MALDI imaging is a powerful tool that allows the acquisition of tissue-specific mass spectra with spatial resolution. Here we applied statistical tools to generate information from MALDI imaging datasets. Unsupervised principal component analysis (PCA) is discussed to access tissue-specific variance in the data, and a supervised classification approach is used to classify specific tissue types.
Methods: Tissue cryosections were thaw mounted on conductive coated glass slides, and the MALDI matrix was applied. Data were acquired on a MALDI-TOF mass spectrometer in linear mode. Statistical analyses were performed with the ClinProTools 2.1 software (Bruker Dal-tonics). Either the peptide masses detected by PCA were displayed in the MALDI images, or the principal components from the classification as such were used to visualize tissue classes rather than the distribution of molecular weights across the tissue.
Results: Unsupervised PCA analysis showed that the variance in the MALDI imaging datasets reflected the variance in the tissue. One limitation of the PCA was that on tumor sections from different patients, the variance between the patients was high, and therefore the PCA reflected largely the difference between the patients rather than variance in individual tissues.
Using supervised classification approaches, it was possible to classify characteristic tissue types and to apply these classifications to unknown tissue. This approach opens the door for diagnostic applications.