Extended Data Fig. 3: Additional analysis on time cell tuning.
From: Medial entorhinal cortex mediates learning of context-dependent interval timing behavior
a. Top, sequence of MEC time cells significantly tuned for S-S trials, sorted by S-S trials and displayed for S-S, S-L and L-S trials. Middle, same as above, expect for significant MEC time cells on S-L trials and sorted by S-L trials. Bottom, same as top, except for significant MEC time cells on L-S trials and sorted by L-S trials. b. A generalized linear model was used to assess whether neurons are tuned to one of three variables- time in the trial, distance travelled from trial start, or licking – or a combination of 2 or 3 variables (see Methods). Analysis performed on n = 695 time cells, collected from 10 behavioral sessions, lead to n = 177 significant models. Boxplot showing log-likelihood increase gained by each variable: time (median = 37.15), distance (median = 14.63) and licking (median = 0.82) (two-sided Kruskal-Wallis test, p = 2.6×10-26; followed by two-sided Wilcoxon rank-sum test with Bonferroni-correction: Time vs. Distance: p = 1.7×10-08; Time vs. Licking: p = 3.7×10-22; Distance vs. Licking: p = 8.8×10-13). Log-likelihood was normalized to recording time in minutes. c. Histogram demonstrating the model that best described the calcium activity of each cell and trial type. d. Boxplot showing adjusted variance explained for models that best describe the calcium activity of each cell for the single variable models: time (median = 0.0512), distance (median = 0.0202) and licking (median = 0.0101). Number of models n = 125,396. (One-sided Wilcoxon signed rank test (median greater than 0), p = 0.0001, p = 0, p = 0.0312). For box plots, the line inside of each box is the sample median. The upper quartile corresponds to the 0.75 quantile and the lower quartile corresponds to the 0.25 quantile. The blue dots in b and d represent outliers. Outliers are values that are more than 1.5 × interquartile range (IQR) away from the top or bottom of the box. The whiskers are lines that extend above and below each box. One whisker connects the upper quartile to the nonoutlier maximum (the maximum data value that is not an outlier), and the other connects the lower quartile to the nonoutlier minimum. e. Proportion of MEC time cells that either remained stable, displayed a time shift, displayed rate remapping, or displayed on/off dynamics across trial types. f. Rank order analysis for shuffle distribution (black) and real data (red). The similarity of the sequences of time cells across trial types is examined by comparing their rank orders. Each time cell is assigned three rank orders, corresponding to its sorting by peak timing for each trial type. Subsequently, the mean difference between rank orders within a cell is compared to a shuffle distribution, generated by shuffling rank order of cells 10,000 times. The p-value is computed as the proportion of shuffle values smaller than the actual data. Notably, p-values are zero for all three comparisons. g. Discriminant Index indicates the extent to which the difference in dF/F between trial types deviates from chance level. S-S vs S-L: Day 1 n = 224, Day N n = 221, z = 2.55, p = 0.01; S-S vs L-S: Day 1 n = 225, Day N n = 231, z = 3.80, p = 1.4×10-4, two-sided Wilcoxon rank sum test. Individual data points with median.
