Using the Ranking-Based KNN Approach for Drug Repositioning Based on Multiple Information
Using effective computer methods to infer potential drug-disease relationships can provide clues for the discovery new uses of old drugs. This paper introduced a Ranking-based k-Nearest Neighbor (Re-KNN) method to drug repositioning for cardiovascular diseases. The main characteristic of the Re-KNN lies in combining conventional KNN algorithm with Ranking SVM (Support Vector Machine) algorithm to get neighbors that are more trustable. By integrating the chemical structural similarity, target-based similarity, side-effect similarity and topological similarity information, Re-KNN method can obtain an improved AUC (Area under ROC Curve) and AUPR (Area under Precision-Recall curve) compared with other methods, which prove the validity and efficiency of multiple features integration.
KeywordsCardiovascular disease Drug reposition Ranking-based KNN Ranking SVM
This work was partially supported by National Basic Research Program of China (Grants No. 2012CB910400), the Fundamental Research Funds for the Central Universities (78260026), and National Science and Technology Support Plan Project (2015BAH12F01).
- 11.Chen, H., Zhang, H., Zhang, Z., et al.: Network-based inference methods for drug repositioning. Comput. Math. Methods Med. 2015, 130620 (2015)Google Scholar
- 14.Chiang, T.H., Lo, H.Y., Lin, S.D.: A ranking-based KNN approach for multi-label classification. JMLR W&CP. 25, 81–96 (2012)Google Scholar
- 22.Tsoumakas, G., Katakis, I.: Multi-label classification: an overview. Depatment of Informatics, Aristotle University of Thessaloniki, Greece (2006)Google Scholar
- 23.Biggs, R., Hayton-Williams, D.S.: Epsilon-aminocaproic acid in the treatment of hemophilia. Br. Dent. J. 124(124), 157 (1968)Google Scholar
- 25.Giardini, O., Cantani, A., Donfrancesco, A.: Vitamin E therapy in homozygous beta-thalassemia. N. Engl. J. Med. 305(11), 644 (1981)Google Scholar