Prediction performance of different feature groups by dividing phosphorylation sites.

The prediction accuracy was simply the percentage of correct predictions among both positive and negative samples. Here “all” means sequence, structure plus the combination of all functional features; “sequence” means sequence and structure features only; “minus” equals to “all” minus “sequence”; “kegg” means sequence, structure plus significant KEGG features; “bp” means sequence, structure plus significant GO Biological Process features; “cc” means sequence, structure plus significant GO Cellular Component features; “mf” means sequence, structure plus significant GO Molecular Function features; “ipr” means sequence, structure plus significant InterPro domain features; “pfam” means sequence, structure plus significant Pfam domain features; “string” means sequence, structure plus significant STRING PPI features.