An Improved Immune Algorithm for the Protein Structure Prediction Problem

The protein structure prediction problem is concerned with predicting the three dimensional native conformation of a protein from the corresponding one dimensional sequence of amino acids. This paper is concerned with solving the protein structure prediction problem on the simple HP lattice model of a protein. Many algorithms and search techniques exist in literature for solving the protein structure prediction problem on the HP lattice model; a very competitive state-of-art algorithm is the clonal selection algorithm inspired by learning and memory in the biological immune system. In this paper an improvement to the standard clonal selection algorithm is proposed: to replace its local search operator with a single point crossover operator from a genetic algorithm. Experiments found that the clonal selection algorithm with crossover reduced the number of function evaluations needed to find the best known solution by an order of magnitude for seven of the nine benchmark sequences tested.