基于深度学习的抗HIV活性QSAR预测

发布时间：2018-01-12 20:25

本文关键词：基于深度学习的抗HIV活性QSAR预测　出处：《计算机工程与设计》2017年01期 　论文类型：期刊论文

【摘要】：为提高抗HIV活性预测的精度,采用深度学习算法,提出一种基于深度信念网络的抗HIV活性预测方法。利用BP神经网络任意精度逼近非线性函数的优点,结合多个受限玻尔兹曼机(restricted Boltzmann machine,RBM)进行非监督逐层贪婪模式训练学习,建立深度信念网络算法模型(deep belief network,DBN)。将抗HIV活性的高温超导源数据(HTS raw data)、高温超导抑制剂(HTS%inhibition(20μM))、最大测定信号释放比率(mean max signal)等特征作为DBN模型的输入,抗HIV平均活性值作为该模型的输出,设计实验对模型进行训练及验证,实验结果表明,DBN模型对抗HIV活性的预测均方根误差小,预测精度高,平均预测精度为93.82%,适用于抗HIV活性评估。
[Abstract]:In order to improve the accuracy of anti-#en0# activity prediction, the depth learning algorithm is used. In this paper, a prediction method of anti HIV activity based on deep belief network is proposed, which utilizes the advantage of BP neural network to approximate the nonlinear function with arbitrary precision. Combined with multiple restricted Boltzmann machines (RBMs), unsupervised greedy training is carried out. An algorithm model of deep belief network is established. The HTS raw data of anti HIV activity are analyzed. High temperature superconducting inhibitor (HTS inhibition 20 渭 m). The maximum signal release ratio (mean max signal) was used as the input of the DBN model, and the average activity value of anti-#en2# was taken as the output of the model. Experiments were designed to train and verify the model. The experimental results show that the root-mean-square error (RMS) of DDBN model against HIV activity is small, and the prediction accuracy is high, with an average prediction accuracy of 93.82%. It is suitable for the evaluation of anti-HIV activity.
【作者单位】：新疆大学软件学院;新疆大学网络中心;
【基金】：国家自然科学基金项目(31160341) 新疆研究生科研创新基金项目(XJGRI2015034)
【分类号】：TP18
【正文快照】： 0引言对抗HIV(human immunodeficiency virus,HIV)活性定量构效关系(quantitative structure activity relation-ship,QSAR)进行活性预测,可以提高抗艾滋病药物发现的效率并降低研发成本,对促进艾滋病药物的设计有重要的意义。抗HIV活性与结构关系是一种复杂的多元、非线性关

【参考文献】