Cross-lattice behavior of general ACO folding for proteins in the HP model

Abstract

The computational investigation of protein folding is one of the most relevant challenges in bioinformatics. In this field, simplified lattice models for proteins like the classical HP model have been proposed, and different lattice types can be employed. A promising approach to find ground state conformations relies on Ant Colony Optimization (ACO), a popular biology-inspired heuristics: several variants have been implemented so far, on square lattices in 2D and 3D.In this paper we propose a general scheme of ACO for HP on both square and triangular lattices in 2D and 3D, including also a novel initialization procedure for the pheromone matrix according to some pre-computed suboptimal conformations. The algorithm behavior, considering the influence of the optional parts and the required parameter tuning, is investigated for the first time with experiments that systematically span different lattice types. The test outcomes are useful in understanding how to operate on the algorithm parameters.The presented results are used to sketch out general guidelines for the practical employment of ACO in conformational studies, depending on the chosen sequences and lattice types.