Figure 1

Effect of mean network correlation. (a) The first column in each row shows the raster plot of the spiking activity from a simulated neuronal network for a firing rate of 20 Hz and different network correlation levels. Histogram of the Ising couplings, partial correlations and cross-correlations for the pairs of neurons that are synaptically connected and not connected are shown respectively in the second, third and fourth columns. The corresponding ROC curves of the three functional connectivity metrics are shown in the last column. The first, second and third rows correspond to mean network correlation levels (ρ) of 0.001, 0.03 and 0.3 respectively. (b) Plot of the AUC values for different mean network correlation levels in scale-free networks of 30 neurons for a fixed firing rate of 20 Hz. Mean value was calculated from ten simulated networks. For weaker correlation levels (0.001 and 0.003), AUC value of Ising couplings was significantly higher than partial and cross-correlations. For stronger correlation levels (0.1 and 0.3), partial correlations had a significantly higher AUC value compared to Ising couplings and cross-correlations (p \( < \) 0.01, two-sample t-tests). (c) True positive rate (TPR) and false positive rate (FPR) for the reconstruction of the structural connections by the three functional connectivity metrics thresholded at a sparsity threshold value of 20%.