基于量子多种群遗传算法的蛋白质二级结构预测

为进一步提高蛋白质二级结构的预测精度,将量子计算和多种群算法融入到传统的遗传神经网络算法中。同时考虑到氨基酸残基的众多理化性质是形成蛋白质二级结构的主要驱动力,构象偏好也是影响蛋白质二级结构形成的重要因素,提出了一种新的基于理化性质和构象信息编码的量子多种群遗传算法。该方法蕴含了丰富的生物信息,可以有效减少网络系统的不确定性。用PDBselect25中的24条蛋白质进行测试,结果表明该算法可以有效的预测蛋白质的二级结构,平均预测精度达到72.10%,分别比SNN、DSC、PREDSATOR方法提高了7.80%、3.70%和3.41%。该方法采用混合编码的形式进行编码,在每个种群内部引入量子计算,形成了以多种群遗传算法来带动量子计算,量子计算反作用于多种群算法的双重优化的方法,可有效提高蛋白质二级结构预测的精确度。

Prediction of Protein Secondary Structure Based on Quantum Multiple Population Genetic Algorithm

In order to improve the prediction accuracy of protein secondary structure, quantum computation and multiple population genetic algorithms were added to the traditional genetic neural network algorithm. At the same time, the chemical and physical properties of amino acid residues was the main driving force to form protein secondary structure, and the conformational preference also greatly affected the formation of protein secondary structure, so a new coding method based on physical and chemical properties and conformation information was presented. This coding method which contains rich biological information could effectively reduce the uncertainty of the network system. This model was employed to predict 24 non homologous protein sequences in PDBSelect25. The result showed that this proposed model improved the prediction accuracy to 72.10%, increasing the prediction accuracy by 7.80%, 3.70% and 3.41% respectively compared with SNN, DSC and PREDSATOR method. In this new method, based on mixed coding and quantum computation, the multiple population genetic algorithms drove the quantum computation and the quantum computation feed back the multiple population genetic algorithms. This kind of drive and feed back mechanism could effectively improve the prediction accuracy of protein secondary structure.