I am working with a dataset of nearly 100 patients. I performed Salmon quantification with genecodeV34 and imported the results with tximeta.
I normalised the TPM salmon output with TPM (using edgeR, CalcNormFactors + cpm).
I took the values for my gene of interest and feed mclust with it. Usually mclust identify several groups (mclust E, univariate, equal variance). Sometimes (with other genes) mclust cannot detect more than one group, so I use means with k=3 (to have the low, int, and high expression groups), but not sure if this is a good practice.
Should I perform clustering (mclust for example) using all the genes before and see how the patients clusters?
Sorry, I am quite lost on biostatistics, so any help will be very welcome!