Figure 5. Computational modeling reveals potential mechanism underlying preferential transmission of coherent input.

(A) Spike-triggered average of white-noise input signal (black), network excitation (red), and inhibition (blue) demonstrates preferential transmission of gamma-frequency input that matches the intrinsic dynamics of the network. White-noise averaging was triggered by spikes of all excitatory neurons; results for inhibitory neurons or all neurons (total MUA) are shown in Figure S8C. Inset depicts a schematic of the PING model.

(B) GC spectrum from white-noise input to total MUA (black), excitatory spikes (red), and inhibitory spikes (blue). Spectra from MUA and spikes to white noise are presented in muted color and overlap near zero.

(C) GC spectra between excitation and inhibition in the network. Spectrum from excitatory spikes to inhibitory spikes (blue) and vice versa (red).

(D) Spike-triggered averages in the model with two excitatory populations. Averages display the spike-triggered white noise (black) driving the second excitatory population and illustrate entrainment by the excitation (red) and inhibition (blue) of the recurrently coupled PING network. Inset depicts a schematic of the model. The PING network shown on top, in lighter colors, contains a first excitatory population and an inhibitory population and generates gamma upon white-noise input. The resulting rhythmic inhibition is fed into a second excitatory population, shown on the bottom, which is driven by independent white noise.

(E) Spike-triggered averages as in (B), but for the PING+M model. Inset depicts a schematic of PING+M model driven by white noise.

E, excitatory neuron pool; I, inhibitory neuron pool; WN, white-noise input.