Entering edit mode
Hi all,
I think it's stated in the F1000 research article that singscore only cares about ranking within samples so TPM/RPKM/FPKM is good enough and TMM normalisation is not essential. I am just wondering if I have interpreted it correctly. If so, can I just use TPM from salmon and calculated by tximport for singscore? Of course I will need to filter the low count genes. I am just wondering if that's the correct interpretation. So something like:
# Read files
txi <- tximport(files, type = "salmon", tx2gene = tx2gene)
# Get TPM
tpm <- txi$abundance
tpm <- tpm[rowSums(tpm) > 2, ] # Filter genes
tpm_ranked = rankGenes(tpm)
And then continue with the workflow?
Many thanks!
As a comment, TMM does not apply here because the authors say in the article that they recommend gene length bias to be removed, and TMM does not do that. What you could do is to use tximport output, run it through the usual
calcNormFactorsfrom edgeR and then use it's rpkm function to get the values you need. Alternatively, DESeq2 has a fpkm function. That way you could be very consistent if you use any of these packages nor differential analysis downstream.