I am pretty sure that you did not show that there was "no change" when normalizing to RSP18. I wonder how you get to this conclusion. I think you made some test (e.g. a t-test comparing the dCt values tumor vs. normal) and the p-value was too large (e.g. >0.05), so that the data was "not significant". This does *not* mean that there was no change. It means that your data is insufficient to conclude the direction (up or down). It may be that both results show the same, but that the data for RPS8 is more noisy (or the sample size was lower), so that it did not reach "significance".*

However, if you have several putative reference genes, then better use them all together (not just one)!

PS: HIF1A is regulated on protein-level (it is produced at high rates and destroyed as it is produced, leaving low steady-state levels, that can rapidly increased when the the decay is inhibited). In order to have a relevant impact on HIF1A signalling, I would expect that HIF1A mRNA levels need to be changed rather drastically (what should be easy to detect even with suboptimal reference genes). If the change on the mRNA level is rather modest, you will need a good explanation how this can be relevant.

----

*

(i) you can use the confidence interval to see if the results are strictly incompatible with "relevant changes"; in this case you demonstrated that the change is at least not relevant.

(ii) you can compare the effect under RPS8 normalization to the effect under ALAS1 normalization to see if the data is sufficient to conclude a lower effect under RPS8 normalization than under ALAS1 normalization. But this is not a relevant question here. The question remaines if RPS8 and or ALAS1 are (slightly) differentially expressed in normal and in tumor cells.

Jochen Wilhelm I used the same sample to assess gene expression for both ALAS1 and RPS18. I didn't do any stats yet. And I only was looking at the "Gene expression bar plots" of the StepOne software. Here, I just look how many x-fold a gene drops or increases and from there decide whether I should keep considering to investigate this gene or not.

For example, no matter what reference gene I use, I do see considerable drop in my target genes upon KD.

Zhanna Orlova

Reference or housekeeping genes can vary based on the effect a gene being knockdown has. I was experiencing the same issue when I was doing a knockdown with certain transcription factors where one TF knockdown the housekeeping gene was fine, while a different TF knockdown effected the reference gene. To solve this, instead of using a reference gene I recommend absolute quantification. You create a standard curve with purified DNA (usually down to 100fg concentration) and plot the standard curve to get an estimate of amount of gene expression. There are alot of protocols out there for absolute quantification but once you get the standard curves down it becomes way easier to work with because then you do not need to worry about a reference gene.

Shawn Krueger thank you for your explanation and suggestion

Zhanna Orlova

Hope it helps, alot of people resist absolute quantification with RT-qPCR because everyone uses reference genes, but when you run into situations where reference genes are changing you have to use a more precise method.