Microsoft word - budapest2003_poster_g.doc

Evaluation of GC/MS Data
Systematic Toxicological Analysis
in Forensic Analyses by
A main objective of forensic toxicological analysis is the identification of general unknown poisons in biological material. A reproducible and automatic extraction procedure for the analysis of Chemometric Methods
drugs in post-mortem samples of blood and tissue has been developed [1]. Extracts are analyzed by GC/MS producing more than 2000 spectra per sample. The resulting total ion current (TIC) chromatograms contain many W. Demuth1*, T. Stimpfl 2
substances from the biological material (liver sample). The distinction ,
between negative and positive cases and identification of unknown W. Vycudilik2
1
, K. Varmuza
poisons is very time consuming using routine software. A successful identification depends on the availability of the 1 Vienna University of Technology,
corresponding mass spectra in the applied libraries. In this work the Institute of Chemical Engineering, Laboratory for Chemometrics,
MS library for toxicologically interesting substances by Pfleger, Getreidemarkt 9/166-2, A-1060 Vienna, Austria
Difficulties arise in cases, where comprehensive mass spectral data of Corresponding authors for chemometrics: the unknown compound are not available. A software was developed [email protected]
to filter the total number of acquired mass spectra in general unknown [email protected]
cases and to simplify the TIC chromatogram by an automated,
2 University of Vienna, Institute of Forensic Medicine,
computer-assisted subtraction procedure. Sensengasse 2, A-1090 Vienna, Austria
This procedure allows the forensic toxicologist to focus on mass Corresponding authors for forensic chemistry: spectra of uncommon compounds, and therefore improves the chance [email protected]
for the identification of a “general unknown” in complex biological [email protected]
* Presenting author: W. Demuth
The Software FORGE (Forensic GC/MS-Data Exploration) was
realized in MATLAB to test new algorithms. Development of a Austrian Science Fund, project P14792-CHE
Stimpfl T., Jurenitsch J., Vycudilik W.: J. Anal. Toxicol. 25, 125 (2001)
Pfleger K., Maurer H.H., Weber A.W.: Mass spectral and GC data of drugs, poisons, pesticides, pollutants and their metabolites; 2nd ed., VCH Publishers, Poster Presentation: Advances in Chromatography and Electrophoresis -
Conferentia Chemometrica (ACE & CC 2003)
Software FORGE
Selection of Characteristic Spectra
Forensic GC/MS-Data Exploration
GC/MS analyses of a blood sample spiked with standard mixture of
toxins. Selected range of TIC after baseline subtraction: Baseline Subtraction
Performed separately for each mass chromatogram. Selection of Characteristic Spectra
Automatic selection of approx. 50 mass spectra from GC/MS data. Retention Index Calibration
Linear or quadratic regression model using alkanes and relevant Selection Method: Selection of characteristic spectra by searching
Tools for Building Negative Libraries
for local maxima in all mass chromatograms. Characteristic spectra of negative reference samples (no Two positive slopes followed by two negative slopes with an intoxication) are collected to eliminat unsuspicious spectra. intensity above a threshold are considered a maximum in a mass chromatogram. A characteristic spectrum must contain maxima in at Subtraction of Spectra present in Negative Library
Spectra of matrix components are filtered to clean the TIC and to Forward Search
Identification of compounds with spectra selected. Target Search
Search for maximal occurrence of selected target spectra in chromatogram with optionally considering retention indices. Calculation of Spectral Features
Various features can be generated from mass spectral data to obtain suitable variables for analysis of spectra-structure Exploratory Data Analysis of Spectra or Features
Selected mass spectra 3 and 4 are very similar because of an overlap Spectral features or MS peak intensities are used for exploratory of two substances. The developed method allows the detection of Identification of Spectra
PCA of Characteristic Spectra
A “foreward search” is used to identify the selected characteristic spectra. The search algorithm tries to fit the library spectrum into the unknown/
spectrum of the unknown. Only small intensities of the library impurities
spectrum are allowed not to be present in the unknown. Higher intensities in the spectrum of the unknown are tolerated (caused by other toxins
Results of Identification for all selected characteristic mass spectra:
All added toxins where identified correctly (spiked sample!) components
PCA: 42 objects (characteristic spectra) 862 features
This representation shows the structural similarity of found compounds derived from mass spectral features. The added Such PCA plots can be used as templates for interpretation of AC …. acetylated (sample preparation) Heroin found in real forensic blood sample.

Source: http://www.lcm.tuwien.ac.at/vk/Manus/Poster-131-Budapest-2003-GCMS-tox.pdf