Predicting residue solvent accessibility from protein sequence by considering the sequence environment