In an earlier study, we demonstrated the use of statistical modeling using mixture models and principal components analysis (PCA) in characterizing samples having a relatively simple chemistry and coupled with topographical features [Proc. Am. Vac. Soc. (in press)]. This study involves the extension to analysis of samples with more complex chemistry, such as with proteins. In this study, polystyrene microspheres were adsorbed with different proteins and statistical models such as PCA and mixture models were used to analyze the ToF-SIMS images. Discrimination of protein spectra from the images was then performed using the afore-mentioned techniques, without a priori information about the type of protein adsorbed onto the sphere surfaces. In one set of experiments, each sample was prepared by coating the microspheres with a certain type of protein. Using PCA, we could differentiate between these samples based on their protein spectra. In another experiment, the sample consisted of a set of microspheres, each coated with a different kind of protein. Mixture models were used to contrast between these microspheres on ToF-SIMS images. (C) 2004 Elsevier B.V. All rights reserved.