Fig 1.
A summary of SLapRLS.
Fig 2.
Firstly, label dataset are derived from Phospho.ELM, and it is split into train dataset and test dataset. Secondly, the model is developed using train dataset and its similarity matrix with SLapRLS, with which the predicted result of test dataset is achieved. Additionally, an independent test dataset is used. The model that predicts the independent dataset is developed with all the label dataset.
Fig 3.
ROC curves of different algorithms.
ROC curves of kinase Erk2, Erk1, CDC2 and PKC alpha achieved by four different algorithms are plotted. The red line, blue line, yellow line and cyan line represent SLapRLS, SVM, BDT and KNN, respectively.
Table 1.
Compared AUC values of the four algorithms: SLapRLS, SVM, BDT and KNN.
Fig 4.
Comparison of the four algorithms at high stringency level (sp = 0.99).
The Sn, Sp, Acc and Mcc values at high stringency level (sp = 0.99) of four algorithms on the S/T and Y kinases.
Fig 5.
Weblogos of S/T substrate kinases and Y substrate kinases.
A and C are the Weblogos of kinase ATM and ck2 alora2 and B and D are the Weblogos of kinase EGFR and INSR.
Table 2.
Comparison among SLapRLS, iGPS and NetworKIN on independent test data.
Table 3.
Pathway enrichment analysis of PKC alpha.