Related Experiment Videos
Prediction of human intestinal absorption using an artificial neural network
1Department of Pharmacy, Zhejiang University City College, Hangzhou, PR China. Fuxc@zucc.edu.cn
Die Pharmazie
|October 15, 2005
Summary
This study introduces an artificial neural network to predict human intestinal absorption (%FA) using molecular structural parameters. The model accurately estimates compound absorption, aiding drug development and safety assessments.
Area of Science:
- Computational chemistry
- Pharmacokinetics
- Artificial intelligence in drug discovery
Background:
- Predicting human intestinal absorption (%FA) is crucial for drug development.
- Accurate %FA prediction informs compound selection and reduces late-stage failures.
- Existing methods may not fully capture complex structure-absorption relationships.
Purpose of the Study:
- To develop and validate an artificial neural network (ANN) model for predicting %FA.
- To identify key molecular structural parameters influencing intestinal absorption.
- To provide a computational tool for rapid %FA estimation.
Main Methods:
- An ANN model was constructed using molecular descriptors such as polar molecular surface area (PSA), fraction of polar molecular surface area (FPSA), and atomic charges.
- The model was trained on a dataset of 85 compounds.
- Model performance was evaluated on an independent test set of 10 compounds.
Main Results:
- The ANN model successfully predicted %FA using selected molecular structural parameters.
- Root mean squared errors (RMSE) for the training and test sets were 8.86% and 14.1%, respectively.
- Key parameters included PSA, FPSA, and various atomic charge sums.
Conclusions:
- The developed ANN model offers a reliable method for predicting %FA.
- This computational approach can streamline early-stage drug discovery and development.
- The identified molecular descriptors provide insights into factors governing intestinal absorption.