Predicting anticancer hyperfoods with graph convolutional networks

Table 3 Summary of results (%) on anticancer drug prediction

Method	ACC	F1	AUPR	Precision ac	Recall ac	Precision non-ac	Recall non-ac
SVM	79.26 ± 4.2	52.12 ± 5.92	53.35 ± 10.97	41.50 ± 6.75	69.12 ± 10.08	96.31 ± 1.06	88.74 ± 3.20
RWR + SVM	81.13 ± 3.79	51.84 ± 5.79	67.43 ± 8.14	38.98 ± 5.38	75.08 ± 6.92	96.90 ± 0.83	86.67 ± 2.37
MLP	80.62 ± 3.81	66.53 ± 5.02	69.05 ± 5.01	69.75 ± 6.74	64.55 ± 8.23	96.02 ± 0.85	96.68 ± 1.30
GCN	80.52 ± 3.33	63.95 ± 3.90	66.45 ± 5.82	63.33 ± 5.72	65.51 ± 7.42	96.08 ± 0.76	95.54 ± 1.38
GraphSAGE	78.27 ± 6.11	59.93 ± 6.53	64.42 ± 9.96	61.04 ± 5.72	61.15 ± 13.48	95.62 ± 1.37	95.38 ± 1.51
ChebNet	83.46 ±2.52	67.99 ±2.87	73.91 ±3.49	65.46 ±4.53	71.27 ±5.58	96.71 ±0.59	95.65 ±0.96
MLP-P	76.72 ± 2.68	54.40 ± 3.56	59.79 ± 7.64	51.67 ± 11.33	60.73 ± 7.81	95.44 ± 0.72	92.72 ± 3.18
GCN-P	78.70 ± 5.36	57.43 ± 7.61	60.03 ± 8.48	52.77 ± 7.69	64.03 ± 11.05	95.83 ± 1.18	93.37 ± 1.72
GraphSAGE-P	77.09 ± 4.18	54.07 ± 4.88	60.55 ± 9.51	48.87 ± 4.06	61.64 ± 9.65	95.53 ± 0.96	92.55 ± 1.95
ChebNet-P	76.10 ± 2.67	55.71 ± 4.46	59.68 ± 9.53	53.72 ± 4.07	57.86 ± 4.96	95.17 ± 0.53	94.35 ± 0.44

ACC = balanced accuracy, F1 = harmonic mean of precision and recall, AUPR = area under the precision-recall curve, ac = anticancer, non-ac = non-anticancer

ISSN: 1479-7364