Multivariate image analysis (MIA) applied to quantitative structure– activity relationship (QSAR) analysis, whose descriptors are derived from pixels of two-dimensional (2D) chemical structures that may be built by using any appropriate drawing software, was used to model 17 glycogen synthase kinase 3 (GSK-3) inhibitors. Calibration was carried out using partial least squares (PLS) regression, and an r 2 value of 0.93 was obtained for four latent variables. Leave-one-out crossvalidation and a robustness test were performed, and the model proved to be a suitable alternative QSAR approach for modeling this series of compounds.