Sequence design in lattice models by graph theoretical methods

A general strategy has been developed based on graph theoretical methods, f or finding amino acid sequences that take up a desired conformation as the native state. This problem of inverse design has been addressed by assignin g topological indices for the monomer sites (vertices) of the polymer on a 3x3x3 cubic lattice. This is a simple design strategy, which takes into acc ount only the topology of the target protein and identifies the best sequen ce for a given composition. The procedure allows the design of a good seque nce for a target native state by assigning weights for the vertices on a la ttice site in a given conformation. It is seen across a variety of conforma tions that the predicted sequences perform well both in sequence and in con formation space, in identifying the target conformation as native state for a fixed composition of amino acids. Although the method is tested in the f ramework of the HP model [K. F. Lau and K. A. Dill, Macromolecules 22, 3986 (1989)] it can be used in any context if proper potential functions are av ailable, since the procedure derives unique weights for all the sites (vert ices, nodes) of the polymer chain of a chosen conformation (graph). (C) 200 1 American Institute of Physics.