Classification of the Seized Cannabis Seeds Cultivated in Greenhouse by Age through Near Infrared Spectroscopy associated with Chemometrics

Cannabis, popularly called as marijuana, is still the most frequently used illicit substance in the world and the cultivation of its herb can be performed in a wide range of environments. In the last years, the cannabis dependence has increase in Brazil as well the number of seized marijuana by Brazilian Federal Police. Cannabis seeds sent in small amounts in post offices have been the new form of drug trafficking in Brazil. The conditions of cultivation of indoor cannabis, such as time of harvesting, can induce the formation of cannabinoids or breakdown products. The aim of this study is evaluate the growning time of indoor cannabis of seized seeds by Brazilian Federal Police by near infrared spectroscopy associated with chemometrics. Experimental Part: Twenty nine seized seeds were cultivated in a homemade greenhouse, at 30°C, for three different periods, on soil consisting of a 1:1:2 fertilized soil (NPK 8-12-10), worm humus and sand, daily irrigation and the light/dark photoperiod was hold for 18 and 6 hours, respectively, using only artificial lighting. The harvest was performed by removing the entire plant from the soil. The dried aerial plant samples (leaves, stems and inflorescence) were grinded in agate mortar and directly analyzed by near infrared spectrometer in triplicate. The spectra of the cannabis were smoothed by the Savitzky-Golay algorithm (1st order polynomial, 15 points per window), normalized by multiplicative signal correction (MSC), derived at first order by the Savitzky-Golay algorithm (2nd order polynomial, 15 points per window) and mean-centered. The principal component analysis (PCA), partial least square discriminant analysis (PLS-DA) and support vector machines discriminant analysis (SVM-DA) were used. The Kennard-Stone algorithm was used to separate the training (two-thirds) and test (one third) samples. The 29 samples were classified by age (5.5 weeks, 7.5 weeks and 10 weeks samples). The software Matlab with PLS_toolbox was used. Results and Discussion: The spectra of the cannabis were divided in 8, 16 and 32 interval with equal number of variables and the PCA was carried out in each interval since the global PCA did not show a tendency of separation among the classes. It was observed that the absorption bands of CH, CH2 and CH3 combinations (4200 to 4600 cm-1), first and second overtones (5600 to 6000 cm-1 and 6950 to 7150 cm-1, respectively) are the intervals that showed a separation by age of the plant. However, the separation obtained by the absorption bands of CH, CH2 and CH3 combinations (4200 to 4600 cm-1) was better. The both discrimination analysis algorithms were carried out in this interval. The PLS-DA showed a worse classification rate, in terms of sensitivity and specificity, in comparison with SVM-DA, that classified correctly all training and test samples. Conclusion: These results showed that, in the early stages of the cultivation of indoor cannabis, the age of the plant can be predicted by near infrared spectroscopy. However, a larger study may be conduct to reiterate this tendency due to low number of samples obtained.