Binary classification of compounds by learning from docking software results and chemical information

Masato Okada, Katsutoshi Kanamori, Hayato Ohwada, Shin Aoki

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

4 Citations (Scopus)

Abstract

This paper proposes a new approach to classification of docking between compounds and proteins for drug design virtual screening. Currently, docking software programs often use real numbers as docking scores; but due to their predictive accuracy, it is difficult for biologists to use such scores in realistic experiments. In contrast, our approach utilizes binary classification that indicates whether a candidate compound is docked to a target protein. This leads to automatic screening of compounds without the input of biologists in drug design. The present method provides consensus use of the scores of existing docking software, yielding higher accuracy of binary classification. In this paper, we discuss several implementations of the proposed method based on Support Vector Classification and Regression. We have created a classification model from docking scores and chemical information. The experiment demonstrates that our method outperforms the existing docking softwares in classification.

Original languageEnglish
Title of host publication5th International Conference on Bioinformatics and Computational Biology 2013, BICoB 2013
Pages125-130
Number of pages6
Publication statusPublished - 13 Sep 2013
Event5th International Conference on Bioinformatics and Computational Biology 2013, BICoB 2013 - Honolulu, HI, United States
Duration: 4 Mar 20136 Mar 2013

Publication series

Name5th International Conference on Bioinformatics and Computational Biology 2013, BICoB 2013

Conference

Conference5th International Conference on Bioinformatics and Computational Biology 2013, BICoB 2013
Country/TerritoryUnited States
CityHonolulu, HI
Period4/03/136/03/13

Keywords

  • Bioinformatics
  • Docking simulation
  • Machine learning
  • Support vector machine

Fingerprint

Dive into the research topics of 'Binary classification of compounds by learning from docking software results and chemical information'. Together they form a unique fingerprint.

Cite this