Using classification structure pharmacokinetic relationship (SCPR) method to predict drug bioavailability based on...

Area of Science:

• Computational Chemistry
• Pharmacokinetics
• Drug Discovery

Background:

• Predicting drug bioavailability is crucial for drug development.
• Quantitative Structure-Property Relationships (QSPR) and Quantitative Structure-Activity Relationships (QSAR) are vital in cheminformatics.
• Classification models can aid in predicting drug properties.

Purpose of the Study:

• To develop and compare classification models for predicting drug bioavailability.
• To utilize molecular descriptors derived solely from molecular structures.
• To assess the reliability of different modeling techniques in drug property prediction.

Main Methods:

• Employed Linear Discriminant Analysis (LDA) and Grid-Search Support Vector Machine (GS-SVM).
• Generated molecular descriptors using CODESSA software from 167 compounds.
• Selected five key descriptors using LDA for model development.

Main Results:

• Both LDA and GS-SVM models demonstrated the discriminative capacity of molecular descriptors for bioavailability.
• GS-SVM achieved a higher total accuracy of 85.6% compared to LDA's 72.4%.
• The results confirm the strong relationship between selected descriptors and drug bioavailability.

Conclusions:

• GS-SVM is a more reliable method for predicting drug bioavailability than LDA.
• The developed models are useful for selecting new drug candidates.
• The approach can be extended to other Classification Structure-Pharmacokinetic Relationship (CSPR) and Classification Structure-Activity Relationship (CSAR) studies.