We present gCOMBINE, a Java-written graphical user interface (GUI) for performing comparative binding energy (COMBINE) analysis (Ortiz et al. J Med Chem 1995; 38:2681-2691) on a set of ligand-receptor complexeswith the aim of deriving highly informative quantitative structure-activity relationships. The essence of the method is to decompose the ligand-receptor interaction energies into a series of terms, explore the origins of the variance within the set using Principal Component Analysis, and then assign weights to selected ligandresidue interactions using partial least squares analysis to correlate with the experimental activities or binding affinities. The GUI allows plenty of interactivity and provides multiple plots representing the energy descriptors entering the analysis, scores, loadings, experimental versus predicted regression lines, and the evolution of parameterssuch as r(2) (correlation coefficient), q(2) (cross-validated r(2)), and prediction errors as the number of extracted latent variables increases. Other representative features include the implementation of a sigmoidal dielectric function for electrostatic energy calculations, alternative cross-validation procedures (leave-N-out and random groups), drawing of confidence ellipses, and the possibility to carry out several additional tasks such as optional truncation of positive interaction energy values and generation of ready-to-use PDB files containing information related to the importance for activity of individual protein residues. This information can be displayed and color-coded using a standard molecular graphics program such as PyMOL. It is expected that this user-friendly tool will expand the applicability of the COMBINE analysis method and encourage more groups to use it in their drug design research programs.
(c) 2009 Wiley-Liss, Inc.