Related Experiment Videos
Classification of dopamine antagonists using TFS-based artificial neural network.
Satoshi Fujishima1, Yoshimasa Takahashi
1Laboratory for Molecular Information Systems, Department of Knowledge-Based Information Engineering, Toyohashi University of Technology, Hibarigaoka 1-1, Tempaku-cho, Toyohashi 441-8580, Japan.
Journal of Chemical Information and Computer Sciences
|May 25, 2004
Summary
The Topological Fragment Spectral artificial neural network (TFS/ANN) effectively classifies and predicts pharmacological activity. This computational chemistry approach achieved high accuracy in identifying dopamine antagonist drug classes.
Area of Science:
- Computational chemistry
- Cheminformatics
- Pharmacology
Background:
- The Topological Fragment Spectral (TFS) method was previously developed for molecular topological structure profiling.
- Accurate classification and prediction of pharmacological activity are crucial in drug discovery.
Purpose of the Study:
- To develop and evaluate a TFS-based artificial neural network (TFS/ANN) for classifying and predicting pharmacological activity.
- To assess the model's performance on known and unknown chemical compounds.
Main Methods:
- Utilized the Topological Fragment Spectral (TFS) method to generate molecular descriptors.
- Developed an artificial neural network (ANN) model trained on a dataset of 1227 dopamine antagonists.
- Validated the TFS/ANN model on a separate set of 137 unknown compounds.
Main Results:
- The TFS/ANN model achieved 89% accuracy in classifying known dopamine antagonists into their respective active classes (D1, D2, D3, D4).
- For an independent prediction set, the model correctly predicted the active classes of 111 out of 137 (81%) compounds.
Conclusions:
- The TFS/ANN approach demonstrates significant potential for accurate classification and prediction of pharmacological activity.
- This method offers a valuable computational tool for accelerating drug discovery and development by identifying potential drug candidates.