[1] All amino acid types are classified as either hydrophobic (H) or polar (P), and the folding of a protein sequence is defined as a self-avoiding walk in a 2D or 3D lattice.
The HP model imitates the hydrophobic effect by assigning a negative (favorable) weight to interactions between adjacent, non-covalently bound H residues.
[3][4] For some model variants/lattices, it is possible to compute optimal structures (with maximal number of H-H contacts) using constraint programming techniques[5][6] as e.g. implemented within the CPSP-tools webserver.
[7] Even though the HP model abstracts away many of the details of protein folding, it is still an NP-hard problem on both 2D and 3D square lattices.
[8] A Monte Carlo method, named FRESS, was developed and appears to perform well on HP models.