|
|
| Type |
Poster Presentation |
| Area |
의약화학 |
| Room No. |
포스터발표장 |
| Time |
4월 20일 (목요일) 11:00~12:30 |
| Code |
MEDI.P-438 |
| Subject |
QSAR Modeling for hERG Inhibition by Learning Machine Methods |
| Authors |
서용일, 남기엽*, 서영주, 문자연, 김민경, 윤정혁1
(주)파로스아이비티 바이오인공지능연구소, Korea
1(주)파로스아이비티 -, Korea
|
| Abstract |
The hERG (the human Ether-à-go-go-Related Gene) is a one of ion channel protein to transport the potassium ions to selective. This gene contributes to the electrical activity of the heart such as QT interval. If the hERG ion channel does not play a role due to genetic or artificial factors, it will cause a long QT syndrome. Therefore, many drug candidates are dropped when they bind to hERG or interfere with the passage of potassium ions. There conduct an experiment in vitro like measuring half maximal inhibitory concentration (IC50) to prevent it. In recently, some compounds predict at risk for hERG inhibition using QSAR model before in vitro test. This study builds QSAR Models to predict hERG inhibitior qualitatively and quantitatively by using various machine learning methods such as MLR, LR, SVM, and ANN. We collect compounds through ChemBL for targeting with hERG and select training set by the experimental method, cell environment, and measurement method. Additionally, the compounds for quantitative prediction were selected so that the distribution of the experimental values was similar. To qualitative prediction, the experimental value of the compound is defined as Active and Inactive by criteria on 10-5 M, and the training set is selected to have as a ratio of 1: 1. |
| E-mail |
tntmc@naver.com |
|
119th General Meeting of the KCS