Extended Data Fig. 3: Nested parallel co-representation of multiple tracks within sleep frames estimated by Bayesian decoding.
From: Nested compressed co-representations of multiple sequential experiences during sleep
a-b, Example multi-track sleep frames analyzed by the Bayesian decoding method. Red outline indicates tracks that were significantly represented (for example, tracks 2&10 in a and 3&4 in b) in the corresponding sleep frame shown on the far left. Decoded probabilities were normalized across all tracks (top) or within each track (bottom). c, Proportion of sleep frames grouped by the number of significantly represented tracks estimated by Bayesian decoding in each sleep session and corresponding shuffle sleep. Pie charts summarize the proportions of frames with 0 (that is, no), 1, 2 and >2 tracks represented significantly in parallel within the same frame. d, Multiple-track parallel co-representation in sleep frames across sleep sessions (that is, from pre- to post-experience sleep) and in corresponding shuffle sleep expressed as proportions of multi-track frames among all frames (left) or among significant p/replay frames (right). e, The minimum total number of tracks investigated after which the average or peak of the distribution of multi-track co-representing sleep frames exceeded those of sleep frames representing single tracks, across all sleep sessions. f-g, Proportion of sleep frames when the threshold of significance was set to 0.002 (f, Bonferroni correction) or frame significance was adjusted by Benjamini-Hochberg correction (g). Left: Pie plots of proportions of the sleep frames grouped by the number of significantly represented (and co-represented) tracks estimated by Bayesian decoding across all sleep sessions (data and shuffle sleep). (f) Wilcoxon signed rank test, two-sided, P=0; n=300 tracks; (g) Wilcoxon signed rank test, two-sided, P=0; n=300 tracks. Middle: Cumulative proportions of significant sleep frames expressed as a function of the number of tracks explored. Right: the difference of cumulative proportion of significant sleep frames between data and shuffle sleep. Data are represented as mean±SEM. ***P<0.005.
