Performance of Canonical Correlation Forest in Phosphorylation Site Predictions

Protein phosphorylation is among the most widely used regulatory mechanisms in eukaryotes. In recent years, several phosphorylation site prediction tools have been developed to identify phosphorylation sites in silico. However, there are still ways to improve the performance of these methods. Here, we report the development of a new predictor, termed Canonical Correlation Forest-based Phosphosite (CCF-Phos) predictor, to predict putative phosphorylation sites on a given protein. The CCF-Phos was evaluated using both 10-fold cross-validation and an independent dataset. During these analyses, CCF-Phos compared favorably to other popular mammalian phosphosite prediction methods.