目的 :考察改进的遗传算法 (GA)在QSAR中的变量优选能力。方法 :用自编程的改进的GA选择变量 ,PLS方法拟合模型 ,对 7,9,10 ,11取代的喜树碱衍生物和取代苯环双三氮唑醇类抗真菌化合物两个数据组进行QSAR研究。结果 :对两个数据组分别得到了一批较常规QSAR分析结果更优的方程。结论 :GA作为一种启发式搜索算法以其高效快速、并行性的特点尤其适合于QSAR研究 ,但因其仅以统计指标为进化依据 ,故需对其实施必要的预处理及结果检查 Objective: To investigate the variable optimization ability of improved genetic algorithm (GA) in QSAR. METHODS: Two data sets of the camptothecin derivatives substituted with 7, 9, 10, 11 and the substituted benzene ring bisotriptyline alcohol antifungal compounds were prepared by using the modified GA selection variables and the PLS method. Conduct QSAR studies. Results: A number of better equations than the conventional QSAR results were obtained for both data sets. Conclusion: GA, as a heuristic search algorithm, is especially suitable for QSAR research because of its high efficiency, rapidity and parallelism. However, because it only evolves based on statistical indicators, GA needs to implement the necessary preprocessing and result checking