If you normalise everything to your WT Control then the measures for all groups become relative to it and you can make a comparison between any subset of groups. That way you can identify any differences between strains that aren't a result of the specific treatment without doing any further calculations. It will also make subsequent statistical analysis easier and make plotted data more intuitive.

Hey Mike,

I'd say why not do both? It sounds like both sets of data are worthwhile and it's nice to have more data around; makes choosing what goes in the manuscript easier!

I think that you should use as calibrator for treated strain its own saline control for ddCt value. After that you should compare data for two different strain and obtain the differences due to specific genetic bacground of the transgenic animals.

If you normalise everything to your WT Control then the measures for all groups become relative to it and you can make a comparison between any subset of groups. That way you can identify any differences between strains that aren't a result of the specific treatment without doing any further calculations. It will also make subsequent statistical analysis easier and make plotted data more intuitive.

Neil I'm leaning towards your solution, however I have nearly 100 targets from an array and it would be difficult to show the saline and the treated for both strains simultaneously in a concise way. It is easier to visualize with Spaska's solution since I wouldn't have to plot the saline. What I may end up doing is to break it down and present multiple graphs of certain families of genes from the array and show all data, including the transgenic animals' saline relative to the WT saline. The genetic differences between strains are profound enough that I don't think I can get away with merely showing the change in expression, that would be misleading. I need the relative levels between the strains too.

Thanks for the input guys, just needed a few more voices on this one!

If you assume that the expression of your "comparator" is truly constant across all of your four groups, then just calculating 2^(-1*dCt) [where dCt = Ct for your gene of interest - Ct of the comparator] will give you a relative expression level for each sample/animal on the same scale across all of your groups--and you can then make any comparisons or statistical calculations that you like. Probably more important to test different "comparators" to be sure that expression really is constant or at least that using a different comparator does not give a different overall result. (In general, performing a ddCt type calculation doesn't change the relative levels; it only affects which group will appear to have a level of "1".)

If you are plotting qPCR data, it is probably best to do it on a gene by gene basis and yes you can then organise the graphs into panels by gene family or some such, especially if you are dealing with large numbers of genes. if you are trying to plot micro-array data then the traditional method is to use a heat map of some description. Alternatively you can put the normalised micro-array data into a program called BioLayout Express 3D. In that you can carry out cluster analyses and you can group data by associated annotation (such as gene family if you add the appropriate information) and it will produce graphs for you if that is any use...

I guess you are referring to what I call “the calibrator” for the calculation of the 2^-ddCt. Sample dCt = Ct endogenous gene – Ct target gene; ddCt = sample dCt – calibrator dCt; calibtrator = dCt of control sample. For example: if you are comparing treatment versus no treatment control, your calibrator is dCt of the no treatment sample; if you are comparing patients versus healthy donors, your calibrator is the median value of the dCts of the healthy donors. In your case, you need to represent the data in two different graphs. In one graph I would show the data as 2^-ddCt values and compare each drug treated strain to its own calibrator (saline control). In another graph I would show data as fold of decrease and/or increase of gene expression in the transgenic strain relative to the wild-type strain.

The fold of decrease and/or increase of gene expression will be calculated as transgenic 2^-ddCt - wild-type 2^-ddCt

Totally agree with Neil. Because of the genetic background, the saline groups may also have a difference, if you calibrate the treated groups with their own saline groups, the difference between treated groups may not be seen.

I agree with Xin, As suggested by most of the researchers you first check for the gene expression with respect to the saline treated animals this will give you the alteration in the gene due the drug followed by the comparison with the othe strain which also is calibrated with its saline treated group. So at the end you will get values exactly showing the level the expression between the two species. All the Best!!!

Mike,

I am also working with gene expression analysis using two to three variables. I was having the same issue as you are having, and I ultimately decided to normalize the target genes by the internal control gene (dCT) and then present this normalized expression. You can then use ANOVA with Tukey's post hoc test to categorize each sample type as the same or different from every other sample type. This is easier to show in one graph.

You need to find a gene that does NOT change in both your control and your experiment for relative controls. You should also consider using a standard curve - a serial dilution of your amplicon of interest previously cloned into a plasmid that can be used repeatively in different experments. This will also permit obtaining copy # of mRNAs - so you can see the variability of expression relative to your "housekeeping genes". Or you could redo using the new digital qPCR from BioRad - that will give you absolute amounts. Check the MIQE guidelines for more information.

Julio- The endogenous controls have virtually identical Ct values before and after treatment and between animal lines so I'm OK there, but many of my targets show significantly different expression levels between animals even before treatment. So if I only show changes in expression (which for instance may be 2 fold in both animals) I will miss the fact that the relative expression may be much higher in one animal compared to another since basal expression levels are so different.

So I seem to be getting answers on both sides! Some would suggest to compare raw expression level changes in gene activation, while others suggest the copy number of transcripts is most important. I guess none of it really matters since transcription does not necessarily equal function anyway. A high basal copy number may indicate a translational block or rapid protein turnover feeding back on the gene itself, so as long as I'm transparent in my approach and cautious in my conclusions hopefully the reviewers won't torpedo me!