摘要
pKa(解离常数)关系到药物分子在生物体内的吸收、代谢等过程。近年来,基于机器学习模型预测药物分子性质在药物筛选中获得广泛应用,神经网络可通过在深度与宽度两个方向上的扩展来增强模型的学习能力。以神经网络在药物分子pKa预测中的应用为例,比较了神经网络的深度与宽度对预测结果的影响。通过分析预测结果的均方差以及预测值与真实值之间的相关系数,系统地评估了模型的深度与宽度对预测性能的影响。基于定量的比较结果,提出了组合的神经网络模型计算方案。计算结果表明:深度神经网络模型在使用组合MACCS和ECFP指纹时,预测准确性超过了单一的宽度或深度神经网络。
The pKa,logarithm of the acid dissociation constant,is related to the absorption and metabolism of drug molecules.In recent years,the prediction of drug molecular properties based on machine learning model has been widely used in drug screening.Neural network can enhance the learning ability of the model by expanding in depth and width.Taking the application of neural network in pKa prediction of drug molecules as an example,the influence of the depth and width of neural network on the prediction results was compared.By analyzing the mean square error of the prediction results and the correlation coefficient between the predicted value and the real value,the influence of the depth and width of the model on the prediction performance is systematically evaluated.Based on the results of quantitative comparison,a combined neural network model is proposed.The results suggested that the prediction accuracy of the model based on the deep neural network is higher than that of the single wide or deep neural network when using the combination of MACCS and ECFP fingerprints.
作者
谢良旭
薛亮亮
李峰
XIE Liangxu;XUE Liangliang;LI Feng(Institute of Bioinformatics and Medical Engineering,Jiangsu University of Technology,Changzhou 213001,China;Jiangsu Sino-Israel Industrial Technology Research Institute,Changzhou 213001,China)
出处
《江苏理工学院学报》
2021年第2期1-8,共8页
Journal of Jiangsu University of Technology
基金
国家自然科学基金项目“构象动力学对酶催化活性影响的自由能计算”(22003020)
江苏省自然科学基金项目“SIRT1激活的分子机理研究及新型别构激动剂筛选”(BK20191032)
江苏省“双创计划”(双创博士)项目
江苏省中以产业技术研究院开放课题“多层动态描述符助力人工智能预测蛋白-配体分子结合自由能”(JSIITRI202009)
常州市重点研发项目“人工智能辅助筛选新冠病毒S蛋白与宿主ACE2蛋白结合抑制剂”(CJ20200045)。
关键词
人工智能
神经网络
深度学习
定量构效关系
药物发现
PKA
artificial intelligence
neural network
deep learning
quantitative structure-activity relationship
drug discovery
pKa
