Dr. Xuegong Zhang earned his BS degree in Industrial Automation in 1989 and Ph.D. degree in Pattern Recognition and Intelligent Systems in 1994, both from Tsinghua University, Beijing, China. He joined the faculty of Tsinghua University, Department of Automation in 1994, where he is now Professor of Pattern Recognition and Bioinformatics now. Dr. Zhang worked at Harvard School of Public Health as a visiting scientist on computational biology in 2001-2002, and in Feb-March of 2006, both in the Biostatistics Department. He visited the program of computational and molecular biology of the University of Southern California in Mar-Apr of 2007. Currently he is the Director of the Bioinformatics Division, Tsinghua National Laboratory for Information Science and Technology (TNLIST), and the deputy director of the MOE Key Laboratory of Bioinformatics. His research interests include biological data mining, gene expression and regulation, alternative splicing, microRNA and RNA regulation, computational analysis of haplotypes and meiotic recombination hotspots, etc..

Email: zhangxg@tsinghua.edu.cn

Homepage: http://www.au.tsinghua.edu.cn/szll/XuegongZhangPage/XZhang_English.htm

Tentative Title

Learning biology with machines: examples from alternative splicing and DNA methylation study

Abstract

Machine learning methods such as support vector machines (SVMs) and artificial neural networks have become a major category of tools in mining today¡¯s high-throughput data in molecular and systems biology. New biological data and questions are emerging very rapidly, which brings challenges for machine learning methods. Enormous efforts have been put for improving performances of existing algorithms and developing new algorithms. These efforts have helped in better solving many biological problems and also advanced the study of the machine learning methods. However, for many current biological questions, the bottleneck may not be the algorithm itself, but the way the algorithm is tuned for the specific question. In this talk, I will present several examples of our recent study to illustrate how SVMs can be better tuned for different types of biological questions, and how the experiments with learning algorithms can help to abtain better understanding to the biological questions. The examples include the discrimination of alternative and constitutive splicing sites and the prediction of CpG island methylation. Not only the performance of the methods were improved in terms of the prediction accuracy, but also new insights were achieved on the regulation mechanism of alternative splicing and patterns of DNA methylation protection.