Laboratory for Molecular Information Systems

Department of Computer Science and Engineering, Toyohashi University of Technology

By Prof. Yoshimasa Takahashi

 

Description
We are doing research works on the development of algorithms and software tools for molecular structure information processing and intelligent systems for drug design and development aiming toward the establishment of domain-specific information technology in chemistry and the related fields.

 

Theme1FStudies on algorithms for molecular information processing
Research Outline

gSimilarity" is very important concept in solving problems in science. This is true in chemistry. Especially structural similarity provides us a lot of information on structure-activity and structure-property problems. There are two different viewpoints:
(1) What is similar among the structures?
(2) How much are they similar?
From these viewpoints, in our laboratory, the fundamental studies on new algorithm and software tools for the evaluation of structural similarity/diversity using a graph theoretical approach.

 

ƒeƒLƒXƒg ƒ{ƒbƒNƒX:  

Visualization of the TFS space and similar structure searching

KeywordFF graph theory, molecular similarity, pattern recognition, pattern clustering, visualization of multivariate data, data mining, graph mining, visualization@of multi-dimensional structural feature space, automatic recognition of 3-dimensional common structural feature@

 

http://www.tut.ac.jp/hs_template/teach/img/images/dot616.gif

Theme2FChemical artificial intelligence system based on machine learning
Research Outline

On the basis of a chemical structure which is drawn on the computer, structural feature of the drug molecule is analyzed automatically, and the feature profile is expressed as digital spectra by TFS (Topological Fragment Spectra) method developed by our laboratory. The correlations between the spectra and activity (or toxicity) of known chemical compounds are trained by the machine learning such as artificial neural network, and by studying the mutual relationship, the development research of the system which presumes the safety and character of the new useful chemical substance is being advanced.

 

ƒeƒLƒXƒg ƒ{ƒbƒNƒX:  
Artificial inteligent system for drug design and development

KeywordFF machine learning, artificial neural network, support vectoring machine, structure-activity relationship, toxicity prediction, data mining, knowledge discovery

 

Others
Digital representation of structural information//Similar structure searching /Visualization of similarity data space /Graph mining for chemical data

Publications:
‘Kentaro Kawai, Yoshimasa Takahashi, Virtual Screening of Antihypertensive Drugs Using Support Vector Machines, J. Comput. Chem. Jpn., 9,167-176(2010).
‘Kentaro Kawai, Yoshimasa Takahashi, Identification of the Dual Action Antihypertensive Drugs Using TFS-Based Support Vector Machines, Chem-Bio Informatics Journal, 9, 41-51 (2009).
‘Kentaro Kawai, Satoshi Fujishima, and Yoshimasa Takahashi, "Predictive Activity Profiling of Drugs by Topological-Fragment-Spectra-Based Support Vector Machines", J. Chem. Inf. Model., 48, 1152-1160 June (2008).