Application of Chemometric Tools on Cannabis Samples Analyzed by the FTIR-ATR Method

Marijuana is the most popular form of Cannabis sativa L. (Cannabaceae), popularly known, in Brazil, as the illicit drug. It is composed of the plant’s aerial parts, such as the leaves and the inflorescences, which are dried, pressed and prepared as a mixture for smoking. Cannabis is the most consumed and illegally trafficked drug in the world, with an increasing number of users every year. The plant can be grown indoor and outdoor, and these differences may influence the drug’s potency. In addition, marijuana can be mixed with diluents and/or adulterants such as aromatic plants, soil, commercial tobacco and feces that may contribute to cases of addiction and lead to serious health risks to its consumers. Studies involving the chemical profile of drug samples are important to provide evidence for trafficking, supporting the materiality of the crimes. The aim of this study is to analyze cannabis and marijuana seized samples by FTIR-ATR (range 1800-880 cm), combined with unsupervised chemometric tools, to differentiate the plant’s cultivation forms and to suggest the use of diluents. PCA and HCA showed relevant trends of separation between seized 478 Brazilian Journal of Forensic Sciences, Medical Law and Bioethics 9(4):477-498 (2020) M. González et al. samples from indoor and outdoor cultivation. Additionally, differences between samples containing pure cannabis and samples including diluents were observed, grouping the samples by their chemical similarity. The use of FTIR-ATR, combined with chemometric tools, can generate fast and sensitive data, providing relevant information for chemical profiles of drug abuse.


