In my experiment I would like to find out if GOIs are differentially expressed in different tissues and if their expression varies through the year. Animal samples were taken at winter, spring, summer and autumn. Each season, samples were taken from four different animals and from each animal four different tissues were taken. qPCR was performed in triplicates. Just to clear up, I have for e.g. winter 48 data points = 4 (biological replicates) * 4 (tissues) * 3 (technical replicates).
During quality control in qbase+ software, technical replicates passed with flying colours. All technical replicates were within 0,5 cycle limit. However the sample quality is another story. In qbase+ manual is written that there are three ways to evaluate samples, i.e. normalization factors, detected targets and average Cq value. Normalization factors allows you to inspect possible experiment problems (different amount of RNA, low cDNA quantity, presence of inhibitors,..). Using equal amounts of equal quality input material and stably expressed reference genes should resulted in similar normalization factor values for all samples. It is written that a variation of 2- to 3- fold is generally acceptable. There is also example provided in manual about potential issue for one sample. Now I'm wondering what is meant by 2- to 3- fold variation? Variation between minimum and maximum value of normalization factors? In all my samples (192) normalization factors range from 0.533 to 1,793. The difference is more than 3-fold as you can see. However if I group my samples by tissue then normalization factors range from 0.926 to 1.074. I'm confused what to do. Do I exclude some samples or not? Based on what do you decide to exclude biological replicate from analysis?
Hi,
I think the factor 2-3-fold is a bit arbitrary and cannot be applied to all experiments. As you already told us- the comparison of the different tissues causes the "high" variation. But, especially if you compare different tissues it will be not easy to find a "universal" set of reference genes that is stably expressed in all tissues. Maybe there are genome wide expression datasets available for the tissues and you can have a look for alternative reference genes.
But the exclusion of biological replicates will likely not improve the situation if you still want to compare different tissues and introduces the risk of some kind of bias.
Maybe you have the chance to validate some differently expressed genes by alternative methods (NorthernBlot, immunohistochemistry) to enhance you confidence in the results.
I Agree with Norman.
You have a very difficult problem to solve. As you want to compare different tissues (i.e. cell types!) it will be difficult to find suitable refernce genes. Note that a "housekeeping gene" is simply a gene that is expressed in all cell types at all developmental stages, but not neccearily always at the same level. In contrast, a suitable reference gene only needs to be expressed in the cells you are analyzing, but at a constant level. Thus, not any housekeeping gene will be a suitable reference gene (and not all suitable reference genes need to be housekeepers). If you would compare the same cell type under slightly different conditions, it is likely that the expression of housekeeping genes will be quite stable. But this is not the case when you want to compare different cell types. So That is your key problem.
You can easily spend years searching for a good set of suiatble reference genes. The common way is to test a whole bunch of candidates and analyze their co-expression patterns. If you see a group of candidates with a clearly recognizeable co-expression pattern over the different conditions/celltypes then it is likely that the pattern is the result of varying input and the genes are not differentially regulated (and, thus, are suitable reference genes).
If you really could control the amout of input material well, then it would be simpler and more effetive to do this and to completely bypass all the use of reference genes. Instead only some control of the Ct calculation should be included. Here you can use a simple standard that is measured in each run (the Ct values of different runs are shifted so that the Ct values of the standard in all runs are equal).
- Badges
- Science method