TOWARDS A SYSTEMATIC CLASSIFICATION OF PROTEIN FOLDS

A lattice model Hamiltonian is suggested for protein structures that c an explain the division into structural fold classes during the foldin g process. Proteins are described by chains of secondary structure ele ments, With the hinges in between being the important degrees of freed om. The protein structures are given a unique name, which simultaneous ly represent a linear string of physical coupling constants describing hinge spin interactions. We have defined a metric and a precise dista nce measure between the fold classes. An automated procedure is constr ucted in which any protein structure in the usual protein data base co ordinate format can be transformed into the proposed chain representat ion. Taking into account hydrophobic forces we have found a mechanism for the formation of domains with a unique fold containing predicted m agic numbers {4,6,9,12,16,18,...} of secondary structures and multiple s of these domains. It is shown that the same magic numbers are robust and occur as well for packing on other nonclosed packed lattices. We have performed a statistical analysis of available protein structures and found agreement with the predicted preferred abundances of protein s with a predicted magic number of secondary structures. Thermodynamic arguments for the increased abundance and a phase diagram for the fol ding scenario are given. This includes an intermediate high symmetry p hase, the parent structures, between the molten globule and the native states. We have made an exhaustive enumeration of dense lattice anima ls on a cubic lattice for acceptance number Z=4 and Z=5 up to 36 verti ces.