how to calculate gene length for TPM

Question

[adi.gershon1]

Hi i have row counts from featurecounts I want to calculate tpm, and for that I want to calculate gene length correctly I'm not sure what is the right way doing it do I need Total genomic span of the gene or Sum of non-overlapping exonic regions

For example running this command on a gtf file is good? awk '$3 == "exon" { match($0, /gene_id "([^"]+)"/, gene); gene_id = gene[1]; length[gene_id] += $5 - $4 + 1 } END { for (id in length) print id, length[id] }' gencode.v34.chr_patch_hapl_scaff.annotation.gtf > gene_lengths_cleaned.tsv

Answer 1

[ATpoint]

featureCounts returns a gene length, use that. There is no naive approach that is better. Here is what it does: https://support.bioconductor.org/p/88133/#88135

The "more correct" way would be the way that tools like salmon uses. They resolve the transcript expression of each gene and then in combination with something like tximport give you the average length of all expressed transcipts. But you don't have that in featureCounts, and there is no naive method that could easily mimic that. So just use what featureCounts gives you.