In-silico target prediction by ensemble chemogenomic model based on multi-scale information of chemical structures and protein sequences

Table 5 Comparison results with alternative state-of-the-art prediction methods

TopK	^aHitPickV2	^aPPB2							^aPPB	^aTargetNet	^aOur model	^bOur model	^bSwisstarget prediction
TopK	^aHitPickV2	NB(ECfp4)	NN(ECfp4) + NB(ECfp4)	NN(ECfp4)	NN(MQN) + NB(ECfp4)	NN(MQN)	NN(Xfp) + NB(ECfp4)	NN(Xfp)	^aPPB	^aTargetNet	^aOur model	^bOur model	^bSwisstarget prediction
1	24.69	16.49	14.89	16.59	21.87	10.65	21.49	16.49	5.18	23.20	26.96	57.00	28.00
3	56.74	35.06	52.31	52.88	52.40	22.43	52.40	30.91	18.85	41.85	56.36	–	–
5	58.43	47.03	60.92	57.96	57.21	26.96	60.30	35.34	25.82	46.37	59.33	–	–
7	60.82	53.35	62.76	61.29	60.04	30.16	61.30	39.21	29.78	48.91	60.89	–	–
10	62.20	60.98	64.75	63.62	63.05	34.68	62.58	45.62	34.40	50.99	63.99	–	–
15	–	–	–	–	–	–	–	–	–	–	–	76.00	72.00

^aRecall rate defined in our article (%); ^b the fraction of compounds for which at least one known target was identified in the top-1 or top-15 of the prediction lists (%)

ISSN: 1758-2946