Artificial intelligence assisted behavioral profiling of synthetic cannabinoids in planarians.

PURPOSE: Freshwater planarians provide a rapid and scalable biological model for detecting drug-induced neurobehavioral effects. This study evaluated whether automated behavioral profiling in Dugesia dorotocephala could distinguish phytocannabinoid and synthetic cannabinoid exposure based on organism-level motor responses. METHODS: Planarians were acutely exposed to Δ⁹-tetrahydrocannabinol (Δ⁹-THC), AB-PINACA, MA-CHMINACA, A-796,260, or JWH-412 at concentrations of 5–60 µg/mL in artificial spring water containing PEG-400. Locomotion and posture were recorded for 5 min and analyzed using LabGym, a supervised deep-learning–based behavioral classification system that quantified gliding, headshake activity, and sustained C-shaped postures. RESULTS: Distinct compound-associated behavioral profiles were observed. Δ⁹-THC and JWH-412 produced marked suppression of total locomotion relative to pooled controls. AB-PINACA and MA-CHMINACA preserved overall movement volume but produced severe disruption of coordinated gliding accompanied by frequent abnormal postural states. A-796,260 produced comparatively mild effects on locomotor organization. These findings revealed separable behavioral patterns that were further resolved in a two-dimensional state space by PCA (91.4% variance captured). These results demonstrate that cannabinoid exposures differing in pharmacological class produce separable behavioral patterns defined by both movement magnitude and organization. CONCLUSIONS: Automated planarian behavioral profiling provides a biologically grounded functional assay capable of distinguishing synthetic cannabinoids based on integrated motor signatures. This scalable non-vertebrate platform may support forensic toxicology by enabling early functional characterization of emerging synthetic cannabinoids and other novel psychoactive substances relevant to regulatory monitoring and public health surveillance.