Extended Data Fig. 7: Additional decoding of active and passive decisions, huddle size, and huddle membership.
From: Cortical regulation of collective social dynamics during environmental challenge
a-d. Performance of SVM decoders (mean ± SEM) trained to classify active entry from speed-matched running (a), active exit from speed-matched running (b), passive entry from rest (c), and passive exit from rest (d). e,f. Cross-decoding performance of SVM decoders trained on active entry predicting active exit (e), or trained on active exit predicting active entry (f) g,h. Cross-decoding performance of SVM decoders trained on passive entry predicting passive exit (g), or trained on passive exit predicting passive entry (h). i,j. Cross-decoding performance of SVM decoders trained on active entry predicting passive entry (i), or trained on active exit predicting passive exit (j). k,l. Cross-decoding performance of SVM decoders trained on passive entry predicting active entry (k), or trained on passive exit predicting active exit (l). m. Example raster plot of huddling behavior for all four animals in one session, color coded by huddle size. n. Example pie charts showing proportion of time for various huddle configurations for huddles of two and three for one group. o. Matrix showing proportion of subject animal’s (x-axis) total huddle time with partner animals (y-axis) for one session. Sum of proportions for one animal can exceed 1 because subjects can huddle with multiple animals simultaneously. p. Partner preference index (maximum preferred partner – minimum preferred partner) for real data vs. shuffled data in which binary vectors containing individual huddle behaviors are circularly shifted relative to each other. q. Schematic of potential huddle memberships for huddles of two during miniscope imaging. r. Performance of multi-class LDA decoders trained to classify huddle membership for huddles of two from dmPFC population activity. Baseline is 33% (three possible memberships). s. Schematic of potential huddle sizes during miniscope imaging. t-v. Performance of SVM decoders trained to classify huddle size of 2 vs. 3 (t), 2 vs. 4 (u), and 3 vs. 4 (v). Box plots: center line–median; box limits–upper/lower quartiles; whiskers–min/max. Statistical tests: two-way ANOVA with Bonferroni post-hoc (a-d) and two-tailed Wilcoxon matched pairs tests (e-l,p,r,t-v). *P < .05, **P < .01, ***P < .001, ****P < .0001. See Supplementary Table 1 for statistical details.
