Dataset	N	5-Gene C-index	4-Gene (w/o NFS1) C-index	Δ C-index	Interpretation
Training set (n=255)	255	0.655	0.651	+0.004	NFS1 contributes minimal value
Internal validation (n=110)	110	0.689	0.686	+0.004	Consistent across splits
Full TCGA-LIHC cohort	365	0.666	0.663	+0.003	ΔC < 0.01 in all datasets