In silico prediction of the structure of membrane proteins :is it feasible?