PhosphoRice: a meta-predictor of rice-specific phosphorylation sites

Table 3 The parameters in the weighted voting meta-predictors selected by a restricted grid search and a conditional random search

Element Predictor	Parameter selected by Restricted Grid search	Random number*	Parameter selected by conditional random search
Predphospho	0	Random (1)	0
NetPhos2.0	1	Random (3)	1.23
NetPhosK_0.5	0	Random (1)	0
NetPhosK_0.7	0	Random (1)	0
KinasePhos_default	3	1+Random (4)	2.75
KinasePhos_90	1	Random (3)	2.76
KinasePhos_95	0	Random (1)	0.79
KinasePhos_100	0	Random (1)	0
DISPHOS_default	3	1+Random (4)	4.25
DISPHOS_ Eukaryotes	1	Random (3)	1.65
DISPHOS_Arabidopsis	1	Random (3)	2.22
KinasePhos2.0_80	0	Random (1)	0.71
Scansite_middle	1	Random (3)	1.6
Scansite_low	3	1+Random (4)	3.9
Scansite_high	1	Random (3)	2.57
T value	8		13.3
ACC (%)	73.5		73.8
MCC	0.469 (2.60E-06)**		0.474 (6.00E-07) **

* Random (3) means the weight could fluctuate from 0 to 3. For instance, by restricted grid search, the weight value of NetphoK 2.0 was 1, and the last grid value and next grid value were 0 and 3, respectively. In a conditional random search, the weight of Netphos 2.0 was set as random (3). The weight value of KinasePhos_default was 3, and the last grid value and next grid value were 1 and 5, respectively. Therefore, its weight was set as '1+random (4)' in a conditional random search.
** P-values in Fisher's Z-transformation test (compared with the MCC of the best element predictor) are shown in parentheses.

ISSN: 1746-4811