A Raman spectroscopy cell-based biosensor has been proposed for rapid detection of toxic agents, identification of the type of toxin and prediction of the concentration used. This technology allows the monitoring of the biochemical properties of living cells over long periods of time by measuring the Raman spectra of the cells non-invasively, rapidly and without use of labels (Notingher et al. 2004 doi:10.1016/j.bios.2004.04.008). Here we show that this technology can be used to distinguish between changes induced in A549 lung cells by the toxin ricin and the chemical warfare agent sulphur mustard. A multivariate model based on principal component analysis (PCA) and linear discriminant analysis (LDA) was used for the analysis of the Raman spectra of the cells. The leave-one-out cross-validation of the PCA-LDA model showed that the damaged cells can be detected with high sensitivity (98.9%) and high specificity (87.7%). High accuracy in identifying the toxic agent was also found: 88.6% for sulphur mustard and 71.4% for ricin. The prediction errors were observed mostly for the ricin treated cells and the cells exposed to the lower concentration of sulphur mustard, as they induced similar biochemical changes, as indicated by cytotoxicity assays. The concentrations of sulphur mustard used were also identified with high accuracy: 93% for 200 microM and 500 microM, and 100% for 1,000 microM. Thus, biological Raman microspectroscopy and PCA-LDA analysis not only distinguishes between viable and damaged cells, but can also discriminate between toxic challenges based on the cellular biochemical and structural changes induced by these agents and the eventual mode of cell death.