Fig. 3: Historical loss remodels the rotation dynamics of vHPC during food competition.

a Left, a comparison of rSLDS states with annotations of food competition behaviors. Right, state transition diagram derived from the empirical transition probabilities. b Composition of behaviors within rSLDS states. c Probability of specific behavioral class aligned to the onset of state 2 and 3 (n = 6 mice for each group). d, e Left, the rotational population trajectories in neural state space of mouse vHPC neurons during competition episodes, and the trajectories are color-coded to correspond with distinct behavioral categories observed in the recorded mice. Green flow field, state 1; Red flow field, state 2; Blue flow field, state 3. Right, the corresponding sequential activity of vHPC neurons during competition episodes. f Calculation of the rotation angle (θ) aligned to the point attractor in state 1 during competition episodes. g Left, empirical cumulative distribution of rotation angle (θ) for various behavioral class (n = 12 mice). Right, quantification of 50% distribution point of rotation angle (θ) for various behavioral classes (****P < 0.0001, n = 12 mice). h Dynamic velocity of Con and HL mice in different discrete states, using the dynamic velocity of state 1 as a baseline (***P < 0.001, n = 6 mice for each group). i Relationship between the dynamic velocity of discrete state and the performance of specific behavioral class. State 2/Snatch duration, P = 0.0108, R² = 0.4943; State 2/Snatch events, P = 0.2050, R² = 0.1549; State 3/Retreat duration, P = 0.5275, R² = 0.0418; State 3/Retreat events, P = 0.0267, R² = 0.4025; n = 12 mice.