For a set of miRNA gene expression data from different cancer patients I produced some heatmaps with R. The goal is to examine which miRNAs are up/downregulated depending on different parameters (state, occurence of brain metastases, etc). I could not figure out whether to use euclidian distance for the calculation or a correlation analysis (spearman).
The patients didn't receive any treatment so I wouldn't consider the shape to be that important but rather the actual distance between the datapoints. On the other hand I read that a lot of people are using correlation analysis for gene expression data. Is it just a matter of personal choice which function to use here or is there a right and wrong way? I am quite new to statistical analyses so I would appreciate any help!
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
1
You can use correlation or distances to see which patients are similar in terms of gene expression. I don't quite understand why you would use distances or correlation for your problem.
I think what you're looking for is a differential expression analysis since you want to find up/down regulated genes in a certain group against another one.
You can find packages on Bioconductor such as DESeq2 or limma to do that. But there's also some specific packages adapted for microRna like AgiMicroRna but are maybe a bit more complicated to use.
You can have a look at this tutorial for generating a heatmap of differential expression analysis.
https://www2.warwick.ac.uk/fac/sci/moac/people/students/peter_cock/r/heatmap/
-
$\begingroup$ Thank you very much for your answer. I asked about correlation and distances not for the analysis itself but for the distance function in the heatmap() argument. The data is from an RNA-microarray and we just examine a subset of miRNAs (55) so I thought it's a viable option to have a look at the similarity of the miRNA-gene expression. $\endgroup$– brahminCommented Oct 20, 2017 at 12:11