Header image  
Tutorials: Nov 13, 2011, Shanghai, China
Main Conference: Nov 14-17, 2011, Shanghai, China
Workshops: Nov 18, 2011, Hangzhou, China
line decor
   Final program of ICONIP2011 and book of abstracts are available now.  
line decor

Data Mining-based Approach for Drug-Target Prediction


Shanfeng Zhu
Fudan University, China

Hiroshi Mamitsuka
Kyoto University, Japan


The primary objective of this tutorial is to introduce a bioinformatics (and chemoinformatics) problem which has a unique setting and has immense influence over broad areas including those both in science and engineering. A wide variety of information works for solving this problem, and currently a lot of relevant data are available. The secondary objective is then to explain the state-of-the-art of the approaches for this problem, particularly focusing on data-intensive technique, including text mining, multi-task learning and frequent pattern mining.

Content and Benefits:

Drugs are chemical compounds bound to biologically working molecules, mainly proteins or gene products, leading to activation/inhibition of biologically wired pathways in cells to create medically effective control in body. Even developing one drug has been done by heavily laborious experimental efforts which need high financial cost and extremely long duration, typically more than ten years. Thus computationally predicting drugs which can be interacting with one key target can bring enormous assistance in drug discovery if the prediction is precise.

Drug-target binding is physical, three-dimensional interaction and at the same time, can be selective, co-occurrence events between two different sets of entities. This tutorial is focused on the latter perspective, where each drug and target has its own features, which can be explained in a variety of manners. For example, drugs can be represented by physico-chemical properties, i.e. numerical vectors, and have chemical structures, corresponding to graphs in computer science. Furthermore drug-target interactions themselves can be represented by a variety of manners, including examples in drug-target databases and co-occurred pairwise terms in biomedical documents. Thus a variety of data and approaches have been proposed for the problem of predicting new interaction of drug-target, while this tutorial will be devoted to three most promising approaches: those based on text mining, multi-task learning and frequent pattern mining.

Generally speaking the problem is to predict a new edge in a bipartite graph between two distinct sets, where a variety of data can be used behind each of two sets as well as edges themselves, depending on the taken approaches. This means that widely different types of directions in data usage will be shown, which will give audience hints of the manner of using data and finding unknown data. At the same time, this tutorial might be a motivation for the audience to start getting interested in bioinformatics and chemoinformatics.

Target Audience:

The problem must be totally new for the general audience, and so the background knowledge and problem setting will be described well. The target audience will then be researchers, engineers and students who are in machine learning and related fields and are interested in application-oriented work whatever level they are at.


Shanfeng Zhu is an associate professor at School of Computer Science, and Shanghai Key Lab of Intelligent Information Processing at Fudan University, China. He obtained his Ph.D.in Computer Science at City University of Hong Kong in 2003. Before joining Fudan University in July 2008, he was a postdoctoral fellow at Bioinformatics Center, Institute for Chemical Research of Kyoto University. His research focuses on developing machine learning and data mining algorithms as well as related applications in information retrieval and bioinformatics.

Hiroshi Mamitsuka is Professor of Bioinformatics Center, Institute for Chemical Research of Kyoto University, Japan, being jointly appointed as Professor of School of Pharmacological Sciences of the same university. He has been working in machine learning, data mining and bioinformatics, having published totally more than 100 publications, which have mainly appeared in the top conferences and journals of the relevant areas.

Copyright© 2010-2011 International Conference on Neural Information Processing. All rights reserved.