基于异构网络的标签传播算法预测药物靶点关系*

识别新的药物靶点作用关系是当前药物研究的关键，在网络标签传播算法的基础上，提出了一种融合异构网络信息的药物靶点预测策略。首先计算药物相似性和靶点相似性，并结合已知的药物靶点作用关系构建异构网络。然后充分融合药物化合物和靶蛋白信息，分别在药物相似性和靶点相似性同构网络中轮流执行改进的标签传播算法，在传播过程中通过异构网络接收另一个同构网络的信息。最后通过在四个经典数据集上测试，并与网络方法BLM-NII，NRWRH算法进行比较，结果表明用本文策略可获得较高的ROC和PR曲线面积，具有较高的预测精度。

Prediction of Drug-Target Interaction by Heterogeneous Network-based Label Propagation Algorithm

Identification of potential drug-target interactions is critical for drug research. Based on the label propagation algorithm, this paper proposed a new strategy which integrated the information of heterogeneous network for identification of drug-target interaction. Firstly, it calculated the similarity between all drugs and similarity between all targets, it integrates three different networks into a heterogeneous network by the known drug-target interactions. Then implements the modified label propagation on drug-drug similarity and target-target similarity network alternately with mutual interaction information derived from the heterogeneous network. Finally, extensively testing was applied to four datasets. Through comparison with other recently proposed BLM-NII and NRWRH, our algorithm attains high ROC and PR curve area and is capable of predicting more reliable DTIs.