Prediction of HIV-1 protease cleavage site using a combination of sequence, structural, and physicochemical features

Table 5 Predictive performance of hybrid features for the four benchmark datasets

Features	DT		LR		ANN
Features	Acc.(%)	AUC	Acc.(%)	AUC	Acc.(%)	AUC
746 Dataset
Seq + Str	78.3	0.788	91.8	0.982	94.5	0.994 ^*
Seq + Phy	83.7	0.838	86.4	0.968	94.5	0.976
Phy + Str	83.7	0.810	78.3	0.860	91.8	0.982
Seq + Str + Phy	75.6	0.841	97.2	0.991	97.2	0.988
1625 Dataset
Seq + Str	89.0	0.910	96.3	0.980	95.1	0.992
Seq + Phy	89.0	0.785	97.5	0.958	98.7	0.990
Phy + Str	91.4	0.940	86.5	0.810	93.9	0.985
Seq + Str + Phy	91.4	0.956	95.1	0.980	97.5	0.990
Schilling Dataset
Seq + Str	90.8	0.845	86.5	0.865	92.0	0.873
Seq + Phy	87.1	0.500	90.8	0.837	88.9	0.825
Phy + Str	85.1	0.500	80.8	0.603	80.8	0.596
Seq + Str + Phy	88.9	0.810	89.5	0.826	91.4	0.895
Impens Dataset
Seq + Str	89.3	0.682	89.3	0.918	93.6	0.918
Seq + Phy	91.4	0.839	87.2	0.889	91.4	0.896
Phy + Str	85.1	0.500	82.9	0.889	93.6	0.932
Seq + Str + Phy	87.2	0.675	87.2	0.889	89.3	0.850

ISSN: 1471-2105