Improving prediction of heterodimeric protein complexes using combination with pairwise kernel

BMC Bioinformatics

Table 1 Performance on the best average precision, recall, and F-measure for each combination kernel and other methods

Method	α	C ⁻	C⁺/C⁻	Precision	Recall	F-measure
Min kernel	0.7	1.0	3.5	0.658	0.671	0.664
MinMax kernel	0.7	1.0	3.5	0.618	0.718	0.664
Normalized Min kernel	0.2	1.0	3.5	0.678	0.628	0.652
Min-MLPK kernel	0.4	1.0	3.5	0.664	0.664	0.664
MinMax-MLPK kernel	0.4	1.0	4.0	0.678	0.669	0.673
Normalized Min-MLPK kernel	0.3	1.0	4.5	0.717	0.657	0.686
Min-TPPK kernel	0.8	1.0	3.5	0.539	0.713	0.614
MinMax-TPPK kernel	0.5	1.0	4.0	0.605	0.643	0.624
Normalized Min-TPPK kernel	0.5	1.0	4.0	0.605	0.643	0.624
MinMax-MLPK-TPPK kernel	0.2	1.0	4.5	0.632	0.691	0.660
loc-Min-MLPK kernel	0.0	1.0	3.5	0.667	0.506	0.527
phy-Min-MLPK kernel	0.8	1.0	3.5	0.612	0.521	0.547
phy-MinMax-MLPK kernel	0.2	1.0	3.5	0.632	0.578	0.604
Domain Composition kernel [27]	0.5	1.0	4.0	0.618	0.644	0.631
naive Bayes [26]	-			0.24	0. 44	0.31
MCL [16]	-			0.017	0. 023	0.020
MCODE [17]	-			0	0	-
RRW [24]	-			0.030	0.32	0.055
NWE [25]	-			0.035	0.33	0.063

The table lists, for each kernel combination, the average precision, recall, and F-measure are obtained in a 10-fold cross-validation experiment. The results by the naive Bayes-based method [26], MCL [16], MCODE [17], RRW [24], and NWE [25] are also shown, where the experiments for these methods were performed in [26]

ISSN: 1471-2105