Quantitative Structure-Activity Relationship (QSAR) has been widely applied and proven to be somewhat successful in predicting activity of biological molecules. In this study, a new 3D-QSAR methodology, based on kernel Partial Least Squares (kPLS) regression, has been developed to model binding affinity of ligands. For a 3D-QSAR model, proper choice of bioactive ligand conformations is essential. Pose conformers of ligands are used and generated using the POSIT algorithm when corresponding crystallographic structures are not available. Shape and color interactions, as produced by ROCS, are used as descriptors in this methodology. The newly developed method is used to build a model for binding affinities of the human beta-secretase (BACE-1) inhibitors. Results show that the performance of the new model is at least comparable, or slightly superior to currently existing 2D and 3D-QSAR models in the literature. Models built with kPLS and ROCS defines the activity as a linear superposition of the training set structures. The simple mathematical expression of the model allows exploration of the binding pocket using probe particles with varied interactions (e.g., hydrogen bond doner/acceptor etc.). These probe particle interactions could be used to identify favorable and unfavorable regions of characteristic interactions that can provide valuable insight and guidance into molecular design and optimization. To further enhance the predictive capability of the model, current efforts are being directed towards improving pose prediction with MMPBSA, incorporating multiple pose conformers, improving descriptors with better estimates of hydrogen bond donor/acceptor interactions, and incorporation of entropy associated with ligand binding.

3D-QSAR model building with kPLS and ROCS

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.