| Separation science is a field covering a broad range of analytical methodologies. Within the sphere of separation science, gas chromatography with its sensitivity, selectivity and resolving power represents an established means of analysis for a wide variety of complex samples. Developments to the field of gas chromatography in recent years have been largely manifest in work towards the development of miniaturized high-speed gas chromatographic systems for reliable rapid analyses as well as in the development of multidimensional chromatographic systems which provide additional separation power than their one-dimensional counterparts. This dissertation addresses developments in both of these aspects of modern gas chromatography. The ability to perform rapid gas chromatographic separations of complex multicomponent mixtures while still retaining quantitative information is explored in the development of a novel gas chromatographic instrument, the gas chromatographic sensor. The main focus of the dissertation is, however, devoted to the development of a multidimensional approach to the study of metabolite samples and other complicated mixtures involving two-dimensional comprehensive gas chromatographic separations combined with time of flight mass spectrometric detection and chemometric data handling. The task of rapidly and easily determining analytes of interest in a multidimensional separation of a complex mixture can be very difficult. Comprehensive two-dimensional gas chromatography combined with mass spectral detection, however, produces a data set that is amenable to analysis using chemometric techniques such as pattern recognition and deconvolution. The development and application of chemometric tools such as these as well as the application of appropriate data preprocessing methods and data handling tools to accurately and effectively analyze the data make up the emphasis of this work. An algorithm for locating analytes of interest in complicated GC x GC/TOFMS separations, called DotMap is presented and evaluated. A method of feature selection using Principal Component Analysis (PCA) loadings plot information is presented, and the use of Parallel Factor Deconvolution (PARAFAC) is demonstrated. |
