Abstract
Cannabis sativa L. is widely used as recreational illegal drugs. Illicit Cannabis profiling,
comparing seized samples, is challenging due to natural Cannabis heterogeneity. The aim of this
study was to use GC–FID and GC–MS herbal fingerprints for intra (within)- and inter (between)-
location variability evaluation. This study focused on finding an acceptable threshold to link seized
samples. Through Pearson correlation-coefficient calculations between intra-location samples, ‘linked’
thresholds were derived using 95% and 99% confidence limits. False negative (FN) and false positive
(FP) error rate calculations, aiming at obtaining the lowest possible FP value, were performed for
different data pre-treatments. Fingerprint-alignment parameters were optimized using Automated
Correlation-Optimized Warping (ACOW) or Design of Experiments (DoE), which presented similar
results. Hence, ACOW data, as reference, showed 54% and 65% FP values (95 and 99% confidence,
respectively). An additional fourth root normalization pre-treatment provided the best results for
both the GC–FID and GC–MS datasets. For GC–FID, which showed the best improved FP error
rate, 54 and 65% FP for the reference data decreased to 24 and 32%, respectively, after fourth root
transformation. Cross-validation showed FP values similar as the entire calibration set, indicating the
representativeness of the thresholds. A noteworthy improvement in discrimination between seized
Cannabis samples could be concluded.
comparing seized samples, is challenging due to natural Cannabis heterogeneity. The aim of this
study was to use GC–FID and GC–MS herbal fingerprints for intra (within)- and inter (between)-
location variability evaluation. This study focused on finding an acceptable threshold to link seized
samples. Through Pearson correlation-coefficient calculations between intra-location samples, ‘linked’
thresholds were derived using 95% and 99% confidence limits. False negative (FN) and false positive
(FP) error rate calculations, aiming at obtaining the lowest possible FP value, were performed for
different data pre-treatments. Fingerprint-alignment parameters were optimized using Automated
Correlation-Optimized Warping (ACOW) or Design of Experiments (DoE), which presented similar
results. Hence, ACOW data, as reference, showed 54% and 65% FP values (95 and 99% confidence,
respectively). An additional fourth root normalization pre-treatment provided the best results for
both the GC–FID and GC–MS datasets. For GC–FID, which showed the best improved FP error
rate, 54 and 65% FP for the reference data decreased to 24 and 32%, respectively, after fourth root
transformation. Cross-validation showed FP values similar as the entire calibration set, indicating the
representativeness of the thresholds. A noteworthy improvement in discrimination between seized
Cannabis samples could be concluded.
| Original language | English |
|---|---|
| Article number | 6643 |
| Number of pages | 17 |
| Journal | Molecules |
| Volume | 26 |
| Issue number | 21 |
| DOIs | |
| Publication status | Published - 2 Nov 2021 |
Bibliographical note
Funding Information:This project received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 833787, BorderSens. The Fund for Scientific Research (FWO), Vlaanderen, Belgium, also supported and financed this research under Grant agreement No. G033816N.
Publisher Copyright:
© 2021 by the authors. Licensee MDPI, Basel, Switzerland.
Keywords
- chromatographic fingerprint
- alignment optimization
- design of experiments
- data preprocessing
- comparison intra- and inter-location samples