Fig. 4: Heterogenic changes in spontaneous behaviors as depressive-like behavioral pathogenesis progresses.

a, b Confusion matrices for the predictive performance of the random forest classifier in discriminating the timing of behavior recordings (left) or the duration of stress exposure (right) based on syllable usage (a) or bigram (b) in exploratory behavior of stressed and non-stressed mice (N = 12 for both groups). The classification rate represents how mice with true labels (row) were classified into predicted labels (column). For the box plots, the boxes span the interquartile range (25th–75th percentiles); center lines indicate the median, and whiskers indicate the minimum and maximum values. c, d The mean LDA points for syllable frequencies (c) and transitions (d) for the four experimental time points (1–4 weeks) in stressed (circle) or non-stressed (square) mice. e A color-coded dot plot showing chronic stress-induced changes in syllable usage across the stress paradigm, where the sizes of the circles indicate the significance level of alteration estimated with Benjamini–Hochberg post hoc following two-way ANOVA. f Top five most impactful syllables in the classification of time points for stressed mice, sorted by overall absolute Shapley value. The ethogram is indicated along the y-axis, with each cluster ID presented next to its corresponding bar. g Beeswarm plot for syllables with the top five highest contributions in distinguishing third-week CUS mice. h Prediction performance of the classifier diagnosing the psychiatric state of individual mice at different time points (1–4 weeks) based on syllable frequencies (top) or transitions (bottom). The top prediction scores in accuracy, precision, or recall are indicated in bold to highlight the pathogenic features of depressive symptoms. i Circular network graph representing significant modifications in transition probability induced by stress (left) or the time point (right) (multiple two-tailed unpaired t-tests). Source data are provided as a Source Data file.