Utilizing deep learning to analyze circadian rhythms, color preferences, and agonistic behaviors among color variants of the freshwater crayfish

The red swamp crayfish Procambarus clarkii is widely used as a model organism for studying aquatic ecosystems, with various behavioral assessment methods developed for this purpose. However, up to now, the assessment still lack of improvement, slow, and low throughput. Thus, this study aimed to investigate multiple behaviors of several color variants of this species by using deep learning approach tools with fully automation. In natural and artificial habitats, P. clarkii occurs in multiple color morphs, including white (W), blue (B), and red (R), which are also popular in the aquarium trade. This study optimized multiple deep learning tools: UMATracker, Idtracker.ai, and DeepLabCut to assess innate color preferences, circadian rhythms, and agonistic behaviors respectively in P. clarkii. With this optimized setup, we compared the behaviors of W, B, and R color variant. It displayed that B exhibited more pronounced behaviors compared to the W and R variants. In circadian rhythm activity, the B morph displayed a significantly higher average velocity than other color variants during both day and night cycles. In terms of color preference, no significant differences were found among the R, B, and W variants. Additionally, R, B, and W showed similar agonistic behaviors, displaying aggression in various ways, such as spreading claws and bracing against opponents. By integrating these machine and deep learning tools we can retrieve faster and reliable results.