Table 4(A-B).

A summary of statistical estimates on performance of single-omics models (4A), multi-omics models and the Radiomic model trained by using MKL algorithm (4B).

Table 4A	Training Set				Hold-out test set
	Avg. AUC	STD	95% CI	p-value		Avg. AUC	STD	95% CI	p-value
Single-omics Model
Radiomics (R)	0.94	0.01	(0.938,0.946)	Ref		0.92	0.03	(0.903,0.933)	Ref
Morphology (M)	0.74	0.03	(0.726,0.754)	<0.0001*		0.64	0.08	(0.608,0.677)	<0.0001*
Contouromics (C)	0.664	0.052	(0.641,0.687)	<0.0001*		0.55	0.082	(0.514,0.586)	<0.0001*
Dosiomics (D)	0.9	0.02	(0.887,0.903)	<0.0001*		0.81	0.03	(0.798,0.824)	<0.0001*
Table 4B	Training Set				Hold-out test set
	Avg. AUC	STD	95% CI	p-value		Avg. AUC	STD	95% CI	p-value
Multi-omics Model
RM	0.99	0.01	(0.983,0.989)	<0.0001*	0.47	0.93	0.04	(0.916,0.952)	0.36	0.62
RD	0.99	0	(0.990,0.994)	<0.01*	<0.001*	0.93	0.03	(0.920,0.947)	0.37	0.64
RC	0.99	0.01	(0.984,0.989)	<0.0001*	0.42	0.93	0.04	(0.909,0.941)	0.14	0.92
RMDC	1	0	(0.995,0.998)	Ref	<0.0001*	0.94	0.03	(0.931,0.956)	Ref	0.21
Radiomic Model (trained by MKL)
R_MKL	0.98	0.01	(0.981,0.988)	<0.0001*	Ref	0.93	0.05	(0.905,0.948)	0.21	Ref

The symbol (*) represents meeting the level of statistical significance (p < 0.05).