Heuristic ant colony optimization algorithm for predicting the structures of 2D HP model proteins

Predicting the structures of a protein, i.e. finding low-energy conformations of a protein is one of the most prominent problems in bioinformatics. A simplified two-dimensional (2D) hydrophobic-hydrophilic (HP) lattice model is studied. Despite the simplicity of the model, the protein structure prediction problem on the HP lattice model has been proven to be NP-hard. The ant colony optimization (ACO) algorithm is a class of global search method. By incorporating the local search method with pull move into the ACO algorithm, a heuristic ACO algorithm (HACO) is put forward for solving 2D HP protein structure prediction problem. Eight general benchmark instances are tested. The numerical results show that the HACO algorithm is as good as or outperforms the other eight methods in the literature for seven out of eight instances. For the longest sequence with length 64, the HACO algorithm achieves the suboptimal solution, which has a difference of -1 from the optimal value. Experimental results show that the proposed HACO algorithm is a powerful method to predict the protein's structure.