File:Model dynamics correlate with empirical cognitive performance. Functional connectivity, synchrony, amplitude and variance of neural population activity depend on excitation-inhibition-ratios.webp

English: "a PMAT24_A_RTCR versus strength of correlation of input currents in the full 379-nodes large-scale BNMs. The models of slower subjects had a higher synchrony between the time series of synaptic currents IEi (p=9.8×10−5). b PMAT24_A_RTCR versus input amplitude. The models of slower subjects had a lower average synaptic current amplitude IEi (p=8.6×10−4)). c E/I-ratio versus parameter settings in the simplified two-node large-scale model. The E/I-ratio of a connection is defined by the quotient of long-range excitation wLREij(black) and feedforward inhibition wFFIij(black). Ji values (green) were obtained by FIC. d E/I-ratio versus FC for active (black) and inactive FIC (blue). A monotonic relationship between E/I-ratio and FC only emerged when FIC was active. e E/I-ratio versus correlation of input currents. f E/I-ratio versus input amplitude. With FIC input amplitudes peaked at relatively low E/I-ratios and then continued to monotonically decrease for increasing E/I-ratios g E/I-ratio versus input variance showed an inverse pattern compared to f. h Amplitude versus variance of inputs. FIC coupled the variance of synaptic inputs with the amplitude of synaptic inputs: the higher the variance (resulting from stronger coupling), the lower the amplitude. i, j Firing rate (Eq. 3) and input current (Eq. 1) time series after injecting 10-Hz sinusoidal waves with increasing variance for active (black) and inactive FIC (blue). FIC compensated higher input variances (which were modulated by the fitting algorithm via the multiplicative coupling parameters wLRE and wFFI) with a lower mean (h). This was necessary as the upper half-wave of the input continued to grow in amplitude for increasing E/I-ratios, while the lower half-wave was bounded by 0 Hz firing (gray to black lines), which required FIC to increase Ji to arrive at the same target average firing rate of 4 Hz. Data in panels c-h are presented as mean values +/- SD derived from N = 100 simulations with different random number generator seeds. Obtained p values of two-sided Pearson’s correlation test: *p < 0.001; including only p values that remained significant after controlling for multiple comparisons using the Benjamini-Hochberg procedure with a False Discovery Rate of 0.1." Study featured in '2023 in science' here.