Abstract
Brain function studies greatly depend on quantification and analysis of behavior. While behavior can be imaged efficiently, the quantification of specific aspects of behavior is labor-intensive and may introduce individual biases. Recent advances in deep learning and artificial intelligence-based tools have made it possible to precisely track individual features of freely moving animals in diverse environments without any markers. In the current study, we developed Zebrafish Larvae Position Tracker (Z-LaP Tracker), a modification of the markerless position estimation software DeepLabCut, to quantify zebrafish larval behavior in a high-throughput 384-well setting. We utilized the high-contrast features of our model animal, zebrafish larvae, including the eyes and the yolk for our behavioral analysis. Using this experimental setup, we quantified relevant behaviors with similar accuracy to the analysis performed by humans. The changes in behavior were organized in behavioral profiles, which were examined by K-means and hierarchical cluster analysis. Calcineurin inhibitors exhibited a distinct behavioral profile characterized by increased activity, acoustic hyperexcitability, reduced visually guided behaviors, and reduced habituation to acoustic stimuli. The developed methodologies were used to identify 'CsA-type' drugs that might be promising candidates for the prevention and treatment of neurological disorders.