feat: Calculate average & delta PAUs for provided contrasts

[SamBryce-Smith]

Alongside using DEXSeq to perform differential usage analysis, it would be good to calculate group-mean/median PAU expressions and usages along with a delta PAU for each contrast, as this relative expression metric is more interpretable than DEXSeq's log2FC in PAS usage.

Another nice to have would be to calculate LABRAT's psi metric for each gene/last exon, as I'm using this fairly often in my analyses and it provides a decent alternative method to summarise general changes in PAS usage for a region.

I'd suggest the following approach...

Inputs:

• qapa quant results table (output of rule qapa_quant)
• contrast name, base & contrast keys (so have groups of interest)
• sample table (so can extract samples of interest)

Snakemake:

• similar set-up to rule dexseq_apa - run for each provided contrast

Outputs:

• TSV of PAS metadata with average expressions and delta expressions. Mean and median TPMs for each condition. Mean and median PAUs for each condition. delta PAUs (contrast vs base condition) using mean and median PAUs.

  • Column names should be 'general' e.g. mean_pau_contrast, mean_pau_base
  • Additional columns of 'contrast_name', 'base_key', 'contrast_key' should probably also be added (contrast name most important)

• (Bonus) - calculate per-sample, condition-wise and overall mean expression for genes/last exons, outputting as a separate TSV. Calculated by simply summing the expression (TPM) of polyA sites of a gene. perhaps needs to be a separate script (only needs to be done once really...)