Is labeling important as a QC? I dont have any DEG across 8 diets which is really weird!

Question

[sshibin]

Hi everyone, I used affymetrix mouse gene 1st array for experiments. I have 8 groups/diets done in triplicates, so a total of 24 chips was run. I used the Partek and TAC software for my analysis. When I import cel or chp files and look for genes I got "no differentially expressed gene". Which is surprising as I did a recent one on heart and liver and was successful. This time I am using RNA from hypothalamus. QC for hybridisation is good but fails in labeling. Does labeling really affect data so badly that i get no DEGs? Is there something I am missing?: I import using RMA setting and have library files. I had to manually import files. I set the contrasts and look for DEG. Histogram looks good. But PCA plot suggests within sample variability. I really don't know where it can go wrong :( Any help would be appreciated. Thanks!

[WouterDeCoster]

Just a stab in the dark: how certain are you that you dissected the hypothalamus for your experiments? Any chance of contamination from other cell types? That would result in within-sample variability. I've no experience with isolating brain parts but I know the neuronal transcriptome is pretty diverse.

[Marge]

I agree, adding that besides contamination from neighbouring regions, the large variability could also be due to the fact that you have a variety of neuronal types in the hypothalamus that secrete different types of hormones. So in principle you could have a lot of variation if you don't sample exactly the same part of the hypothalamus. In practice though I am not so sure you are so sensitive.

One more point to consider is whether the differential expression you observe in liver and heart are the only indications that you should find differential expression in the hypothalamus. I can easily imagine that liver goes through heavy changes in response to diet (after all it's a central player in metabolic regulation). I can also imagine that both liver and heart go through structural changes in response to diet, and this as well would be linked to important changes in gene expression. The hypothalamus however might have small changes in terms of hormones secreted, but the expression profile is going to be dominated by the fact that hypothalamus is hypothalamus.

Unfortunately I am not familiar with the labelling failure you describe.

Answer 1

[sshibin]

It is the first time we tried hypothalamus, we confirmed that it is hypothalamus with the specific genes present but also found one gene in relation to pitituary gland. How about the labeling after the fragmentation process, could that have anything to do with it? I mean could it affect in detecting the intensities across arrays because each and every probe have the same intensity across all diets and triplicates..

[WouterDeCoster]

Please use ADD REPLY to answer to an earlier comment, as such this thread remains logically structured and easy to follow.

My knowledge of technical aspects of microarrays is... very limited, so that's not a question I can answer.

Answer 2

[sshibin]

It is the first time we tried hypothalamus, we confirmed that it is hypothalamus with the specific genes present but also found one gene in relation to pitituary gland. How about the labeling after the fragmentation process, could that have anything to do with it? I mean could it affect in detecting the intensities across arrays because each and every probe have the same intensity across all diets and triplicates..

[mastal511]

I don't really know much about the QC for labelling, I assume there are some control probes on the microarray which should indicate whether the labelling process worked as expected or not. However, the labelling step determines what intensities you get for each spot on the array, so if it didn't work properly for some reason, then your data is unreliable.