I am not aware of any tools to predict masked residues and a bit skeptical that such models would give reliable prediction in case of phosphorylations, given that latter can induce protein refolding. May be you can start from other direction and check for identified phosphorylations in databases, like phosphosite.org. Also there are many kinase site prediction tools which may help to identify at which residues phosphorylations may occur (for example GPS 5.0, PhosphoPredict or PhosphoNET.