The programming and simulation of a lattice model of protein folding begun last year was completed by the ICOT collaborators. The solvent environment of protein was represented by short-range interactions around the amino acid sidechains. Temperature modulated simulated annealing was used in an attempt to drive the protein folding. The short range character of the forces employed prevented the simulation from succeeding. In trying to understand the role of the solvent in folding, we began to think about various longer range forms of organization which the water could take. The polarizability of the amide bond lead to considering a chain of fluctuating hydrogen bonds between the waters linking amide nitrogen to the carbonyl oxygen. Investigating the structure of solvated unfolded protein sequences and desolvated folded helices of typical proteins lead to the realization that the topological number of the unfolded and folded states was the same. By considering the number of free-electrons in each of the 20 amino acids we were able to formulate a connectivity based model of protein structure in which all the transformations involved in folding conserve the topological number. This representation combines the hydrophobic and hydrophilic character of each amino acid in consistent fashion. The typical transformation event which leads to protein folding is analogous to what is known in physics as a Feynman diagram. The exchange of two hydrophilic water loops is conditioned by the proximity of a non-polarized hydrophobic water loop. The Feynman exchanges occur throughout the protein sequence after the logical equivalent of ribosomal synthesis. Each type of amino acid along the protein sequence adds specificity to the pattern of water loop exchanges. A new program has been written at NIH to implement this topological model of protein folding. Secondary structure in the form of helices and beta sheet strands has been achieved and the packing of these objects is now under way. The computer capacity required for this folding method is quite modest, an early version of the program forms the secondary structure for TIM (Triose Phosphate Isomerase), a typical 248 amino acid protein in about 6 hours on a personal computer. Complete packing and three-dimensional optimization should take perhaps 2-3 times more computational power.
