's response to the stimulus, as opposed to a mere byproduct, such as an unrelated downstream effect of the actual response. A reasonable indicator for this is in fact that the fold change is not too small, and a plausible veto might be that the change is unlikely to cause changes in protein abundance, even though I do not think that saturation of the translation machinery is a common effect. (Look at ribosomal proteins: Why would the cell produce such masses of mRNA, orders of magnitude more than average genes, if it hadn't the resources to translate them all.)

Finally: The mass-spec folks have made some amazing advances recently, and comparing mRNA and protein levels in at least a semi-genomewide fashion will be a big topic soon, I think. A limitation here will be that we need time courses for this. After all, we do not expect mRNA levels to be proportional to protein levels, but to be proportional to the derivative with respect to time of the protein level.

[Xi Wang]

Interesting discussion, and good idea to occasionally go back to the basics here on the forum.

A few more points:

If a cell want to change its protein inventory in reaction to a stimulus, it can do so by changing the production, the translation or the degradation of mRNA. So, if you want to know how the cell reacts to a stimulus, one would want to measure all three processes genome-wide in control and treatment samples and check where one sees a statistically significant change. Of course, we can only measure mRNA level (i.e., the combined effect of production and degradation) in a convenient way, so we may miss the cell's main response mechanism. However, if we find something, we can be sure that it is in fact a genuine response of the cell to the stimulus (even though we should not jump to the conclusion that it is the functional aspect of the response rather than a byproduct).

Just to be clear: By statistical significance, I mean that we can be confident that the change in expression level when comparing treatment with control is large enough that it is unlikely to be due to the random variation in expression that we also see when comparing replicates. To test for this, we need a test that uses replicates to estimate this within-group noise, and Fisher's exact test (mentioned in the original post) does not fulfill this criterion and hence is not suitable.

By biological significance, we mean that the observed effect is likely to be a functional part of the cell's response to the stimulus, as opposed to a mere byproduct, such as an unrelated downstream effect of the actual response. A reasonable indicator for this is in fact that the fold change is not too small, and a plausible veto might be that the change is unlikely to cause changes in protein abundance, even though I do not think that saturation of the translation machinery is a common effect. (Look at ribosomal proteins: Why would the cell produce such masses of mRNA, orders of magnitude more than average genes, if it hadn't the resources to translate them all.)

Finally: The mass-spec folks have made some amazing advances recently, and comparing mRNA and protein levels in at least a semi-genomewide fashion will be a big topic soon, I think. A limitation here will be that we need time courses for this. After all, we do not expect mRNA levels to be proportional to protein levels, but to be proportional to the derivative with respect to time of the protein level.

I totally agree with Simon's points. It is more important to go back to the original biological questions when analyzing biological data.

In my mind, once people can image all the processes that occur in cells between molecules, how DNA goes into protein products will be clear. However, nowadays, due to the limitation of current technologies, we can only measure gene expression levels at the RNA level and seldom at the protein level, which is far from the whole picture. The lack of knowledge makes inferences very noisy, and many discoveries are not believable. The combination of RNA-seq and mass-spec will be a way helps to solve this issue in the near future, although not thoroughly.

[Xi Wang]

yes thats right.
The gene i investigated is not the sole gene related to the phenotype. but my project is to determine the effect of the expression of this one gene on the phenotype.
gene expression was carried out using real time qPCR measuring relative expression (using TaqMan MGB probes)
What do you mean with "2) how about the ones with extremely low expression levels? "
thanks

By "2) how about the ones with extremely low expression levels? " I mean how about the 6 subjects in the group with lower expression levels. Why did you also analyze this group? Some different phenotype?