Protein Structure Prediction with Large Neighborhood Constraint Programming Search

Protein structure predictions is regarded as a highly challenging problem both for the biology and for the computational communities. Many approaches have been developed in the recent years, moving to increasingly complex lattice models or even off-lattice models. This paper presents a Large Neighborhood Search (LNS) to find the native state for the Hydrophobic-Polar (HP) model on the Face Centered Cubic (FCC) lattice or, in other words, a self-avoiding walk on the FCC lattice having a maximum number of H-H contacts. The algorithm starts with a tabu-search algorithm, whose solution is then improved by a combination of constraint programming and LNS. This hybrid algorithm improves earlier approaches in the literature over several well-known instances and demonstrates the potential of constraint-programming approaches for ab initio methods.