QSAR models are usually constructed from a single class, or type, of descriptors.  In particular, it is infrequent that 1D/2D descriptors and 3D/4D descriptors are jointly used as a trial descriptor pool in building a QSAR model.  Yet there is no reason for not using such joint trial descriptor pools, and one can ascertain situations where multiple classes of descriptors are needed.  The optimum QSAR models for in vitro receptor-binding endpoints are often realized using only 3D/4D descriptors, but 1D/2D descriptors capturing solvation, entropic and hydrophobic properties of the ligand-receptor binding event may also be needed.  Moreover, cell-based potency and in vivo endpoints both may require 1D/2D descriptors to take into account a modulation in endpoint potency due to transport restrictions, as well as well as 3D/4D descriptors to describe ligand-receptor binding.  ADME transport properties, like cell penetration, may require 1D through 4D descriptors to capture both the magnitude and direction of the solute trajectory through the cell membrane.  Examples of QSAR models built from different classes and types of trial descriptor pools will be given to demonstrate a) enhanced QSAR model quality in some instances using joint classes of descriptors, and b) the opportunity to better explore possible mechanisms of action by comparing and contrasting QSAR models for the same endpoint built from single and from joint sets of descriptor classes.

Combining multiple classes and types of descriptors to optimize QSAR models

243rd National Meeting (2012)

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