中央研究院 資訊科學研究所

活動訊息

友善列印

列印可使用瀏覽器提供的(Ctrl+P)功能

TIGP -- Prediction for Adverse Drug Events by Chemical Descriptors and Statistical Learning

:::

TIGP -- Prediction for Adverse Drug Events by Chemical Descriptors and Statistical Learning

  • 講者盧鴻興 教授 (交通大學統計所)
    邀請人:TIGP Bioinformatics Program
  • 時間2013-04-09 (Tue.) 14:00 ~ 15:10
  • 地點資訊所新館106演講廳
摘要

In addition to the medicine treatment effect, side effects are complex undesired phenomena due to the bio-activity of pharmaceutical
compound.For each compound, the chemistry informatics can delineate its intrinsic chemical formula into chemistry informatics
features.Based on the assumption that the chemical structure is critical to the biological perturbation in the human system, we
investigate different adverse drug events with associated chemistry informatics features of marketed drugs.In this research, we identify
1,384 ADEs with corresponding associated chemistry informatics features by decision tree.With an automatic analysis workflow, we can
obtain a concordant drug subset with satisfying 10-fold cross-validation accuracy.The accuracy of selected 35 ADEs in the test
experiment is higher than 80%. For example, there are three ADEs of interest and their accuracy: Diabetes Mellitus (0.871), Renal Failure
Acute (0. 910) and Renal Impairment (0. 946). This is a joint work with Pei-Ling Liu and Yi-Jing Wu.