An Efficient Improved PERM Method For Protein Folding Problem

Protein structure prediction problem is to predict the dimensional folding configurations from its amino acid sequence. Being a kind of important biological macromolecules, proteins undertake the majority of life activities. Only when their amino acid sequences fold correctly into certain 3-dimensional configurations, can proteins function properly. The ultimate purpose of studying proteins is to modify natural proteins or to design artificial proteins in order to meet the requirement of humans, the precondition of which is to solve the protein structure prediction problem.Initially,the primary measure used to predict the structures of proteins is by experiments. However, this kind of method is largely confined because of the difficulty in getting the protein crystals. Researches have shown that proteins dimensional structure is decided by their anmino acid sequences, which makes it feasible to get the proteins structure by theoretical computation. Many algorithms have been proposed for HP lattice folding problem, such as the genetic algorithm, simulated annealing, SISPER and PERM. All these methods make some progress, but still need to be improved.PERM is a kind of chain growth algorithm. By formulating some certain judgment criteria, it enriches those promising branches and prunes those ones with poor qualities. PERM is analyzed by quasi-physical and quasi-sociological methods and then two strategies are provided: population control strategy and life prediction strategy. By learning the strategies of population control, Some new definitions are presented, such as weight and thresholds. By analyzing the rule of life prediction, the improved PERM tend to decrease the thresholds in the prophase of polymer growth, and to increase the thresholds in the anaphase. Computational results indicate that the improved PERM performs efficiently with benchmarks of protein folding problem.