Total Ion Spectra versus Segmented Total Ion Spectra as Preprocessing Tools for Gas Chromatography – Mass Spectrometry Data |
| |
| Authors: | Lawrence A. Adutwum Ph.D. Robin J. Abel MSc. James Harynuk Ph.D. |
| |
| Affiliation: | Department of Chemistry, Univeristy of Alberta, Edmonton, Alberta, Canada |
| |
| Abstract: | Alignment of fire debris data from GC‐MS for chemometric analysis is challenged by highly variable, uncontrolled sample and matrix composition. The total ion spectrum (TIS) obviates the need for alignment but loses all separation information. We introduce the segmented total ion spectrum (STIS), which retains the advantages of TIS while retaining some retention information. We compare the performance of STIS with TIS for the classification of casework fire debris samples. TIS and STIS achieve good model prediction accuracies of 96% and 98%, respectively. Baseline removal improved model prediction accuracies for both TIS and STIS to 97% and 99%, respectively. The importance of maintaining some chromatographic information to aid in deciphering the underlying chemistry of the results and reasons for false positive/negative results was also examined. |
| |
| Keywords: | forensic science fire debris analysis arson gas chromatography‐mass spectrometry chemometrics variable selection cluster resolution partial least squares— discriminant analysis total ion spectrum segmented total ion spectrum |
|
|
