I want to compare gene expression(37 genes) patterns in mitochondrial among 3 different samples and I come up with those methods:

1. I've got the TPM value of these 37 genes in different samples. Considering TPM can account for the relative expression ratio among genes, I did z-score normalization for 37 genes TPM in one sample and repeated this process to all three samples. After that, I plotted normalized expression in one figure like this. Thus we can compare the expression patterns.
2. I can use the raw read counts as input and use TMM to correct library size. Then I calculate the TPM value. Will library-size-corrected TPM value can be used to compare gene expression between and within samples?
3. I just do differential expression analysis of these 37 genes using tools like edgeR.

How do these methods sound?