Count Matrix normalisation for downstream analysis and for creating heatmap of targeted genes

Question

[Vikram]

Hello Everyone!

I have a count matrix generated from stringtie (from FPKM to readcount using prepDE.py3 of stringtie). I would like to create heatmap of targeted genes across samples.

My questions are:

1. Before creating heatmap do we need to normalize the read count data
2. If it is necessary, I was thinking of normalizing something like this, given gene-x and its read count across control, treatment1,treatment2 in triplicates, I'll calculate the average of all the read counts of gene-x across samples and then with that ill divide each read count by averaged read count and use that resulting value for plotting

Is this right way to proceed ?

If you have any other methods please suggest.

Thank you!

Answer 1

[rpolicastro]

Instead of using the stringtie counts you should quantify your reads against your assembled/merged transcriptome with Salmon or Kallisto. This will give you more accurate TPM abundance estimates at the gene and transcript level. For normalization after this DESeq2 rlog/vst, or edgeR TMM all work well. See their respective documentation for more information.