| Profiling of GC and GC-MS Data Sets by PCA and Self-Organizing Maps (SOM) |
|
for Correlations of Multiple Data Attributes Alexej Nikiforov1; Barbara Remberg2 and Robert Mistrik3 1Institute of Organic Chemistry, University of Vienna, Waehringerstr. 38, 1090 Vienna, Austria Profiling (multiple comparisons) of samples represented by simple chromatographic data with regard to their underlying (but not directly measurable) attributes is difficult. This is particularly true for multivariate data. In all such cases, provision has to be made to enable the alignment of chromatograms, and to ensure that corresponding components of each sample in a data set are the same. This does not pose a problem for comparisons of non-chromatographic data since the assignment of sample components to particular mass numbers in always unequivocal. In the case of coupled chromatographic-spectroscopic data, the latter. e. g., MS data are used to ensure sample component identity.
This poster presents a semi-automatic approach to the profiling of simple chromatographic data and subsequent correlation of sample attributes. The approach involves the alignment of data, the selection of corresponding peaks in each data set, and the comparison of resulting component profiles by PCA and/or SOM. Visualization of the correlation of SOM classifications whit up to three underlying sample attributes is also presented as a rapid means to assess the appropriateness of the selected peaks to reflect sample attributes.
The example presented is the profiling of GC data of ephedrine, a precursor in illicit drug production, to link samples depending on their source. Also presented are SOM-based approaches to the classification of mass spectra from a given chromatogram, and the search for possible substructures for spectra that are not contained in MS libraries. The software used was HighChem Mass Frontier in combination with Microsoft Excel.
Contact: This e-mail address is being protected from spam bots, you need JavaScript enabled to view it |