Introduction
Marijuana is composed of the aerial parts of the plant Cannabis sativa L., such as the leaves and the inflorescences, which are dried, pressed and prepared as a mixture for smoking cigarettes, pipes and/or hookahs 1  Furthermore, cannabis is a chemically complex plant with a diversity of compounds including flavonoids, mono and sesquiterpenoids, steroids, nitrogenous compounds and cannabinoids, a characteristic class of terpenophenols for the plant 7,8 . Over the years, cannabis has undergone enhancement of genetic and cultivation techniques, allowing the increase of its psychotropic cannabinoid, Δ9-Tetrahydrocannabinol (THC), along with its modulator CBD. The higher content of THC, and the addition of diluents and/or adulterants, frequently found in marijuana, may contribute to cannabis addiction 7 .
Cole et al. in a review study of illicit drugs, the addition of diluents and / or adulterants was disregarded in clinical and forensic toxicology studies; the effects caused by these substances being ignored in the face of the effects of drugs 8 .
Adulterants and diluents are deliberately added to increase bulk, enhance or mimic a pharmacological effect, or to facilitate drug delivery 8 .
For marijuana, the main diluents added are: aromatic plants because they have a strong odor; soil from roots during cultivation and from an inadequate storage location, as accidental contamination; animal feces such as cows and horses, as the waste comes from a herbivorous diet and that dry visually resemble marijuana, being added to make the drug bulky; syrupy liquids like molasses, because cannabis has an oily extract to give it a dense and compact appearance for sale; aluminium for unknown reason, may have resulted from impure water supply and glass powder also for unknown reason, but potentially to improve apparent quality and increase weight 8,9,10 . Sometimes it is possible to use adulterants, like Tobacco, used to increase the volume and the addiction, due to nicotine presence [9][10][11] . Other adulterants come from the form of cannabis cultivation, such as pesticides or fungus that develop by a natural biological process and by the poor storage conditions, and can cause damage to health 9 , but are not the direct objective of this study now.
The drug trafficking industry has become professionalized, requiring new analytical methodologies, capable of identifying and tracing its origin by the police force and the forensic scientists 12 . Therefore, the use of chemical profile studies are tools that could assist cannabis identification, providing sensitive data for tracking and grouping of seized samples, that can be used as evidence of trafficking, proving the materiality of the crime 12,13 . Infrared spectroscopy (IR) is a reliable methodology for detecting fingerprint regions of different compounds and can be widely used to analyze any sample that has organic functional groups (CH, NH, SH and OH) 14 . The FTIR-ATR methodology relies on the Fourier Transform, a mathematical operation that, through software, separates the frequencies of individual absorptions contained in the sample interferogram, also subtracting the background interferogram that is made from atmospheric gases active in IR (carbon dioxide and water vapor), producing a spectrum identical to that obtained by a dispersive spectrometer 15 . This is a desirable technique for this kind of research, since it requires only a small quantity of samples for the analysis and it is a non-destructive method, allowing sample re-processing if necessary. In addition, it has a faster sample preparation and a low cost, compared to other available methodologies 16 . These advantages fit the reality of different police forces throughout various regions of the world. Combined with exploratory data analysis, FTIR-ATR becomes a powerful tool in Forensic Science 10,14 . This method is already consolidated for cocaine analysis and its adulterants, for medicines falsification assessment and for adulteration of documents 17 , for example. Chemometrics is the application of statistical algorithms to chemical data. Chemometric algorithms have the advantage of tolerating overlapping peaks, so the models do not need to include the concentration of every chemical species present, and multiple analytes can be easily determined 18 . Using principal components analysis (PCA) and hierarchical cluster analysis (HCA) for data related to cannabis and marijuana samples it was possible to create, groups and/or isolates samples by criteria of their chemical and sectoral similarity can be identified 18 . Thus, it is possible to differentiate samples, and to infer the use of diluents and/or adulterants for drug yield increase.
FTIR-ATR presents a challenge for the infrared spectrum interpretation.
Unique sample information is in the fingerprint region, which is in approximately 700 -1800 cm -1 . In this region, it is possible to compare the spectrum of a standard sample to the questioned sample, which allows the sample's identity confirmation 19 . In this study, all cannabis and marijuana samples analyzed did not show a significant signal in the 2000-1800 cm -1 range and the range of 3600 -2800 cm-1 was also not used, as the angular stretches of aromatic groups are observed in the fingerprint region, so this region of aromatic overtones was not included for the multivariate analysis. As a guide for identifying the main molecular clusters present in the samples in this study, we follow the Lopes and Fascio 19 scheme.
Thus, this study combines instrumental analysis methodologies from FTIR -ATR with exploratory tools to perform the analysis of marijuana and cannabis samples seized by the BFP with the addition of the following diluents: basil, cilantro, oregano, horse feces, soil and commercial tobacco. These diluents were chosen following the casuistry of the forensic institutes in the country and taking into account available and inexpensive materials that when added to cannabis had volume and aspect similar to marijuana without making the final product more expensive, aiming to mimic an adulterated real sample. This is a pioneering study for cannabis analysis that aims to pave the way for studies with adulterated real samples.

Seized samples and diluents
The cannabis and marijuana samples were provided by the BFP. All the research was observed by a federal criminal expert. Samples were separated into groups according to the region where they were seized and/or its geographic location.
Twenty-nine cannabis samples were provided from previous research on seed trafficking 16,20 . The seeds were grown in indoor way at the BFP station in Porto Alegre, Rio Grande do Sul, Brazil (30° 2' 53.30" S 51°12' 54.26" W), with authorization from the judiciary authorities 16,20 . Ten samples were seized in Manaus, Amazonas, Brazil (3° 7' 50"S 60° 1' 23" W) and they were sent to FPD in Rio Branco,  Table 1 shows all the information from the seized samples.  (Table 2).

Sample preparation
All the cannabis, marijuana and diluents samples were prepared using the same protocol. The samples were dried with heat at 60 °C for one hour in an oven (Biomatic®), and crushed using a hand crusher, followed by homogenization with a gral and a pistil. Then, they were sifted using a tamper and packed in a 1.5 ml plastic tube. Diluents samples were mixed according to the classes described in Table 2, with no distinction by origin.

