Three closely related cancer cell lines have been analysed with ToF-SIMS using a C-60(+) primary ion beam. Principal component-discriminant function analysis (PC-DFA) has been applied for spectral classification. Various spectral pre-processing methods are discussed and assessed for optimum discrimination of this data set. The sum-normalised PC-DFA spectral model produced sensitivities as high as 83.3% and specificities as high as 100% at the 99% confidence limit. At this confidence limit only one errant spectrum was misclassified. The resulting loadings plots suggest that a range of lipid and amino-acid related signals are responsible for the cell line discrimination. (C) 2008 Elsevier B. V. All rights reserved.