Table 4.

The performance of tumour type classification on the GDC multi-omics dataset with different omics type combinations.

Omics Type	Macro-F1	Macro-TPR	Macro-PPV	Accuracy	Macro-ROCAUC
Gene expression (a)	0.9518 ± 0.0053	0.9522 ± 0.0055	0.9558 ± 0.0069	0.9676 ± 0.0027	0.9982 ± 0.0003
DNA methylation (b)	0.9273 ± 0.0167	0.9253 ± 0.0192	0.9333 ± 0.0181	0.9650 ± 0.0040	0.9985 ± 0.0001
miRNA expression (c)	0.9274 ± 0.0140	0.9268 ± 0.0134	0.9320 ± 0.0148	0.9544 ± 0.0057	0.9983 ± 0.0004
Multi-omics (a + b)	0.9675 ± 0.0083	0.9669 ± 0.0077	0.9643 ± 0.0077	0.9753 ± 0.0040	0.9988 ± 0.0009
Multi-omics (a + b + c)	0.9683 ± 0.0020	0.9684 ± 0.0026	0.9705 ± 0.0047	0.9771 ± 0.0027	0.9991 ± 0.0002

The best results were marked in bold.