============================================================
WRN-MSI Pan-Cancer Atlas — Phase 1: MSI/MMR Classifier
============================================================

OVERVIEW
  Total cell lines with lineage annotation: 2119
  Cell lines with CRISPR data: 1186
  MSI-H classified: 70 (3.3%)
  MSS classified: 2049

CLASSIFICATION TIERS
  tier3_MSH2_LOF: 30
  tier2_MLH1_silenced: 22
  tier1_annotated: 12
  tier1_annotated+molecular: 6

MSI-H SUBTYPES (mechanism)
  MMR-mutated: 41
  MLH1-methylated: 16
  MLH1-methylated+MMR-mutated: 10
  annotated-only: 3

MLH1 SILENCING THRESHOLD: 1.0 (log2 TPM+1)
  Tier 1: Model.csv annotation (ModelSubtypeFeatures = MSI)
  Tier 2: MLH1 expression < threshold + no MLH1 LOF (methylation)
  Tier 3: MSH2 LOF mutation (dominant mutation-driven MSI)
  MSH6/PMS2-only LOF: recorded but not sufficient for MSI-H call

QUALIFYING CANCER TYPES (>=5 MSI-H + >=10 MSS with CRISPR)
------------------------------------------------------------
  Uterus: 8 MSI-H / 26 MSS (CRISPR), 11/44 total (25.0%)
  Bowel: 12 MSI-H / 51 MSS (CRISPR), 22/99 total (22.2%)

ALL CANCER TYPES WITH MSI-H LINES
------------------------------------------------------------
  Uterus: 11 MSI-H / 33 MSS (CRISPR: 8/26) *
  Bowel: 22 MSI-H / 77 MSS (CRISPR: 12/51) *
  Vulva/Vagina: 1 MSI-H / 4 MSS (CRISPR: 1/1)
  Ovary/Fallopian Tube: 5 MSI-H / 70 MSS (CRISPR: 4/55)
  Esophagus/Stomach: 6 MSI-H / 98 MSS (CRISPR: 4/65)
  Eye: 1 MSI-H / 23 MSS (CRISPR: 0/15)
  Cervix: 1 MSI-H / 25 MSS (CRISPR: 1/17)
  Breast: 3 MSI-H / 93 MSS (CRISPR: 2/51)
  Pleura: 1 MSI-H / 35 MSS (CRISPR: 0/21)
  Lymphoid: 7 MSI-H / 256 MSS (CRISPR: 3/90)
  Head and Neck: 2 MSI-H / 92 MSS (CRISPR: 2/75)
  Lung: 4 MSI-H / 256 MSS (CRISPR: 3/123)
  Pancreas: 1 MSI-H / 67 MSS (CRISPR: 0/48)
  Kidney: 1 MSI-H / 73 MSS (CRISPR: 0/35)
  Skin: 2 MSI-H / 148 MSS (CRISPR: 2/73)
  Myeloid: 1 MSI-H / 87 MSS (CRISPR: 1/42)
  CNS/Brain: 1 MSI-H / 126 MSS (CRISPR: 1/88)

POWER ANALYSIS NOTE
  Qualifying threshold: >=5 MSI-H + >=10 MSS with CRISPR data.
  This ensures adequate statistical power for Cohen's d effect
  size estimation with bootstrap CI in Phase 2.
