Chemotyping: Classifying cannabis strains by chemical composition

An approach to optimizing the chemotyping of cannabis strains has been outlined by researchers at Ohio University and Chemistry Mapping Inc. Specifically, they evaluated the impact of several mass spectrometry (MS) data pre-processing techniques to identify which strategy helps to provide the most accurate and useful chemical fingerprint of cannabis samples. Their findings were published in Talanta. The research highlights the importance of carefully evaluating and selecting data pre-processing parameters.

A rapid method for characterizing botanicals 

From cultivators and breeders wanting to patent specific strains with specific chemical compositions, to clinicians demanding more information on the medical benefits of individual cannabis strains, an accurate view of the chemical profile of any given cannabis strain is of interest for many reasons.

Speaking to us about the story behind the group’s work, corresponding author, Peter Harrington, Professor of Chemistry, Ohio University, explained that the team was first recruited to work on characterizing botanicals using chemical profiling by the United States Department of Agriculture (USDA). Following the publication of many papers on characterizing ginsengs and black cohosh, they were enlisted by Chemistry Mapping Inc. to apply their techniques to cannabis products. Since then, the team have published several papers on cannabis. Dr Harrington highlighted two in particular relating to a high-throughput method of extracting plant material into deuterated chloroform and then characterizing it by nuclear magnetic resonance spectroscopy.

Prof. Harrington told us “The goal is to develop a quick method of measuring the chemical composition of cannabis, so we use spectroscopic methods to analyze extracts, and skip a chromatographic separation step that usually takes longer.  Instead of identifying and quantifying each component in the spectrum, the spectrum is treated as a fingerprint.  Using chemometrics and machine learning, we then can group the samples into classes based on their observed chemical composition.  We refer to this procedure as chemotyping.  The goal is to correlate these groups with desired pharmacological properties, so that industry can have some quality control over products and provide an avenue to achieve personalized medicine.”

Prof. Harrington explained “Our findings demonstrate that a chemotyping approach avoids the inherent pitfalls of genotyping and using plant morphology to identify and characterize cannabis products.  In addition, we have demonstrated that low-cost methods such as UV spectroscopy can work just as well as more expensive high-resolution methods of nuclear magnetic resonance spectroscopy and MS.

High-resolution mass spectrometry for cannabis chemotyping 

The groups most recent paper “Effect of preprocessing high-resolution mass spectra on the pattern recognition of Cannabis, hemp and liquor” looks at MS-based analysis rather than spectroscopy. High resolution MS (HRMS), was combined with pattern recognition to characterize twenty-five cannabis samples, twenty hemp samples and eight liquor samples. The team were keen to explore the use of HRMS as it offers a sensitive and selective option for detecting and quantifying plant metabolites down to low parts per million (ppm) levels. However, the challenge with this kind of MS is a product of its powerful sensitivity. The authors explained in the paper “The large dynamic range of the sensitive measurements may bias the classification model to favour large peaks at a low mass that may convey less characterizing information.” 

To overcome this challenge, the researchers set out to identify the best strategy for pre-processing their data and reduce the dynamic range. In brief, they found that proportional binning combined with square root transformation of the data yielded the best results. The study confirmed that “… data pre-processing methods such as different transformations, binning strategies, and resolving powers are important factors to be optimized for HRMS direct infusion measurements combined with pattern recognition to be an authentication and characterization tool for various products”. 

Moving toward personalized cannabis medicine

The team are now looking to partner with clinicians to improve our understanding of the medicinal utility of different cannabis strains. Their plan is to collect data regarding the pharmacological effects of cannabis on individual patients and attempt to correlate that data with their chemotyping results; potentially enabling specific strains with specific chemotypes to one day be prescribed for specific conditions. 

Prof. Harrington added “In addition, we seek to scale up our experiments so as to demonstrate that there exists a small set of canonical chemotype classes for the hundreds of cannabis cultivars.” This line of enquiry may help to address the long-standing debate around how many “types” of cannabis exist. Exact figures vary but, growers, cultivators and users have claimed there are many hundreds, if not thousands of different strains of cannabis. Grouping them by major phenotypic features via chemotyping may prove useful in ascertaining how strains are related, and the kind of medical benefits you can expect to see with a strain from a given chemotype class. 

Prof. Harrington concluded by saying “We hope to continue to work with existing partners at Ohio University, like Black Elk Biotechnology and Chemistry Mapping, to expand our reach and help harness the benefits of cannabis and other natural medicines.”