Fig. 2: The mesoscale brain-wide fluctuation analysis (MBFA) framework.

A Sample LFP traces recorded concurrently from eight implanted brain regions. Brain activities were obtained during social interaction behavior. B Schematic of the workflow for neural oscillation analysis of multiple brain regions signals. For each mouse, the PSD ratio of eight brain areas during SIT across the designated trial was calculated. C Schematic of the workflow for network analysis of multiple brain regions signals. For each mouse, the network of eight brain areas during SIT across the designated trial was constructed based on the directional transfer function. Global efficiency and information flow were then calculated. D, E Schematic of SVM workflow. There were two separate binary classifications and one multi-class classification. The features used in the SVM included behavioral data (social index) and LFP characteristics (PSD Ratio, global efficiency, and information flow across different frequency bands between brain regions). We selected features that exhibited significant differences in LFPs among the comparison groups (Control vs. CSDS, Ket vs. Sal, Control vs. CSDS vs. Ket) and significant correlation in LFPs characteristics-SI. The dataset was randomly parsed into a training set (70%) and a test set (30%). Accuracy, ROC curves, and AUC metrics were employed for assessing model performance.