This PR does some major restructuring around the resources used to prepare for breakpoint search, particularly in discarding the workaround method to calculate expected values and adding some flexibility to add constraint-related annotations to synonymous and nonsense variants.
Major changes:
Change expected value calculation from workaround method using mu_snp and total_exp to the actual formula as defined by Konrad in LoF constraint analyses, and remove code thus rendered obsolete
Restructure filtered_context to contain missense, nonsense, and synonymous variants and all constraint annotations including observed and expected
This necessitates changes to process_context_ht and process_vep
Command-line arguments can be used to determine whether constraint_prep contains missense, nonsense, and/or synonymous variants
Perform break search on all transcripts, meaning constraint outliers are not filtered out of filtered_context or constraint_prep
Add filtering downstream of RMC to remove constraint outliers from steps like MPC generation
Move Table prep code out of regional_constraint.py pipeline script and into utils
Change break search-associated util/annotation names and structures to be more intuitive
Minor changes:
Update RMC freeze version to 7
Revise parser structure in regional_constraint.py to use subparsers for the different steps
Update flagship LoF bucket path
Remove unused resources and constants
Add new constants for nonsense and synonymous VEP labels
Save coverage and plateau models used to calculate expected values to HailExpression
Remove exome_coverage field and retain coverage field for constraint annotations
Remove gene from groupings used to calculate expected values (transcript should be sufficient)
Remove unused lines in add_obs_annotation
Move section annotation creation into constraint prep HT
Change name of function + function parameters used to calculate chisq of section OE being different from 1 to be more intuitive
This PR does some major restructuring around the resources used to prepare for breakpoint search, particularly in discarding the workaround method to calculate expected values and adding some flexibility to add constraint-related annotations to synonymous and nonsense variants.
Major changes:
mu_snpandtotal_expto the actual formula as defined by Konrad in LoF constraint analyses, and remove code thus rendered obsoletefiltered_contextto contain missense, nonsense, and synonymous variants and all constraint annotations includingobservedandexpectedprocess_context_htandprocess_vepconstraint_prepcontains missense, nonsense, and/or synonymous variantsfiltered_contextorconstraint_prepregional_constraint.pypipeline script and into utilsMinor changes:
7regional_constraint.pyto use subparsers for the different stepsexome_coveragefield and retaincoveragefield for constraint annotationsgenefrom groupings used to calculate expected values (transcript should be sufficient)add_obs_annotationsectionannotation creation into constraint prep HT