Extended Data Fig. 2. Corner score shuffling and spatial decoding.
(a) Distributions of corner scores calculated without applying the penalty (Methods) for all subiculum neurons recorded in the triangle (n = 4774 cells from 18 sessions of 9 mice), square (n = 4685 cells from 17 sessions of 9 mice), and hexagon (n = 4774 cells from 18 sessions of 9 mice) environments. The red lines represent the 95th percentile of the distributions in (b). Note: This calculation is for display purposes only and was not used for the analyses in this paper, also see Methods for details. (b) Distributions of corner scores calculated with penalty applied (Extended Data Fig. 1g), which was used in the analyses. The red lines represent the 95th percentile of the corresponding distributions. (c) Distributions of shuffled corner scores calculated without applying the penalty, which was used in the analyses. Each cell was shuffled 1000 times. The red lines represent the 95th percentile of the distributions in (b). The red box over (b) and (c) indicates the method used in corner score calculation and shuffling procedures in the current work, which resulted in the most stringent corner score criteria for defining corner cells. (d) Same as c, but distributions of shuffled corner scores calculated with penalty applied. The red lines represent the 95th percentile of the distributions in (b). Note: this method was not used in this paper. (e) An example of the true vs. decoded spatial x-y position using the full decoder (all recorded subiculum neurons). (f) Decoding performance of the full decoder in different environments. The decoder was trained and tested within each session using 10-fold cross-validation. Each dot is a session, black lines represent the median. Decoding errors were compared with the corresponding shuffle within each condition (two-tailed Wilcoxon signed-rank test: all p < 0.0001; n = 15, 18, 17, and 18 sessions for circle, triangle, square, and hexagon, respectively, from 9 mice). (g) Decoding error ratio from a triangle (top) or square (bottom) session, color-coded for larger (white) and smaller (black) error ratios. The error ratio was obtained by taking the error map from the corner cell removed decoder and dividing it by the error map of the full decoder. Corner areas (orange boxes) showed higher error ratios than the center area (blue box). (h) Quantitative comparisons of the decoding error ratios between the corner and the center areas (shown in g) for all sessions. Each line is a session (two-tailed Wilcoxon signed-rank test: triangle: p < 0.0001; square: p < 0.0001; hexagon: p = 0.0004; n = 18, 17, and 18 sessions for triangle, square, and hexagon, respectively, from 9 mice).