A data analysis approach for locomotor behavioral responses in developmentally exposed zebrafish

The larval zebrafish (Danio rerio) locomotor response light/ dark (LMR L/D) assay may inform developmental neurotoxicity (DNT) hazard, but a lack of standardized data analysis techniques has limited cross-laboratory data comparisons and integration. The present work proposes a concentration-response modeling workflow for evaluating chemicals tested in the LMR L/D assay, comprising 13 endpoints measuring different types of behavioral response (habituation, average speed, startle response, and range of activity). Data were previously collected for 61 chemicals tested in the LMR L/D assay, including 47 chemicals with evidence of in vivo DNT and 14 putative negatives. Response values were normalized using a Box-Cox transformation to address challenges such as inter-fish variability and non-normally distributed data. We found that the commonly studied endpoints in the LMR L/D assay (four in this study) were the most active overall; however, we identified ten chemicals that were only active in the less commonly studied endpoints (nine in this study), suggesting the potential added value of these endpoints for detecting behavioral changes in developing zebrafish. Importantly, we found that bidirectional curve-fitting and biphasic concentration-response models were informative for modeling zebrafish behavior, particularly for effects occurring at lower concentrations. This standardized concentration-response modeling workflow for the LMR L/D assay successfully detected chemical-induced changes in zebrafish behavior and contributes to a global effort to increase data transferability, reproducibility, and integration across LMR L/D labs and other DNT assay technologies.