Caco-2 Permeability Modeling: Feature Selection via Sparse Support Vector Machines
The Virtual Screening Problem
Caco-2 Data
PPT Slide
Approaches to feature selection
Practical Issues
Validation Methodology
Validation Methodology continued...
Feature Selection via Sparse SVM/LP
Potential Pitfalls
Bagged Feature Selection
Bagged SVM (RBF) Caco-2 - 718 Variables
Bagged SVM (RBF) Caco-2 : 31 Variables
Model mining
Star Plot of ABSDRN6
Starplot Caco-2 : 31 Variables
Chemistry In/Out Modeling
Bagged SVM (RBF) Caco-2 - 15 Variables
Caco-2 : 15 Variables
Starplot of a.don
Starplot of SlogP.VSA0
Chemical Insights
Hybrid TAE/SHAPE
Wrapper Feature Selection Results
Iterative Bagged Feature Selection Results
Summary
Acknowledgements
Reserve Slides
Linear m-SVM with 1-norm
Nonlinear Regression
Nonlinear Support Vector
Email: brenec@rpi.edu
Home Page: http://www.drugmining.com