BEEhaviourLab: A high-throughput platform for sublethal stressor screening in insects

Chemical risk assessment for insects relies largely on mortality endpoints in a few model species, limiting detection of ecologically relevant sublethal effects and cross-taxon comparisons. Behaviour is a sensitive indicator of neurotoxic stress, but scalable, standardised measurement has remained challenging. We present BEEhaviourLab, a low-cost, automated platform for high-throughput behavioural phenotyping. The system integrates parallel video and audio recording with computer vision tracking, acoustic classification, experimental control, and automated analyses, enabling long-duration experiments across many replicates and species. A single lightweight object-detection model tracks multiple untagged insects simultaneously, including individuals from different taxa (e.g., bees and hoverflies). Using this platform, we quantify species-specific circadian activity patterns and assess toxicity of the widely used veterinary pesticide moxidectin in the bumble bee (Bombus terrestris). Acute contact exposure caused dose-dependent reductions in locomotion and buzzing at sublethal concentrations, with acoustic measures more sensitive than video. These results demonstrate that scalable, multimodal behavioural phenotyping can detect biologically meaningful neurotoxic disruption well below lethal thresholds, providing a practical path to integrate sublethal endpoints into chemical risk assessment.