Remove redundant salmon TPM output

[j-andrews7]

Description of the bug

Currently, the pipeline generates multiple TPM output from salmon, e.g. salmon.merged.gene_tpm.tsv, salmon.merged.gene_tpm_length_scaled.tsv. These are redundant, as the TPM values don't change. These files are output on both a per-sample and merged basis, so it creates a fair amount of clutter.

Expected behaviour

A single TPM file should be produced instead.

[grst]

Actually, they should not be the same...

• TPM is scaled by effective gene length per sample
• TPM length-scaled is scaled by average effective gene length across all samples

(see tximport Vignette)

So yes, on a per-sample basis, they are redundant. The merged files should be (slightly?) different.

[j-andrews7]

No, TPMs are TPMs. The counts derived from them will change with those different parameters, the TPM values themselves do not. The salmon.merged.gene_counts.tsv, salmon.merged.gene_counts_length_scaled.tsv, etc files are still necessary and will change slightly as you describe.

[grst]

Are you sure? That's not how I understand this section...

We could alternatively generate counts from abundances, using the argument countsFromAbundance, scaled to library size, "scaledTPM", or additionally scaled using the average transcript length, averaged over samples and to library size, "lengthScaledTPM"

[j-andrews7]

Yes. That section applies to the counts that are derived from the abundance (TPMs). The abundance values themselves do not change. You can run with countsFromAbundance set to any option - the abundance matrix will be the same for them all.

This is also confirmed by Mike Love here.

[grst]

ah, now I got it! Thanks for clearing that up!

[drpatelh]

FIxed in #598