pirl-unc / hitlist / 25326535363
80%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 28
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 71.489% (+0.2%) from 71.336%
25326535363

push

github

web-flow 
v1.30.33: Coarse condition_category for sample-level training (#222)

Curators write the perturbation as free text. The same biological
intervention shows up under multiple phrasings:

  ERAP1 deletion
  ERAP1 CRISPR/Cas9 knockout
  ERAP1 shRNA knockdown
  ERAP1 pharmacological inhibition
  ERAP1 variant comparison
  ERAP1+ERAP2 double CRISPR/Cas9 knockout
  ERAP1/ERAP2 variant combinations

For training a condition-aware model these all want one bucket,
not seven. Added hitlist/condition_categories.py:

  categorize_condition("ERAP1 deletion") == "ERAP1_perturbation"
  categorize_condition("ERAP1 shRNA knockdown") == "ERAP1_perturbation"
  ...

Two-pass categorization:

1. APM gene flags (already classified by hitlist.apm) promoted to
   APM-level categories. When a perturbation matches multiple APM
   genes, an explicit precedence list resolves ties — B2M loss
   wins (collapses MHC-I regardless of co-treatment), then TAP
   family, then tapasin, PLC chaperones, ERAP, IRF/SPPL3,
   proteasome, class-II APM, then cytokines/TLR last so
   "B2M KO + IFN-gamma" buckets as MHC-I_loss_B2M.

2. Regex patterns for non-APM contexts: bacterial/parasite
   infection (priority before generic "infection" so Listeria
   doesn't fall under viral), AAV/retroviral/lentiviral
   transduction (BEFORE viral infection so AAV-transduced
   allogeneic MHC-I doesn't bucket as infection_viral), specific
   named viruses, drug exposure, transfection, transplant, cell
   activation, biomaterial contact, metabolic stress, labeling
   controls.

The full categorical set (24 categories observed in current corpus):

  unperturbed                       1,159,110
  virus_inactivated_control            79,102
  IFN_gamma_treatment                  78,105
  ERAP1_perturbation                   69,875
  CIITA_transduction                   49,832
  HLA-DM_perturbation                  36,553
  cell_activation                      36,141
  drug_exposure                        31,795
  ... (continued)

3884 of 5433 relevant lines covered (71.49%)

0.71 hits per line

Coverage Regressions

Lines Coverage ∆ File
107
81.66
 0.05% export.py
Jobs
ID Job ID Ran Files Coverage
1 25326535363.1 28
71.49
 GitHub Action Run
Source Files on build 25326535363