(P26-027-25) Identification of Bioactive Minor Cannabinoids From Cannabidiol-depleted Hemp Extract by Using Machine Learning

Objectives: Hemp (Cannabis sativa L.) contains more than 100 cannabinoids, but most of these cannabinoids go to waste in the CBD industry. Minor cannabinoids refer to all cannabinoids other than cannabidiol (CBD) and tetrahydrocannabinol (THC). Recently, there has been growing interest in the therapeutic benefits of minor cannabinoids, such as their effects on low-grade inflammation and pain. Thus, the objective of this study was to predict bioactive minor cannabinoids from CBD-depleted hemp extract by investigating their effects on low-grade inflammation and integrating a machine-learning approach.

Methods: Minor cannabinoids in the hemp extract were identified using liquid chromatography with high-resolution hybrid quadrupole-time-of-flight mass spectrometry. The hemp extract was fractionated into 47 fractions using a Sephadex LH-20 column. Fractions were analyzed by loop-injection combined with electrospray ionization and high-resolution accurate mass spectrometry (LI-ESI-TOF). RAW264.7 murine macrophage cells were treated with fractions (1 μg/ml) and lipopolysaccharide (LPS, 100 ng/ml) for 24 h to assess their effects on inducible nitric oxide synthase (iNOS)-mediated nitric oxide production. Based on the mass peak intensities of fractions from LI-ESI-TOF and bioassay results, a random forest model with R packages randomForest and caret was used to predict which minor cannabinoids contribute to the bioactivity.

Results: A total of 16 cannabinoids were identified by comparing them with our in-hand library of 35 standards. While LPS treatment significantly increased iNOS-mediated nitric oxide production compared to the vehicle control, LPS-stimulated nitric oxide production was significantly inhibited by fractions #20-45 (p < 0.05). Eight peaks among 63 peaks showed importance from random forests analysis, with cannabinoid acids showing the most importance.

Conclusions: This study demonstrates the utility of machine learning to accelerate the identification of bioactive cannabinoids. The discovery of bioactive minor cannabinoids from waste products of the CBD industry will benefit both hemp-based phytochemical supplement development and a sustainable agro-industry.

Funding Sources: This research was supported by National Center for Complementary and Integrative Health.