The Study On Swarm Intelligent Optimization Algorithm Of RNA Secondary Structure Prediction

RNA's various functions are closely linked to their specific structures, in order to explore their function further, we should start from understanding the structure of RNA. It's not easy for experiment methods to determine RNA 3D-structures and not valid to all molecules. Therefore, combined with the knowledge of molecular structure and its functional properties, through computer simulation and calculation to"predict"the structural information, a certain valuable results can be gotten with lower costs and shorter time.The free energy minimization algorithm is the most widely used method for the RNA secondary structure prediction. Based on local stem search and with the free energy as evaluation function of it is a typical combinatorial optimization problem. As one of effective ways to solve such problems, the Swarm Intelligent (SI) optimization algorithms have already made some achievements. This thesis introduces two SI algorithms: Particle Swarm optimization (PSO) and Shuffled Flog Leaping Algorithm (SFLA).The reasonable algorithm frameworks designed to achieve predictive effects which provide some inspiration for solving such problems.The mathematical definition of RNA secondary structure is introduced with the thermodynamic model which used in predicted model.The SetPSO is analyzed in detail first. Based on SetPSO algorithm, an improved immune particle swarm optimization algorithm is designed for the RNA secondary prediction problem. An immune memory operator is used to avoid falling into local optimal and increases the diversity of particle swarm. Simulation results show that the improved algorithm can get better prediction accuracy in shorter time.An effective DPSO is redesigned. Based on the characteristics of discrete variable, particle's position, velocity and their operation rules are redefined. From the artificial immune systems, immune variation operator is designed to keep the diversity of particle swarm, and inject vaccine operator is defined to improve the algorithm's refinement ability. The simulation results show the algorithm can tackle the problem effectively.The SFLA algorithm's principles, applications and improvements are presented. Redefines the Discrete SFLA (DSFLA) to solve the combinatorial optimization problem, including individual's position and the operation rules. The simulation results show the new DSFLA algorithm can be used to tackle the RNA secondary structure prediction effectively and efficiently.Through the above three algorithms show that the SI algorithm is effective. Future research will include the prediction of the pseudoknot and the parallel calculation to improve forecast accuracy and speed.