Fig. 3: Abnormal spontaneous fine behavioral features in Shank3 mice.

A Normalized classification matrices highlight the performance of a linear classifier in distinguishing Shank3 mutant mice from WT mice. The classifiers were trained via various behavioral metrics, including position, speed, a combination of two-dimensional measures (“scalars”), or Motion Sequencing (MoSeq) syllables. The optimal classifier performance is indicated by white diagonal fields, with black fields representing frequencies of 1, reflecting perfect accuracy in distinguishing between the two groups. B Heatmaps of behavioral data derived from position, speed, height, length, and MoSeq syllables, comparing Shank3 and WT mice. C Plots of syllable usage with ordered based on the differences of changes. Increased syllables are on the left, whereas decreased on the right (n = 10–12 mice in each group; F_{(71, 1496)} = 10.50, p < 0.0001, RM two-way ANOVA). D Similarity dendrogram of the 71 syllables of MoSeq demonstrates hierarchical organization. E Histogram of syllable usage of eight key subtypes (F_{(7, 158)} = 11.32, p < 0.0001, RM two-way ANOVA). Data are presented as the means ± SEMs. *p < 0.05, **p < 0.01, ***p < 0.001 by Tukey’s post hoc test of RM two-way ANOVA.