# Supplementary Table S6: Internal Validation via Bootstrap Resampling
# Method: Non-parametric bootstrap (1,000 resamples) on full TCGA-LIHC cohort (n=365)
# Metric: Harrell's concordance index (C-index)
# Seed: 42 (fixed for reproducibility)

Metric	Value
C-index – Full TCGA-LIHC (n=365)	0.67
C-index – 95% CI (bootstrap, 1000 resamples)	[0.62, 0.72]
  Lower bound (2.5th percentile)	0.62
  Upper bound (97.5th percentile)	0.72
C-index – Training subset (n=255)	0.67
C-index – Internal validation subset (n=110)	0.64
Optimal cutoff method	Median of training-set risk score distribution
Bootstrap seed	42 (fixed)
Note	Percentile-based confidence interval from non-parametric bootstrap.
     C-index values rounded to two decimal places per journal convention.