klmr / codons Goto Github PK

Refactor differential expression calculation to unify code

• Rewrite scripts/write-de-results to scripts/differential-expression, to be the equivalent of scripts/go-enrichment.
• Rewrite Makefile rules to make te depend on de and generate necessary results for all.
• Rewrite supplements.make rules to use the new script instead.
• Rewrite codon-anticodon-correlation.rmd.brew to use calculated DE results instead of recalculating them, equivalent to its use of gsa results.
• Decide where to calculate and save top 200 upregulated genes.
• See whether supplements.make can be made redundant.

File paths to data files should be specified via module_file to ensure that they work even when the scripts are run from a directory that is not codons. This can be trivially solved by using module_file.

Change organisation such that each script creates a single output that constitutes a target in the Makefile. Then the flow will be (partially; other parts not affected);

• Rewrite differential-expression script to create single output
• Implement overexpressed-genes to take this as input, and create single output
• Rewrite go-enrichment to create single output (and rename to reflect GSA …)
• Rewrite codon-anticodon-correlation such that
  • It is a script rather than a notebook
  • It reads the inputs generated above, and creates a single output
  • It can be configured to work with tAI or wobble-correlation instead
• Optionally create a new notebook taking this as input and generating progressive plots
• Rewrite Makefile to implement the dependencies listed above

Calculating the codon–anticodon comparison for different conditions (in particular the “mismatching” set) seems to have exponential memory requirement and runtime, and makes the code hard to run (fails on laptop).

This can probably be refactored “easily” by running the calculation for different sets sequentially in a loop rather than having them in the same data.frame simultaneously, and using dplyr grouped operations.