Fig. 2: Persistent trajectory-specific tuning over weeks.
From: A persistent prefrontal reference frame across time and task rules
a Example of x/y-coordinates, linearized position, and normalized calcium activity (sorted by location of peak activity during left trajectories). s: sampling; r: reward. b Spatial tuning functions during odd and even left and right runs within the first recording day (n = 1109 neurons). c Top: Spatial correlation of odd vs. even runs (consistency) is higher within trials of the same vs. opposite direction (t = 18.21, p = 3*10−7). Bottom: Spatial information (SI) is comparable during left and right trials (t = −0.28, p = 0.791). Two-sided paired t-tests. d Spatial information (SI) remains consistent across days during left (top, r = 0.654, p = 10−62) and right runs (bottom, r = 0.648, p = 10−61, Spearman’s correlation coefficients day 24 vs. day 1, n = 502 neurons). 2-sided correlation tests. e Examples of spatially binned activities during five superimposed individual left runs on day 1 and day 24. f Spatial tuning functions sorted for the peak location on day 1 (for left and right trajectories, respectively) show large spatial stability across days. g Top: Average trajectory-specific spatial correlation to day 1. Correlations decayed over time (F = 16.48, p = 10−6 for the correlation-time interaction) but remained significant vs. shuffled data to the last day (t = 11.29 to 18.42, p = 10−5 to 3*10−7, 2-way repeated measures ANOVA followed by paired t-tests with Šidák correction). Bottom: Distribution of correlation coefficients of individual cells. h Decoding of the animals’ linearized position. Left: Example of the prediction on day 24 using a model trained on day 1 (blue). Right: Predictions decayed over time (F = 3.48, p = 6*10−4 for the decoding error-time interaction) but remained significant vs. shuffled data (gray, t = −7.32 to −13.00, p = 2*10−4 to 6*10−6, 2-way repeated measures ANOVA followed by paired t-tests with Šidák correction). Boxes show median and upper/lower quartiles. Circles are individual mice (c) or cells (d). Lines in (g) and (h) show mean ± sem, thin lines are individual mice. Left and right trajectories were correlated separately and pooled subsequently (n = 8 for all comparisons). Source data are provided as a Source Data file.