Instrumentation
The infrared spectra of all seized samples and diluents were obtained in a Thermo Fisher Nicolet Avatar 370 DTGS Infrared Spectrometer (Thermo Fischer, San Diego, CA, USA) using a universal attenuated total reflectance (ATR) sampling accessory.
Absorbance was measured in the spectral range of 4000 -400 cm −1 , but the region of analysis chosen was 1800 -700 cm -1 , corresponding to the fingerprint region, in consensus, the authors observed that there is no relevant information in the regions of overtones and other expressive information such as aromatic groups can be seen in the fingerprint region 21,22. The ground samples were directly analyzed. The spectra were acquired at random, in triplicate, with 32 scans and resolution of 4 cm −1 .

Chemometrics
ChemoStat ® software was used for exploratory data analysis 23 . To test repeatability, all cannabis, marijuana and diluent samples analyses was performed in triplicate and its average spectra was used. The analysis range was set at 1800 -880 cm −1 for it is a relevant region with less noise interference. The data was preprocessed using the Savitzky-Golay (SVG) algorithm (1st order polynomial, 13 points per window) and normalization, using the Chemostat ® software. The standard normal variate (SNV) was applied to the spectra to remove vertical shifts, before exploratory analysis, and the spectra were mean-centered. PCA and HCA analysis were performed in the preprocessed spectra to investigate the similarities between the samples.

Analysis script
Exploratory data analysis was separated into stages, in order to know the differences

Results and discussion
Plant materials like Cannabis sativa L. are rich in lignocellulosic biomass that is an abundant renewable resource that can provide biopolymers, fibers, chemicals and energy 24 . The tricky part of applying Beer's Law to cannabis analysis is that it contains many different molecules, and it is not always possible to find an infrared peak that is solely due to a specific analyte 25 . Factors that may alter the infrared spectrum in plant samples are: (i) different soil compositions; (ii) differences in harvest time; (iii) use of nitrogen fertilizers 22,24 . Figure 3 shows a cannabis FTIR-ATR spectrum after range selection of 1800 -800 cm -1 .   However, in solid analyzes, the use of triplicates is recommended, since each sample may have a different behavior depending on the discussed conditions. These factors may change the response to analysis. When applying the pretreatment, it is possible to visualize the spectral signals uniformly (Figure 4b).  Table 3 shows a summary of the main assignments of the signals.       Adding diluent to the analysis was an attempt to differentiate marijuana samples that did not clearly separate in the initial analysis ( Figure 9). First, the diluents were analyzed without any marijuana sample, and a pool of each diluent was analyzed in triplicates using data preprocessing as it was used for cannabis samples. In order to analyze all samples (seized drugs, mixtures and diluents) it was necessary to average the results triplicate for each sample, before the pretreatment.
Due to the number of samples and their similarity, the results overlapped, and a clear identification of the groups was not observed (Figure 11a).
Considering all sample set, some trends are disclosed, such as: 1) PC1 showed 89.68% of variance and clearly separates soil and cannabis mixed with soil samples from the others, as shown in the loadings graph (Figure 11b and 11c).

Conclusions
A methodology of separation for cannabis plants and marijuana, with additional analysis of possible diluents present in the seized samples, was proposed by performing FTIR-ATR analysis and exploratory data analysis. Unsupervised methods were used because of the amount of samples available; the supervised methods were not used because they depend on a set of samples for the construction of the analytical models and another set of samples for external validation, which was not the case. The unsupervisioned methods of PCA and HCA showed that it is possible to separate cannabis and marijuana samples and to differentiate indoor culture from outdoor samples. Diluent analysis disclosed that contamination of seized marijuana samples is possible. FTIR-ATR methodology is quick and easy to apply, it requires a small sample volume and, mainly, it preserves the sample from destruction, an important detail in Forensic Sciences. Chemometrics, obtained through infrared analysis, were reliable and satisfactory, gathering several relevant pieces of information for the chemical profile of cannabis, in a short period of time.