【作者】 吴景恒； 梅娟； 廖思燕； 陈锦灿； 沈勇；

【Author】 Jingheng Wu,Juan Mei,Siyan Liao,Jincan Chen,Yong Shen (School of Chemistry and Chemical Engineering of Sun Yat-sen University,Guanzhou 510275, P.R.China,)

【机构】 中山大学化学与化学工程学院；

【摘要】 利用遗传算法(GA)实现前向人工神经网络(BP-ANN)网络初始权重的优化及药物定量构效关系(QSAR)的分子描述子的选择。通过引入留一法交叉验证(LOO-CV)作为遗传算法衡量建立ANN模型有效性的方法实现选择较优的描述子组合,通过引入新的评价函数避免ANN训练过度,建立了基于遗传算法的自适应人工神经网络模型。把该方法应用到两组近来基于逐步多元线性回归(MLR)模型的QSAR研究当中,通过自适应GA-ANN模型选择描述子建立的非线性人工神经网络QSAR模型在LOO-CV中表现较高的交叉验证系数及较低的均方根偏差(RMSD)。而进一步引入留多法交叉验证(LMO-CV),Y-随机性检验及外部验证的检验表明,建立的GA-ANN模型在稳健性和预测能力中明显优于MLR模型。研究表明自适应GA-ANN模型为改进QSAR模型提供了新的有效的方法。更多还原

【Abstract】 Based on the quantitative structure-activity relationships(QSARs) models developed by artificial neural networks(ANNs),genetic algorithm(GA) was used in the variable-selection approach with molecule descriptors,and helped to improve back-propagation(BP) training algorithm as well.The Cross Validation techniques of Leave-One-Out(LOO-CV) investigated the validity of the generated ANN model and preferable variable combinations derived in the GAs.A self-adaptive GA-ANN model(Fig.1) was successfully established by employing a new estimate function for avoiding over-fitting phenomenon in ANN training.Compared to the variables selected in two recent QSAR studies that were based on stepwise multiple linear regression(MLR) models,the variables selected in self-adaptive GA-ANN model are superior in constructing ANN model,as they revealed a higher cross-validation coefficient(Q~2) and a lower root-mean-square deviation(RMSD) both in the established model and biological activity prediction.The introduced methods for validation,including Leave-Multiple-Out(LMO), Y-randomization and external validation,proved the superiority of established GA-ANN models over MLR models in both stability and predictive power.Self-adaptive GA-ANN showed us a possibility to improve QSAR model.更多还原