Correlation vs differential expression testing

Question

[prabin.dm]

Hi,

I have list of 25 genes for which I have done qPCR across 3 different samples. When I do statistical analysis only few only them are differentially expressed.

However, if I do a sample-wise correlation analysis, the gene expression data show clear correlation between 2 samples.

My question is, would correlation analysis (probably with heatmap) be an acceptable way to present the data if only few genes are statistically differentially expressed ? If so, please suggest some references regarding that.

Answer 1

[n,n]

It all depends on what the intent of your project is. Differential Expression Analysis will tell you if certain genes are notably expressed in some of your samples relative to all samples. If you do correlation between samples, the information you are getting isn't quite the same; with correlation between samples you find out how similar the expression profiles (taking only the genes of your analysis into account) of your samples are (careful since this could be for biological reasons or because of batch effects).

[prabin.dm]

Thank you. I have 3-4 replicates for each gene, so I hope I am excluding batch effects. And similarity in gene expression profile is what I intended to show.

Answer 2

[Devon Ryan]

You can certainly present the results in a heatmap, but of course that won't change the fact the there's no significant difference in the results. If there being obvious correlation is sufficient to support the broader argument that you're trying to make then great, otherwise it may not help you much.

[prabin.dm]

Thats great to know. Thank you.

The correlation is very important for our study, as with DGE too I would have gone on to demonstrate the same thing, that 2 samples are have similar expression prolifes and the third one stands out.