Over eight hundred cannabis strains characterized by the relationship between their psychoactive effects, perceptual profiles, and chemical compositions

Summary

Machine learning analysis of a large public dataset where users freely reported their experiences with cannabis strains, combined with chemical composition data, reveals that cannabis strains can be reliably classified into three major clusters corresponding to Cannabis sativa, Cannabis indica, and hybrids based on self-reported effect and flavor tags. Terpene profiles matched users' perceptual characterizations and could predict associations between different psychoactive effects, while cannabinoid content was variable even within individual strains. The findings suggest that flavor perception, reflecting robustly inherited terpene content, could serve as a reliable marker to predict psychoactive effects, offering a data-driven approach to strain classification for the growing medicinal and recreational cannabis market.

Study at a glance

Abstract

Abstract Background Commercially available cannabis strains have multiplied in recent years as a consequence of regional changes in legislation for medicinal and recreational use. Lack of a standardized system to label plants and seeds hinders the consistent identification of particular strains with their elicited psychoactive effects. The objective of this work was to leverage information extracted from large databases to improve the identification and characterization of cannabis strains. Methods We analyzed a large publicly available dataset where users freely reported their experiences with cannabis strains, including different subjective effects and flavour associations. This analysis was complemented with information on the chemical composition of a subset of the strains. Both supervised and unsupervised machine learning algorithms were applied to classify strains based on self-reported and objective features. Results Metrics of strain similarity based on self-reported effect and flavour tags allowed machine learning classification into three major clusters corresponding to Cannabis sativa , Cannabis indica , and hybrids. Synergy between terpene and cannabinoid content was suggested by significative correlations between psychoactive effect and flavour tags. The use of predefined tags was validated by applying semantic analysis tools to unstructured written reviews, also providing breed-specific topics consistent with their purported medicinal and subjective effects. While cannabinoid content was variable even within individual strains, terpene profiles matched the perceptual characterizations made by the users and could be used to predict associations between different psychoactive effects. Conclusions Our work represents the first data-driven synthesis of self-reported and chemical information in a large number of cannabis strains. Since terpene content is robustly inherited and less influenced by environmental factors, flavour perception could represent a reliable marker to predict the psychoactive effects of cannabis. Our novel methodology contributes to meet the demands for reliable strain classification and characterization in the context of an ever-growing market for medicinal and recreational cannabis.