Table 2.

Prediction performance obtained in the different cancer datasets.

	Cox_en	RSF	Deep_surv	TRCN* a	TRCN	ATRCN
BRCA	0.553 (±0.081)	0.571 (±0.075)	0.588 (±0.103)	0.596 (±0.091)	0.617 (±0.082)	0.652 (±0.078)
HNSC	0.539 (±0.071)	0.547 (±0.064)	0.565 (±0.077)	0.573 (±0.072)	0.585 (±0.056)	0.602 (±0.064)
LIHC	0.570 (±0.074)	0.582 (±0.070)	0.636 (±0.095)	0.654 (±0.088)	0.667 (±0.073)	0.696 (±0.079)
LUAD	0.552 (±0.067)	0.555 (±0.062)	0.572 (±0.083)	0.580 (±0.074)	0.590 (±0.068)	0.605 (±0.070)
STAD	0.542 (±0.054)	0.541 (±0.048)	0.560 (±0.058)	0.555 (±0.060)	0.564 (±0.055)	0.583 (±0.051)
Average	0.551	0.559	0.584	0.592	0.605	0.628

BRCA, breast invasive carcinoma; HNSC, head and neck squamous cell carcinoma; LIHC, liver hepatocellular carcinoma; LUAD, lung adenocarcinoma; STAD, stomach adenocarcinoma; Cox_en, Cox regression model with elastic net regularization; Deep_surv, deep Cox neural network without transfer learning; RSF, random survival forest; ATRCN, adaptive transfer-learning-based deep Cox neural network.

^aTRCN* selected the farthest cancer cluster from the target for pre-training.