A Biophysical–Computational Framework for EEG Simulation of Connectivity Subtypes in ASD Associated with CPEB3 and CPEB4 Dysfunction

Abstract

Simulation and analysis code accompanying the manuscript "EEG Simulation of Connectivity Subtypes in ASD Associated with CPEB3 and CPEB4 Dysfunction" (under peer review). Author information has been omitted for double-anonymous review and will be added upon acceptance. This repository contains the complete, self-contained Python code that reproduces all numerical results and figures reported in the manuscript. The model is a modified Jansen–Rit neural-mass network over six ASD-relevant brain regions (anterior cingulate cortex, thalamus, amygdala, dorsolateral prefrontal cortex, cerebellum and basal ganglia), coupled through a fixed structural connectivity matrix. Region-specific CPEB3/CPEB4 functional-impact indices scale the excitatory gain, recurrent excitation, pyramidal-to-interneuron drive and inhibitory gain through fixed molecular-to-synaptic coupling coefficients. The system is integrated with a second-order Heun scheme (dt = 1 ms) under a Gaussian firing-rate drive sampled at 1 kHz, and every reported quantity is averaged over ten random-seed realizations. The script runs a 121-combination CPEB3/CPEB4 dysfunction sweep, computes band-limited coherence (alpha 8–13 Hz, gamma 30–45 Hz) and a global connectivity index, identifies hypo- and hyperconnectivity subtypes by k-means clustering (with silhouette validation), evaluates five genetic-variability scenarios, and performs one-at-a-time sensitivity and multi-seed robustness analyses. Outputs include the headline results file (verified_results.json), a per-region-pair coherence-change table (CSV), the coupling-coefficient sensitivity file, and all figures as high-resolution TIFF. Requirements: Python 3.10 with NumPy 1.24, SciPy 1.10, Matplotlib 3.7, Numba and scikit-learn 1.2. Run with python cpeb_model_v3.py; results and figures are written to the output directory. The script is deterministic given its fixed internal seeds, so a clean run reproduces the reported values. License: MIT.