September 10 - 14, 2006, ACS Fall 2006, San Francisco, CA, USA

Active Algorithm Training—A Key to Accurate Physicochemical Predictions Applied to Lead Optimization

Karim Kassam, Ed Kolovanov, Sanji Bhal, Greg Pearl

Abstract

Is there a price to be paid for the application of inaccurate predictions? Training is much easier than you may think. As a result of the additive-constitutive fragmental approach taken by ACD/Labs for physicochemical property prediction, the software offers algorithm training tools for pK_a, logP, logD, and, most recently, solubility. Accuracy Extender and System Training allow chemists to utilize in-house knowledge from experimental measurements to improve predictions for proprietary/novel compounds. In turn, improved accuracy of prediction can have a major impact on the use of these values. In this presentation, we shall be discussing the details of algorithm training and illustrating the effective application of predictors in ACD/Structure Design Suite. This software tool rapidly identifies appropriate structural modifications for optimization of physicochemical properties and select ADME parameters of compounds in the lead optimization process.

Download the presentation in MS PowerPoint (1.52 Mb ZIP file) or Adobe Acrobat format (994 Kb PDF file).