Proposal: AlphaMissense Ascendant

[MattWellie]

Re-work the categories to adjust to the facts:

• Support (determined through in silico predictors) largely overlaps with Cat6 (AlphaMissense prediction)
• c6/AM is noisy for cohorts without family structure to better filter for MOI matches

Proposed changes

1. Replace the CADD/REVEL involvement in Cat. 2 with AM pathogenicity
2. Remove the CategorySupport logic. The code placeholders to trigger 'support' logic will remain, but we just won't exercise those for now.
3. Add back-filtering logic to remove variants after MOI application where they are Cat6 only, and the variant-gene has not been phenotype-matched to this participant

Future

• Turn off Category 2? This will be done in config for now, but after evaluation we may drop this category. The first seen/latest report takes care of what we wanted to accomplish with this logic, and is less volatile (results coming & going for each run)

[MattWellie]

As a prototype I'm going to leave the possibility of a Cat6-only compound het pair without phenotype matching. The logic to remove those will be fiddly, and might only affect a tiny number of variants