Norm Glassbrook1 and Scott J Campbell2
1Vector Research Ltd, NC, USA; 2SpectralWorks Ltd, The Heath Business & Technical Park, Runcorn, UK
First Published: PittCon 2004
Traditionally the analysis of complex mixtures by various analytical techniques has either been qualitative, trying to assess what is in a sample by use of a mass spectrometer and library searches, or quantitative, where the analysis has usually been conducted in an extremely targeted manner, with non target components being ignored. To aid lowering detection limits whilst using mass spectrometry selected ion monitoring (SIM) has been commonly used for quantification analysis. With the advent of newer generations of more sensitive mass spectrometers the need for using SIM is reduced and hence data files become far more information rich. This allows both the qualitative and quantification analysis of complex samples, which can be extremely advantageous in both the research and discovery environments, as well as areas such as quality control. In this paper we show the use of new data mining algorithms to simultaneously analyze a set of complex data acquired from various tobacco samples both qualitatively and quantitatively to maximize the information extracted from the data. This allows a unique characterization of each sample and subsequent authentication of ‘unknown’ samples.
Chemical profiling is the systematic, comprehensive analysis of the chemical compositions and processes of test systems. Profiling involves cataloguing the chemical components of the test system and measuring the absolute or relative concentrations of those components.
Although classical compound analyses are part of the analytical approach used for chemical profiling, generalized analytical methods are often used to quickly catalog and monitor a wide range of compounds. These generalized methods produce more complex sample extracts and data than traditional methods. As the data becomes more complex, data processing with software designed for target compound analysis becomes prohibitively labor intensive and time consuming.
Vector Research is currently characterizing the chemical composition of tobacco and tobacco smoke. Here we report on the use of instrument vendor-supplied software and ‘second-source’ data processing software for the analysis of profiling data from the GC/MS analysis of a reference cigarette.
Materials and Methods
Different sample weights of Kentucky reference cigarettes (2R4F) were extracted in organic solvent and centrifuged to remove particulate.
An aliquot of the underivatized supernatant was analyzed by GC/MS.
The remaining extract and sample pellet were treated to produce TMS derivatives and the supernatant was analyzed by GC/MS.
Chromatographic components were catalogued by retention index (Kovats), and mass spectrum of each component.
Peak areas for various components were plotted versus sample weight to verify linearity over the range of sample loading.
Test Samples and Reference Materials
2R4F, finely ground, 3-30mg/tube in duplicate
300uL extraction solvent with ISTDs per tube
Paraffins, C10-C40 even, at 4ug/mL ea as retention STDs
Agilent 6890 Gas Chromatograph
Agilent 5973N Mass Selective Detector
SCAN MODE, m/z 40-540
SGE BPX5 Column: 25m X 0.22mm, 0.25um df
Helium Carrier Gas at ~40cm/sec
Programmed Oven Temperature 60°C to 360°C
Inlet 280°C, 1µL Injection, 5:1 Split
ChemStation, Agilent Technologies
AnalyzerPro, SpectralWorks Ltd
Results and Discussion
Figure 2. Paraffin Mix Retention Standard
Figure 3. 2R4F Underivatized Extract, Highest Sample Loading
Figure 4. 2R4F TMS Derivatized, Highest Sample Loading
Figure 5. Peaks Not Chromatographically Resolved
Figure 6. Example output for tocopherol shown
Typical data processing software supplied by instrument vendors is adequate for automated target compound analysis and routine quantitation. In fact, this software can process very complex data files surprisingly fast from large target compound lists once the list has been generated. But although these programs can be used for peak finding and cataloging of peak characteristics, they are not well suited for data mining of complex samples because they require a tremendous amount of user interaction to generate the target compound lists, refine spectra and export data. Also, processing from a fixed target list overlooks new components that may appear in test samples.
Data processing software designed for mining complex MS data can automate peak finding and characterization, as well as correctly format results for export to other programs such as spreadsheets and databases for parsing and user-customized processing.
Automated calculation of retention indices and refinement of component spectra were particularly useful for proper compound identification.