Abstract
Phytocannabinoids are a diverse class of bioactive compounds produced by Cannabis sativa, including both major and a growing number of minor constituents with pharmacological relevance. However, their comprehensive annotation in untargeted high-resolution mass spectrometry (HRMS) data sets remains a significant analytical challenge due to their structural similarity, low abundance, and the complexity of plant matrices. In this study, we present a comparative evaluation of Kendrick Mass Defect (KMD)-based filtering workflows for the efficient untargeted annotation of minor phytocannabinoids. Three data processing strategies were implemented using Compound Discoverer: (i) KMD filtering before the "Compound Detection" tool, (ii) KMD filtering after the "Compound Detection" tool, and (iii) a pseudo-KMD approach based on the generation of expected compounds. These workflows were tested and compared using a data set comprising 50 Cannabis inflorescence samples analyzed in an untargeted fashion, taking into account the phytocannabinoid coverage, false positive rates, computation burden, and versatility. A total of 61 phytocannabinoids were annotated, including a full series of alkyl homologues (C1-C7), cis/trans isomers, O-methylated derivatives, and sesquicannabinoids. Statistical analyses revealed meaningful chemical differentiation based on seed origin, chemovar classification, and reproductive strategy (dioecious vs monoecious), highlighting the biological significance of minor cannabinoids. Overall, the results demonstrate that KMD filtering significantly enhances the throughput and accuracy of untargeted HRMS workflows for structurally related classes of compounds.