Table 3.

AUC of the baseline models and our proposed deep learning model (DCNN_Concat as shown in Table 1) for multiclass classification. The results are shown for SVM and RF models using individual CNA data (CNA_SVM, CNA_RF) and gene expression data (Gene_SVM, Gene_RF) as well as the integration of both data sources (SVM_Concat, MK_SVM_Concat and RF_Concat). We ran each RF model with the best found parameter for 10 times and reported the mean AUC because of the stochastic nature of RF. In addition, we also reported the standard deviation (SD) for the accuracies and AUCs from these 10 RF models. The results for DCNN_Concat using the selected top genes are also shown. The best results for models with different number of top genes selected by χ² for CNA data and ANOVA for gene expression data are shown in bold color.

Model (top genes)	Test		AUC based on top selected genes
			100	150	200	250	300	350	400	450	500
CNA_SVM	χ2		0.589	0.590	0.632	0.630	0.629	0.636	0.629	0.630	0.633
CNA_RF	χ2	AUC	0.644	0.647	0.633	0.642	0.649	0.655	0.651	0.655	0.669
		SD	0.003	0.002	0.001	0.004	0.003	0.004	0.002	0.006	0.003
Gene _SVM	ANOVA		0.804	0.818	0.812	0.799	0.808	0.807	0.814	0.814	0.819
Gene _RF	ANOVA	AUC	0.805	0.810	0.805	0.802	0.809	0.802	0.803	0.79	0.801
		SD	0.003	0.002	0.001	0.003	0.002	0.003	0.002	0.003	0.002
SVM_Concat			0.810	0.815	0.810	0.818	0.810	0.810	0.815	0.814	0.814
MK_SVM_Concat			0.741	0.760	0.753	0.725	0.734	0.820	0.730	0.746	0.792
RF_Concat		AUC	0.803	0.808	0.803	0.802	0.801	0.804	0.801	0.807	0.808
		SD	0.001	0.002	0.001	0.001	0.001	0.001	0.002	0.002	0.001
DCNN_Concat			0.810	0.817	0.815	0.817	0.821	0.811	0.829	0.834	0.852