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

LAURA ALETHIA DE LA FUENTE; ENZO TAGLIAZUCCHI; FEDERICO ZAMBERLAN; ANDRÉS SÁNCHEZ FERRÁN; FACUNDO CARRILLO; CARLA PALLAVICINI

Congreso; XXXIV Reunión Anual SAN 2019; 2019

Commercialcannabis strains have multiplied in recent years dueto regional changes inlegislation for medicinal and recreational use. The lack ofstandardized systemsto label plants and seeds hinders the consistent identification ofstrains with elicited psychoactive effects.We analyzed a publicly available dataset where users reportedexperienceswith different strains, including effects and flavor tags.Metrics of strain similarity based on thesetags allowed machine learning classification into three majorclusters associated with species (sativa,indica,and hybrids). Synergy between terpene and cannabinoid content wassuggested by significantcorrelations between psychoactive effects and flavortags. The tags werevalidated by the application of Latent Semantic Analysis (LSA) tounstructured reviews, also providing breed-specific topics consistentwith their medicinal and subjective effects. While cannabinoidcontent was variable even within individual strains, terpene profilesmatched perceptual characterizations made by users and could be usedto predict psychoactive effects. Our work represents the firstdata-driven synthesis of self-reported and chemical information on alarge number of cannabis strains. Since terpene content is robustlyinherited and less influenced by environmental factors, flavorperception could represent a reliable marker for the prediction ofpsychoactive effects of cannabis. p { margin-bottom: 0.1in; line-height: 120%; }