New Method Offers High-Throughput, Cost-Effective Chemotyping of Cannabis
Researchers at the Center for Intelligent Chemical Instrumentation in Ohio University have proven the effectiveness of ultraviolet microplate readers with multivariate classifiers as a practical and accurate way of chemotyping different strains of cannabis and hemp plant material. It is thought that the method could also be effectively used for classifying a variety of other botanical extracts from other plant species.
The need for accurate chemotyping
A chemotype, or chemical genotype, generally refers to a subspecies of a plant. Chemotypes belong to the same genus and species but will have observably different chemical composition, usually with respect to the presence of secondary small molecule metabolites. For example, the United Nations Office on Drugs and Crime recognizes three different chemotypes of the species Cannabis sativa, differentiated by the proportion of the cannabinoids tetrahydrocannabinol (THC) and cannabinol (CBN), relative to cannabidiol (CBD). This difference in metabolite expression in varying chemotypes can often lead to differing pharmacological properties and effects when the cannabis is used.
By using chemical profiling methods it is possible to study the chemotype of a sample, and therefore, its potential pharmacological or therapeutic properties. Being able to quickly and cost-effectively profile samples belonging to different chemotypes could have potentially helpful applications in identifying the best botanical extracts and materials for a specific therapeutic use. Additionally, in some cases forensic chemotyping can assist in positively identifying the origin of a plant sample, which could make it possible to identify batches of cannabis material that have been illegally smuggled across geographic regions.
Previously, the chemotyping of cannabis and hemp plants has been done using high-performance liquid chromatography (HPLC), nuclear magnetic resonance spectrometry (NMR) and high-resolution mass spectrometry (HRMS). All of these methods require advanced and specialized knowledge of the method of analysis, as well as time-intensive sample preparation steps that must be taken before sample analysis is possible. The methodology published by the Ohio State researchers for chemotyping using ultraviolet microplate readers offers significant advances over HPLC, NMR, and MS techniques in terms of both speed and cost-effectiveness.
Running a spectral measurement
Using 15 unique cannabis extracts and 20 unique hemp extracts (supplied by Colorado-based Chemical Mapping Inc.) and predetermined amounts of chloroform (CDCl3), researchers created two different homogenized dilutions of each extract, at an extract/solvent ratio of 1-10 and 1-20 respectively. These dilutions, along with a blank for standardization, were then loaded into the microwells of a microplate reader and ultraviolet spectra were recorded for each sample. The wavelength range studied for the cannabis extracts was between 260-400 mn, and for hemp 260-460 nm. The lower limit was necessary due to the cut-off wavelength of the chloroform solvent, and the upper limit was elected in order to minimize the run time of the analysis, as the absorption spectra above 400 nm or 460 nm were broad and featureless for the respective extracts. Each extract was characterized five times over five successive days to ensure reliability.
In general, UV spectra contain less detail than the likes of HPLC, NMR, or MS, but characteristic information can be seen by overlapping absorbance bands when the correct mathematical methods are applied to the data. In this methodology, a range of different multivariate classifiers were used to determine the chemotype of each sample, namely: Fuzzy Rule-Building Expert System (FuRES); Super Partial Least Squares-Discriminant Analysis (sPLS-DA); plus one unmodified and two modified variants of Support Vector Machine (SVM) algorithms, known as SVM, SVMtreeG and SVMtreeH respectively. Respectively, these are examples of a minimal neural network, a self-optimized partial least squares regression technique, and a form of supervised machine learning.
Bootstrapped Latin partitions — the name given to the splitting of data into smaller subsets which maintain a similar class distribution to the wider data while still being random — were used as validation sets to measure the average classification accuracy of each multivariate classifier.
UV microplate readers as a chemotyping method
For the chemotyping of cannabis extracts, SVMtreeG and SVMtreeH were identified as the most accurate classifiers, with both recording a classification accuracy of 99.1±0.4% for the 1-10 dilutions. At a dilution ratio of 1-20, the SVMtreeG classifier remained the best approach, with a classification accuracy of 97.1±0.3%. Across all classifiers, the 1-10 dilutions gave the best classification accuracy. With respect to the hemp extract samples, the unmodified SVM algorithm was the most accurate, with a classification accuracy of 97.4±0.6, followed by the SVMtreeG and FuRES classifiers.
The total cumulative time needed to run the 75 spectra measurements for the 15 cannabis extracts was around 3 hours. For the 20 hemp extracts, which required 100 spectra to be collected, the analysis took around 4 hours. For comparison, a single NMR analysis run for chemotyping takes around 30 minutes, meaning it would take around 37-50 hours to do the same number of cannabis and hemp extract analyses done here using NMR apparatus.
These classification accuracies indicate that using a UV microplate reader coupled with multivariate classifiers is indeed a good method for chemotyping extracts of cannabis and hemp products, though in theory the method should be effective for other botanical extracts as well. The short sample run-time and minimal sample prep should help to make the process a very attractive technique for chemotyping experiments which require a large volume of samples to be run.