Molecular modeling based on a hybrid evolutionary optimization and an information condensation algorithm, called GHOST, of spin label ESR spectra was applied to study the structure and dynamics of membrane proteins. The new method is capable of providing detailed molecular information about the conformational space of the spin-labeled segment of the protein in a membrane system. The method is applied to spin-labeled bacteriophage M13 major coat protein, which is used as a model membrane protein. Single cysteine mutants of the coat protein were labeled with nitroxide spin labels and incorporated in 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) bilayers. The new computational method allows us to monitor distributions of local spatial constraints and molecular mobility, in addition to information about the location of the protein in a membrane. Furthermore, the results suggest that different local conformations may coexist in the membrane protein. The knowledge of different local conformations may help us to better understand the function-structure relationship of membrane proteins.