The main objective of our research project consists in deriving an ab initio method able to predict, at an acceptable level of accuracy , folding and topology of membrane proteins. This will be carried out by improving FILM (Folding in Lipid Membranes), a knowledge-based method previously developed to predict 3-D structures of small membrane proteins. With this aim, energy functions able to describe folding and packing interactions inside the membrane will be derived by the approach used in the current version of FILM. A data set composed of macromolecules with known topology and high-resolved 3-D features will be created for each class of membrane proteins. Furthermore, topology,folding and packing will be studied by statistical analysis of each set and converted in interaction energy functions by applying the inverse Boltzmann equation. The novel interaction energies will be added to the earlier force field and the improved method will be validated on proteins having high-resolved 3-D structure available at PDB. Finally, also peptides forming oligomeric channels, such as melittin and alamethicin will be considered and a proper energy functions will be developed, so that the method could be reliably applied to the whole family of membrane proteins.