量子粒子群优化智能算法的高精度分子对接方法研究

A high-accuracy molecular docking method with Intelligent Algorithm

分子对接是药物发现与设计的重要方法,采用计算机优化和模式识别方法在三维结构数据库中搜索几何、化学特性与特定药物结合位点相匹配分子的计算机辅助药物筛选是当前分子对接的研究热点,这种问题可以归为参数优化问题。本文提出了一种基于改进的量子粒子群(quantum-behaved particle swarm optimization,QPSO)算法的分子对接方法,用于处理大自由度的分子对接计算,并与基于标准QPSO算法和经典拉马克遗传算法的分子对接方法进行了比较,实验结果表明新方法无论是在对接能量还是对接准确性上,明显优于其它2种方法,尤其是在配体复杂性不断增加的情况下,非常适用于高柔性分子对接问题。

The molecular docking is of great importance method for rational drug design. The computer mode identification and optimization technique are used to search the special molecule in three dimensional small molecular database, the special molecule matches with the active site of the specially designed target relating disease in geometry, physics and chemical characteristics, in this way, the computer screening can be used to drug design. Molecular docking can be formulated as a parameter optimization problem. We present a QLDock based on improved quantum-behaved particle swarm optimization （QPSO） for flexible molecular docking which allows to handle a large number of degrees of freedom. Our hybrid method combines a QPSO with a Solis and Wets method for local optimization of molecular complexes. We compared the performance of our new hybrid method to basic QPSO and lamarckian genetic algorithm （LGA） on 17 protein/ligand complex structures benchmarking test set. The result shows that the novel approach is clearly superior to the other two. The new algorithm features a lower energy and gives substantially better results, especially with increasing complexity of the ligands. Thus, it may be used to dock ligands with many rotatable bonds with high efficiency.