Datasets and scripts for "Attractor dynamics of a whole-cortex network model predicts emergence and structure of fMRI co-activation patterns in the mouse brain"

We provide datasets and scripts for reproducing the results of "Attractor dynamics of a whole-cortex network model predicts emergence and structure of fMRI co-activation patterns in the mouse brain". The datasets include the 34 x 34 anatomical connectivity of the mouse brain between excitatory populations, the empirical fMRI time series of 15 mice, and their organization into 6 co-activation patterns (CAPs). The Python script “Empirical_VS_Model.py” implements 3 whole-cortex models with their corresponding best-fit parameters (our full model with directed connectivity and non-linear neural dynamics, an undirected model with non-linear dynamics but undirected anatomical connections, and a linear model with directed anatomical connections but linear activation function). The script uses those models to calculate network statistics averaged over long-time scales (e.g. the functional connectivity matrix), and it finally compares them to the corresponding empirical statistics of the real mouse brain. The script also plots the figures showing the balance between the excitatory and inhibitory currents in our full model. To conclude, the Python script “Attractors_and_CAPs.py” implements our full best-fit model with directed connectivity and non-linear neural dynamics, and it uses the model to calculate the topography and probability of occupancy of its activation state attractors. The script also contains our mapping algorithm, which reconstructs the probability of occupancy of the attractors from the empirical data and compares it with the model distribution. Then the script uses linear combinations of the model attractors to reconstruct the topography of CAPs, and it finally compares the model topography to the empirical one obtained from the real mouse brain.

本数据集与代码脚本用于复现《全皮层网络模型的吸引子动力学可预测小鼠脑功能磁共振共激活模式的出现与结构》一文的研究结果。数据集包含小鼠脑兴奋性神经元集群间的34×34结构连接矩阵、15只小鼠的实测功能磁共振成像（fMRI）时间序列，以及基于这些序列得到的6种共激活模式（co-activation patterns, CAPs）。Python脚本`Empirical_VS_Model.py`实现了3种全皮层模型及其对应的最优拟合参数：本研究的完整模型（含定向连接与非线性神经动力学）、仅具备无向结构连接与非线性动力学的无向模型，以及仅具备定向结构连接与线性激活函数的线性模型。该脚本利用上述模型计算长时程平均的网络统计量（如功能连接矩阵），并将其与真实小鼠脑的对应实测统计量进行对比；此外，该脚本还可绘制用于展示完整模型中兴奋性与抑制性电流平衡关系的图像。Python脚本`Attractors_and_CAPs.py`实现了本研究的最优拟合完整模型（含定向连接与非线性神经动力学），并利用该模型计算其激活状态吸引子的空间分布与占据概率。该脚本还集成了本研究提出的映射算法：该算法可从实测数据中重构吸引子的占据概率，并将其与模型预测的分布进行对比。随后，该脚本利用模型吸引子的线性组合重构共激活模式的空间分布，并最终将模型得到的空间分布与真实小鼠脑的实测分布进行对比。