Tsinghua University News

Scientists from interdisciplinary fields published a new machine learning algorithm for predicting drug-target interactions in Nature Communications

Scientists from interdisciplinary fields published a new machine learning algorithm for predicting drug-target interactions in Nature Communications


A collaborative team from interdisciplinary fields, including Prof. Jianyang Zeng in the Institute for Interdisciplinary Information Sciences (IIIS), Prof. Ligong Chen’s Lab in the School of Pharmaceutical Science at Tsinghua University, and Prof. Jian Peng, in the Department of Computer Science at the University of Illinois at Urbana-Champaign, recently developed a novel machine learning algorithm for predicting drug-target interactions. The paper of this work has been published in the journal Nature Communications.

Prediction of drug-target interaction is a crucial step of drug discovery and drug repositioning. Large-scale genomic, chemical and pharmacological data provide new opportunities for drug-target interaction prediction. However, systematic integration of these heterogeneous datasets remains a challenge. The proposed algorithm, called DTINet, employs a new machine learning approach to integrate these heterogeneous networks. The algorithm represents each drug and gene by compact patterns to remove the background noise of biological data and reveal the topological properties of drugs and genes, which is essential for improving the prediction performance. DTINet was found to outperform several state-of-the-art prediction methods. Furthermore, most of the novel drug-target interactions predicted by DTINet can be supported by known evidence in the literature. In addition, experimental assays had been performed to validate those predicted drug-target interactions that were rarely reported in previous studies, including those interactions between three drugs and two proteins (PTGS1 and PTGS2). The validation results demonstrated new potential applications of these drugs in preventing inflammatory diseases, which provides new insights into drug repositioning.

Nature Communications, with an impact factor 12.124 in year 2016, is a scientific journal published by the Nature Publishing Group. This work is a joint work with Prof. Ligong Chen’s Lab in the School of Pharmaceutical Science at Tsinghua University and Prof. Jian Peng’s Group in the Department of Computer Science at the University of Illinois at Urbana-Champaign. Yunan Luo (Yao Class 20 at Tsinghua University, now a PhD candidate in the University of Illinois at Urbana-Champaign), Xinbin Zhao and Jingtian Zhou (Class 20 at Tsinghua University, now a PhD candidate in the University of California, San Diego) are the joint first authors, and Jianyang Zeng, Ligong Chen and Jian Peng are the corresponding authors of the paper.

The paper is available at https://www.nature.com/articles/s41467-017-00680-8.

 (Edited by Guo Lili)