Representing and Comparing Molecules using Spherical Harmonic Expansions
Lazaros Mavridis, Brian D. Hudson, D. A. Ritchie · 2004
Performing quantitative structure-activity relationship (QSAR) studies is often a vital early step in the drug development process. If detailed experimental data for the molecules under study is not available, their structure and chemical properties can often be calculated satisfactorily using fast semi-empirical quantum mechanical techniques [1]. Currently, corporate and public databases now contain hundreds of thousands, or even millions of compounds, and it is becoming feasible to calculate and store the structures and properties for entire data sets. The ability to search such databases using 3D shape-based or property-based ''scaffold-hopping'' queries offers would then offer the intriguing possibility of finding novel or unexpected drug leads [2]. We are investigating the use of spherical harmonic (SH) expansions as an efficient way to represent and compare small ligand molecules. It has been shown previously that the spherical harmonic representation is particularly well suited to calculating rotational correlations in order to superpose similar molecular shapes [3]. Here, we describe early work on “SpotLight,” a novel 3D molecular superposition and database system. SpotLight can compare, superpose, and cluster the 3D shapes and surface properties of small ligands using only their spherical harmonic expansion coefficients. Introduction