基于随机森林算法的川芎成分-靶点-疾病网络的预测研究

Study on prediction of compound-target-disease network of Chuanxiong Rhizoma based on random forest algorithm

该研究搜集KEGG数据库中小分子药物及其酶、离子通道、G蛋白、核蛋白等4类药靶数据作为训练集,建立基于随机森林法的药物-靶点相互作用模型;采用十折交叉验证评价模型精度,获得4类药靶模型的预测正确率分别为71.34%,67.08%,73.17%,67.83%。利用该模型对川芎26个化学成分进行靶点预测,并构建其成分-靶点-疾病网络,所得结果得到了较好的文献验证。该文所建模型具有较高的预测精度,可用于发现其他中药成分的潜在作用靶点。

To collect small molecule drugs and their drug target data such as enzymes, ion channels, G-protein-coupled receptors and nuclear receptors from KEGG database as the training sets, in order to establish drug-target interaction models based on the random forest algorithm. The accuracies of the models were evaluated by the 10-fold cross-validation test, showing that the predicted success rates of the four drug target models were 71.34%, 67.08%, 73.17% and 67.83%, respectively. The models were adopted to predict the targets of 26 chemical components and establish the compound-target-disease network. The results were well verified by literatures. The models established in this paper are highly accurate, and can be used to discover potential targets in other traditional Chinese medicine ingredients.