Related Experiment Videos
Combinatorial QSAR modeling of P-glycoprotein substrates.
Patricia de Cerqueira Lima1, Alexander Golbraikh, Scott Oloff
1Division of Natural and Medicinal Chemistry, The Laboratory for Molecular Modeling, School of Pharmacy, CB# 7360, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599-7360, USA.
Journal of Chemical Information and Modeling
|May 23, 2006
Summary
This study introduces a combinatorial quantitative structure-activity relationship (QSAR) approach to improve drug resistance models. By combining various methods and descriptors, it identifies superior predictive models for P-glycoprotein substrates, enhancing drug discovery efforts.
Area of Science:
- Computational chemistry
- Cheminformatics
- Drug discovery
Background:
- Quantitative structure-activity relationship (QSAR) and quantitative structure-property relationship (QSPR) models are vital in predicting molecular properties.
- Traditional QSAR/QSPR models often use a single technique and descriptor type.
- P-glycoprotein (P-gp) plays a critical role in multidrug resistance, making its substrates and nonsubstrates an important research area.
Purpose of the Study:
- To explore a combinatorial QSAR approach for modeling P-glycoprotein (P-gp) substrates and nonsubstrates.
- To evaluate multiple combinations of modeling techniques and molecular descriptor types for predictive accuracy.
- To identify the most effective QSAR models for predicting P-gp interaction.
Main Methods:
- Applied a combinatorial QSAR approach to a dataset of 195 diverse P-gp substrates and nonsubstrates.
- Utilized various modeling methods: k-nearest neighbors classification, decision tree, binary QSAR, and support vector machines (SVM).
- Employed diverse descriptor sets: molecular connectivity indices, atom pair (AP) descriptors, VolSurf descriptors, and molecular operation environment descriptors, testing all 16 combinations.
Main Results:
- All 16 combinations yielded high correct classification rates for the training set (CCR(train)).
- Predictive models with high accuracy on the test set (CCR(test)) were generated for specific combinations.
- The best models utilized SVM classification with either AP or VolSurf descriptors, achieving CCR(train) = 0.94/0.88 and CCR(test) = 0.81/0.81.
Conclusions:
- The combinatorial QSAR approach identified models with superior predictive accuracy compared to previous studies for the same dataset.
- This methodology offers a more robust alternative to single-method QSAR/QSPR modeling.
- The combinatorial QSAR approach is proposed as a standard practice for future QSPR/QSAR modeling endeavors.