EnsembleQSAR: A QSAR method based on conformational ensembles and metric descriptors

Abstract Quantitative structure–activity relationship (QSAR) is the most versatile tool in computer‐assisted molecular design. One conceptual drawback seen in QSAR approaches is the “one chemical–one structure–one parameter value” dogma where the model development is based on physicochemical description for a single molecular conformation, while ignoring the rest of the conformational space. It is well known that molecules have several low‐energy conformations populated at physiological temperature, and each conformer makes a significant impact on associated properties such as biological activity. At the level of molecular interaction, the dynamics around the molecular structure is of prime essence rather than the average structure. As a step toward understanding the role of these discrete microscopic states in biological activity, we have put together a theoretically rigorous and computationally tractable formalism coined as e QSAR. In this approach, the biological activity is modeled as a function of physicochemical description for a selected set of low‐energy conformers, rather than that's for a single lowest energy conformation. Eigenvalues derived from the “Physicochemical property integrated distance matrices” ( PD‐matrices ) that encompass both 3D structure and physicochemical properties, have been used as descriptors; is a novel addition. e QSAR is validated on three peptide datasets and explicitly elaborated for bradykinin‐potentiating peptides. The conformational ensembles were generated by a simple molecular dynamics and consensus dynamics approaches. The e QSAR models are statistically significant and possess the ability to select the most biologically relevant conformation(s) with the relevant physicochemical attributes that have the greatest meaning for description of the biological activity. © 2011 Wiley Periodicals, Inc. J Comput Chem, 2011