论文部分内容阅读
脂肪组织中,激素敏感脂肪酶(HSL)被认为是调节脂肪酸代谢的关键限速酶.HSL在糖尿病的发病过程中起重要作用,抑制HSL活性有助于糖尿病的治疗,因此探索新颖的HSL抑制剂成为当前研究的热门.在激素敏感脂肪酶的作用机制和三维结构缺乏的情况下,需要发展预测HSL抑制剂的方法.本文采用几种机器学习方法(支持向量机(SVM)、k-最近相邻法(k-NN)和C4.5决策树(C4.5DT))对已知的HSL抑制剂与非抑制剂建立分类预测模型.252个结构多样性化合物(123个HSL抑制剂与129个HSL非抑制剂)被用于测试分类预测系统,并用递归变量消除法选择与HSL抑制剂相关的性质描述符以提高预测精度.本研究对独立验证集的总预测精度为75.0%-80.0%,HSL抑制剂的预测精度为85.7%-90.5%,非HSL抑制剂的预测精度为63.2%-68.4%.支持向量机方法给出最好的总预测精度(80.0%).本研究表明支持向量机等机器学习方法可以有效预测未知数据集中潜在的HSL抑制剂,并有助于发现与其相关的分子描述符.
In adipose tissue, hormone-sensitive lipase (HSL) is considered as a key rate-limiting enzyme in the regulation of fatty acid metabolism.HSL plays an important role in the pathogenesis of diabetes.It inhibits HSL activity and contributes to the treatment of diabetes.Therefore, we explored novel HSL inhibition Agents become the hot topic in the current research.In the context of the mechanism of action and lack of three-dimensional structure of hormone-sensitive lipases, there is a need to develop methods for predicting HSL inhibitors.In this paper, several machine learning methods (SVM), k- (K-NN) and C4.5 decision tree (C4.5DT)) to classify known prediction models of HSL inhibitors and non-inhibitors.Among the 252 structurally diverse compounds (123 HSL inhibitors and 129 One HSL non-inhibitor) was used to test the classification prediction system and the recursive variable elimination method was used to select the property descriptors related to HSL inhibitors to improve the prediction accuracy.In this study, the overall prediction accuracy of the independent validation set was 75.0% -80.0% , The prediction accuracy of HSL inhibitor was 85.7% -90.5%, and that of non-HSL inhibitor was 63.2% -68.4% .SVM method gave the best overall prediction accuracy (80.0%). This study shows that the support vector Machine learning methods can have Predict potential HSL inhibitors in unknown data sets and help to find molecular descriptors associated with them.