Machine Learning-Guided Discovery of Cannabidiolic Acid Ethyl Ester as a Bioactive Anti-Inflammatory Cannabinoid from CBD-Depleted Hemp Extract

Abstract Industrial processing of hemp (Cannabis sativa L.) produces side-streams, such as CBD-depleted “mother liquor” (CBD-ML), that are rich in structurally diverse but underexplored minor cannabinoids. To accelerate the discovery of bioactive cannabinoids within this complex matrix, we applied a computational pipeline integrating loop-injection high-resolution time-of-flight mass spectrometry (HR-TOF MS) chemical profiling with anti-inflammatory bioassay data. Using an ensemble of random forest models, we correlated mass spectrometric features with the inhibition of nitric oxide (NO) production in LPS-stimulated RAW264.7 macrophages. This predictive model identified an ion at m/z 357.2071 as a top predictor of bioactivity. Targeted isolation and structural elucidation by HRMS and NMR identified the bioactive compound as cannabidiolic acid ethyl ester (CBDA-EE), which likely arises as a byproduct during industrial processing of hemp. Validation studies confirmed that CBDA-EE modulates oxylipin production in stimulated macrophage cells, significantly reducing proinflammatory prostaglandins, including prostaglandin E2, while maintaining levels of prostaglandin D2 (PGD2) and its downstream metabolite 15-deoxy-PGD2. These findings demonstrate the utility of machine learning in streamlining the discovery of unstudied bioactive natural products from complex botanical mixtures.