Related Experiment Videos
Application of non-parametric regression to quantitative structure-activity relationships.
Jonathan D Hirst1, T John McNeany, Trevor Howe
1School of Chemistry, University of Nottingham, University Park, NG7 2RD, Nottingham, UK. jonathan.hirst@nottingham.ac.uk
Bioorganic & Medicinal Chemistry
|February 12, 2002
Summary
Non-parametric regression models show promise for quantitative structure-activity relationship (QSAR) studies. A Nadaraya-Watson kernel estimator achieved the best predictive performance for Syk inhibitors.
Area of Science:
- Medicinal Chemistry
- Computational Chemistry
- Pharmacology
Background:
- Quantitative Structure-Activity Relationship (QSAR) modeling is crucial for drug discovery.
- Evaluating diverse regression techniques is essential for optimizing QSAR model performance.
Purpose of the Study:
- To assess the efficacy of various non-parametric regressors for QSAR modeling.
- To benchmark non-parametric methods against traditional regression techniques.
Main Methods:
- Applied non-parametric regressors including Nadaraya-Watson kernel estimator.
- Utilized multilinear regression and smoothing splines for comparison.
- Explored variable selection via systematic combinations and principal components.
Main Results:
- The best two-descriptor model, a multi-variate Nadaraya-Watson kernel estimator, achieved a q2 of 0.43 on a training set of 539 Syk inhibitors.
- This model demonstrated predictive power on an independent test set of 371 compounds.
- Non-parametric methods outperformed other approaches evaluated.
Conclusions:
- Non-parametric regression is a powerful approach for developing predictive, low-dimensional QSAR models.
- Further descriptors can improve predictive ability but may reduce model interpretability.