The development of FEDUPP: feeding experimentation device users processing package to assess learning and cognitive flexibility.

Cognitive flexibility, the ability to adapt behavior in response to changing contingencies, is a key component of adaptive decision-making and is impaired in multiple neuropsychiatric disorders. Traditional rodent assays of cognitive flexibility are conducted in experimenter-controlled sessions in restrictive environments, limiting ecological validity and temporal resolution. Here, we developed a fully automated, home-cage paradigm using the Feeding Experimentation Device 3 (FED3) and a companion open-source analysis pipeline, the Feeding Experimentation Device Users Processing Package (FEDUPP), to assess learning and cognitive flexibility with minimal experimenter intervention. The paradigm combines a single-day fixed-ratio 1 (FR1) task with a multi-day, reversal learning task in which active port assignment switches every 25 pellets collected. FEDUPP implements multi-scale learning metrics, including overall accuracy, an 80% accuracy milestone, and a machine learning-based classification of meal accuracy to capture motivated, goal-directed feeding. In wild-type mice, the paradigm detected rapid FR1 acquisition and progressive within-block adaptation during reversal. Application to mice with dorsal hippocampal knockdown of the scaffolding protein CASK revealed faster FR1 acquisition and higher accuracy. In addition, a faster onset of the first accurate meal after reversal suggests an improvement in updating goal-directed feeding behavior. These findings demonstrate that FEDUPP enables high-resolution, continuous assessment of learning and cognitive flexibility in ethologically relevant settings, and that meal-based accuracy provides a sensitive metric for detecting subtle changes in flexibility not captured by traditional measures.