Related Experiment Videos
Prediction of partition coefficient based on atom-type electrotopological state indices
J J Huuskonen1, A E Villa, I V Tetko
1Division of Pharmaceutical Chemistry, Department of Pharmacy, POB 56, FIN-00014 University of Helsinki, Finland.
Journal of Pharmaceutical Sciences
|February 9, 1999
Summary
Atom-type electrotopological state indices effectively estimate octanol-water partition coefficients (log P) for drug compounds. Artificial neural networks showed superior prediction accuracy compared to multilinear regression for screening chemical databases.
Area of Science:
- Computational chemistry
- Cheminformatics
- Drug discovery
Background:
- The octanol-water partition coefficient (log P) is crucial for drug development, influencing absorption, distribution, metabolism, and excretion.
- Accurate log P prediction is essential for screening large chemical libraries and optimizing drug candidates.
- Atom-type electrotopological state indices offer a way to represent molecular structure computationally.
Purpose of the Study:
- To evaluate the efficacy of atom-type electrotopological state indices for predicting log P values.
- To compare the performance of multilinear regression and artificial neural networks in log P estimation using these indices.
Main Methods:
- Utilized a dataset of 345 drug compounds and related structures.
- Developed predictive models using multilinear regression and artificial neural networks.
- Employed atom-type electrotopological state indices and molecular weights as input parameters.
Main Results:
- Both multilinear regression and artificial neural networks yielded reliable log P estimations.
- Artificial neural networks demonstrated superior prediction ability for both training and test datasets.
- Atom-type electrotopological state indices proved to be valuable parameters for rapid log P evaluation.
Conclusions:
- Atom-type electrotopological state indices are effective for predicting log P values in drug-like molecules.
- Artificial neural networks offer enhanced predictive performance for log P estimation.
- These indices facilitate efficient screening of large chemical databases, aiding in drug discovery efforts.