Predicting aspects of feeding behavior.

A. Temporal dynamics of feeding behavior characterized on a population level across the four groups: baseline binge-like feeding of palatable chow (Base); binge-like feeding of palatable chow after 24-hours of food deprivation (Dep24) and 48-hours of food deprivation (Dep48); and limited access to house chow after 24-hours of food deprivation (Chow). Session lengths were normalized and the percent of animals feeding was smoothed for visualization. B. Kilocalories (kCal) consumed across conditions. More calories were consumed in Dep48 (n = 8) than Base (n = 12), and in both Dep24 (n = 12) and Dep48 than Chow (n = 8). Red bar indicates group average. * p C. Overview of analysis flow. Recordings were broken into non-overlapping 5 second bins and were assigned to one of 4 categories: not-feeding, rest, feeding, and pre-feeding. Power and coherence were calculated from each bin and averaged together according to category. Average brain activity of each animal during feeding was then normalized by activity during rest and used as predictor variables with behavioral metrics from that animal used as outcome variables. Outcome variables were then permuted to create distributions of ‘by-chance’ performances. D-E. Performance of models in predicting behavioral metrics; black distributions are performances of models using actual outcome variable assignment and white distributions are performances of models using permuted outcome variables. Distribution statistics reported as mean±95% confidence interval and effect size between Actual and Permuted given as Cohen’s d. D. Mean absolute error (MAE) of predicting baseline food consumption in grams (gm). Actual = 2.6±0.08 gm; Permuted = 3.2±0.05 gm; d = 0.43. E. MAE of predicting percent change (%) in food consumption in Dep24 and Dep48 compared to Base (%): Actual = 0.8±0.01%; Permuted = 1.2±0.01%; d = 1.1. F. Percent accuracy (%) in predicting if house chow or palatable food was being consumed: Actual: 59±1%; Permuted: 50±1%; d = 0.52.