We show on Imaging Mass Spectrometry (IMS) data that the Random Forest classifier can be used for automated tissue classification and that it results in predictions with high sensitivities and positive predictive values, even when inter-sample variability is present in the data. We further demonstrate how Markov Random Fields and vector-valued median filtering can be applied to reduce noise effects to further improve the classification results in a post-hoc smoothing step. Our study gives clear evidence that digital staining by means of IMS constitutes a promising complement to chemical staining techniques.

Imaging mass spectrometry (IMS) is a promising technology which allows for detailed analysis of spatial distributions of (bio)molecules in organic samples. In many current applications, IMS relies heavily on (semi)automated exploratory data analysis procedures to decompose the data into characteristic component spectra and corresponding abundance maps, visualizing spectral and spatial structure. The most commonly used techniques are principal component analysis (PCA) and independent component analysis (ICA). Both methods operate in an unsupervised manner. However, their decomposition estimates usually feature negative counts and are not amenable to direct physical interpretation. We propose probabilistic latent semantic analysis (pLSA) for non-negative decomposition and the elucidation of interpretable component spectra and abundance maps. We compare this algorithm to PCA, ICA, and non-negative PARAFAC (parallel factors analysis) and show on simulated and real-world data that pLSA and non-negative PARAFAC are superior to PCA or ICA in terms of complementarity of the resulting components and reconstruction accuracy. We further combine pLSA decomposition with a statistical complexity estimation scheme based on the Akaike information criterion (AIC) to automatically estimate the number of components present in a tissue sample data set and show that this results in sensible complexity estimates.

Surface metallization by plasma coating enhances desorption/ionization of membrane components such as lipids and sterols in imaging time-of-flight secondary ion mass spectrometry (TOF-SIMS) of tissues and cells. High-resolution images of cholesterol and other membrane components were obtained for neuroblastoma cells and revealed subcellular details (resolving power 1.5 mum). Alternatively, in matrix-enhanced SIMS, 2,5-dihydroxybenzoic acid electrosprayed on neuroblastoma cells allowed intact molecular ion imaging of phosphatidylcholine and sphingomyelin at the cellular level. Gold deposition on top of matrix-coated rat brain tissue sections strongly enhanced image quality and signal intensity in stigmatic matrix-assisted laser desorption/ionization imaging mass spectrometry. High-quality total ion count images were acquired, and the neuropeptide vasopressin was localized in the rat brain tissue section at the hypothalamic area around the third ventricle. Although the mechanism of signal enhancement by gold deposition is under debate, the results we have obtained for cells and tissue sections illustrate the potential of this sample preparation technique for biomolecular surface imaging by mass spectrometry.

It has been suggested that the bacterial GroEL chaperonin accommodates only one substrate at any given time, due to conformational changes to both the cis and trans ring that are induced upon substrate binding. Using electrospray ionization mass spectrometry, we show that indeed GroEL binds only one molecule of the model substrate Rubisco. In contrast, the capsid protein of bacteriophage T4, a natural GroEL substrate, can occupy both rings simultaneously. As these substrates are of similar size, the data indicate that each substrate induces distinct conformational changes in the GroEL chaperonin. The distinctive binding behavior of Rubisco and the capsid protein was further investigated using tandem mass spectrometry on the intact 800-914 kDa GroEL-substrate complexes. Our data suggest that even in the gas phase the substrates remain bound inside the GroEL cavity. The analysis revealed further that binding of Rubisco to the GroEL oligomer stabilizes the chaperonin complex significantly, whereas binding of one capsid protein did not have the same effect. However, addition of a second capsid protein molecule to GroEL resulted in a similar stabilizing effect to that obtained after the binding of a single Rubisco. On the basis of the stoichiometry of the GroEL chaperonin-substrate complex and the dissociation behavior of the two different substrates, we hypothesize that the binding of a single capsid polypeptide does not induce significant conformational changes in the GroEL trans ring, and hence the unoccupied GroEL ring remains accessible for a second capsid molecule.

Leptomeningeal metastasis (LM) is a devastating complication occurring in 5% of breast cancer patients. However, the current 'gold standard' of diagnosis, namely microscopic examination of the cerebrospinal fluid (CSF), is false-negative in 25% of patients at the first lumbar puncture. In a previous study, we analyzed a set of 151 CSF samples (tryptic digests) by MALDI-TOF and detected peptide masses that were differentially expressed in breast cancer patients with LM. In the present study, we obtain for a limited number of samples exact masses for these peptides by MALDI-FTICR MS measurements. Identification of these peptides was performed by electrospray FTICR MS after separation by nano-scale LC. The database results were confirmed by targeted high mass accuracy measurements of the fragment ions in the FTICR cell. The combination of automated high-throughput MALDI-TOF measurements and analysis by FTICR MS leads to the identification of 17 peptides corresponding to 9 proteins. These include proteins that are operative in host-disease interaction, inflammation and immune defense (serotransferrin, alpha 1-antichymotrypsin, hemopexin, haptoglobin and transthyretin). Several of these proteins have been mentioned in the literature in relation to cancer. The identified proteins alpha1-antichymotrypsin and apolipoprotein E have been described in relation to Alzheimer's disease and brain cancer.

Imaging mass spectrometry (IMS) is a promising technology which allows for detailed analysis of spatial distributions of (bio)molecules in organic samples. In many current applications, IMS relies heavily on (semi)automated exploratory data analysis procedures to decompose the data into characteristic component spectra and corresponding abundance maps, visualizing spectral and spatial structure. The most commonly used techniques are principal component analysis (PCA) and independent component analysis (ICA). Both methods operate in an unsupervised manner. However, their decomposition estimates usually feature negative counts and are not amenable to direct physical interpretation. We propose probabilistic latent semantic analysis (pLSA) for non-negative decomposition and the elucidation of interpretable component spectra and abundance maps. We compare this algorithm to PCA, ICA, and non-negative PARAFAC (parallel factors analysis) and show on simulated and real-world data that pLSA and non-negative PARAFAC are superior to PCA or ICA in terms of complementarity of the resulting components and reconstruction accuracy. We further combine pLSA decomposition with a statistical complexity estimation scheme based on the Akaike information criterion (AIC) to automatically estimate the number of components present in a tissue sample data set and show that this results in sensible complexity estimates.

Electron capture dissociation (ECD) of proteins in Fourier transform ion cyclotron resonance mass spectrometry usually leads to charge reduction and backbone-bond cleavage, thereby mostly retaining labile, intramolecular noncovalent interactions. In this report, we evaluate ECD of the 84-kDa noncovalent heptameric gp31 complex and compare this with sustained off-resonance irradiation collisionally activated dissociation (SORI-CAD) of the same protein. Unexpectedly, the 21+ charge state of the gp31 oligomer exhibits a main ECD pathway resulting in a hexamer and monomer, disrupting labile, intermolecular noncovalent bonds and leaving the backbone intact. Unexpectedly, the charge separation over the two products is highly proportional to molecular weight. This indicates that a major charge redistribution over the subunits of the complex does not take place during ECD, in contrast to the behavior observed when using SORI-CAD. We speculate that the ejected monomer retains more of its original structure in ECD, when compared to SORI-CAD. ECD of lower charge states of gp31 does not lead to dissociation of noncovalent bonds. We hypothesize that the initial gas-phase structure of the 21+ charge state is significantly different from the lower charge states. These structural differences result in the different reaction pathways when using ECD.