Prediction Of Protein Folding Based On Improved Genetic Simulated Annealing Algorithm

Prediction of protein folding is predicting 3D structure based on the amino sequence. It's one of the most important subjects of Bioinformatics. The most common way to solve the problem is adopting simplified models and using effective searching methods.HP off-lattice is one of the simplified models, it consists the hydrophobic forces as the main driven force of the protein folding, it can reflect the real protein molecular conformation. Genetic algorithm and simulated annealing algorithm are combinatorial optimization methods, and they are applied in protein folding prediction.On the basis of the disadvantages of Genetic Algorithm and Simulated Annealing Algorithm, and the HP off-lattice model consisting of massive conformations and multi-extremum and multi-parameter, a new genetic simulated annealing algorithm is presented. By improving the fitness function and select operator to increase the population's diversity; and presents a new adaptive crossover and mutation rate consisting generation and fitness, avoiding premature converge; adopting effective temperature descending manner and accepting rule to enhance the efficiency of the annealing operation.Finally verify the algorithm by using the Fibonacci and two protein sequences. The results demonstrate that the new algorithm can predict the conformations of the protein models, and reach more accurate energies, also can avoid local poles, and have a better convergent performance.