Applications of chemometric algorithms to ion mobility spectrometry and matrix assisted laser desorption/ionization time-of-flight mass spectrometry

Advanced chemometric algorithms have been used for data preprocessing, feature selection, classification, and data compression of ion mobility spectrometry and matrix assisted laser desorption/ionization time-of-flight mass spectrometry data. Ion mobility spectrometry (IMS) furnishes sensitive, portable, robust, and inexpensive sensors that have a variety of potential applications. Matrix assisted laser desorption/ionization mass spectrometry (MALDI-MS) has proved useful for the characterization of bacteria and the detection of biomarkers.;A novel method that couples multivariate curve resolution (MCR) and artificial neural networks was first developed to identify toxic industrial chemicals using fused positive and negative ion mobility spectra. This new method enhances the capability of IMS significantly for identification of toxic industrial chemical. Seven toxic industrial chemicals which are difficult to identify using drift time window method were successfully identified with this new developed method. Simple-to-use interactive self-modeling mixture analysis (SIMPLISMA) was used to separate the analytical peaks in the ion mobility spectrum from the reactant ion peak (RIP). The SIMPLSIMA analytical components of the positive and negative ion peaks were combined together in a butterfly representation (i.e., negative spectra are reported with negative drift times and reflected with respect to the ordinate). Temperature constrained cascade correlation neural network (TCCCN) models were built to classify toxic industrial chemicals. Seven common toxic industrial chemicals were used to evaluate the performance of the algorithm. Ten bootstrapped Latin partitions demonstrated that the classification of neural networks using the SIMPLISMA components were statistically better than neural network models trained with fused IMS spectra.;The MCR techniques were further improved for modeling IMS spectra with the help of thermodynamic and non-negativity constraints. Based on the equilibria in the ionization region of IMS, two more constraints, mass and charge conservation, were used to optimize the concentration profiles and spectra of components. The feasibility of thermodynamic modeling for quantitative analysis was investigated using a vapor generator.;In MALDI-MS, a key challenge for measurements of bacteria is overcoming the relatively large variability in peak intensities. A soft tool, combining analysis of variance and principal component analysis (ANOVA-PCA), was first applied to investigate the effects of the experimental factors associated with MALDI-MS studies of microorganisms. The growth ages was found to be a significant factor and several ribosomal proteins were found using variable loading from the PCA. ANOVA-PCA is useful to facilitate the detection of biomarkers in that it can remove the variance corresponding to other experimental factors from the measurements that might be mistaken for a biomarker. Four strains of Escherichia coli at four different growth ages were used for the study of reproducibility of MALDI-MS measurements.;The linear discrete wavelet transforms (DWT) and non-linear DWT were used to denoise and compress the MALDI-MS spectra to address the two key challenges: the relatively high noise level and the underdetermined data set. It is the first time that a new approach based on multivariate projected difference resolution has been used to evaluate the effects of linear and non-linear DWT on classification. The cross-validation study has proved the classification accuracy increased after data compression with DWT. Supervised feature selection based Fisher's criterion was first applied to choose wavelet coefficients. The high compression ratio is of importance for building MALDI-MS library for database search.