Extended Data Fig. 2: Track discrimination estimated by Bayesian decoding probability.
From: Nested compressed co-representations of multiple sequential experiences during sleep
a, Place maps across the 15 tracks used for Bayesian decoding in one example rat. Neurons were ordered based on the location of their peak firing on track 6. Place maps were normalized for each neuron based on its activity on the concatenated 15 tracks. b, Example of Bayesian decoding of animal position based on place cells spikes during one lap of animal run on one track. c, Distribution of Bayesian decoding error of animal position on individual tracks during run based on the activity of the place cells on the corresponding tracks. Black line indicates the mean error value. d, Confusion matrix of decoded locations and actual rat locations during run. e, Distribution of correlation coefficients between the overall decoded posterior probabilities within a track and the sequence score for the corresponding track across all sleep sessions (Pearson’s correlation, R=0.02, P<0.0002, N=20). f-g, Cosine angle (f) and Euclidian distance (g) between the population vectors of place cell activity across different tracks during run and the corresponding neuronal activity across different frames during sleep (Wilcoxon signed rank test); (f): P = 0.14; 9×10-5; 0.33; (g) P=0.04; 9×10-5; 0.97; n=20 decoding sessions. Blue lines depict average values of corresponding data in grey. One run session connects to corresponding 4 sleep sessions. h, Signal to noise ratio of track discrimination using decoded probability on tracks during run and during sleep (all animals). Wilcoxon signed rank test, P= 9×10-5; n=20 decoding sessions. Blue lines depict average values. i, Cross correlation of sequence scores for the same tracks across temporally nearby sleep frames as a function of temporal lag between them. j, Proportion of significant sleep frames as a function of the temporal difference between the run session and the sleep session when the frames were restricted to those co-occurring with sharp wave ripples. Data are represented as mean±SEM. ***P<0.005. *P<0.05. ns=not significant.
