Furthermore, we could reduce the effect of these perturbations by incorporating connections between excitatory PNs and the inhibitory LNs (Figure 6C, bottom panel) that would mimic a typical biological network like the insect AL consisting of interacting excitatory and inhibitory neurons. The back-and-forth interaction between excitation and inhibition tends to promote synchrony in both sets of neurons. Each group

of LNs spiked in alternation, thus respecting Selumetinib the coloring of the network as a constraint. (Börgers and Kopell, 2003) have also demonstrated that increasing the strength of excitatory to inhibitory neurons tends to mitigate the influence of heterogeneity due to random connectivity. This synchronization mechanism has been demonstrated in the olfactory bulb of rats where the interaction between mitral cells and granule cells results in the emergence of rapid synchrony in the network (Schoppa, 2006) (but see Galán et al., 2006). In previous simulations we had introduced small variations in the excitability of individual neurons to ensure that the solutions would be robust to parameter noise. We also added a small selleck chemicals amplitude noise term (approximately 10% of the amplitude of the DC input to each neuron) (see Supplemental Information for details). The role of the coloring

of the network on the dynamics remained robustly evident in spite of these variations. However we kept a key parameter, the timescale of adaptation, constant across all neurons. Adaptation allows the LNs to switch from a spiking to a quiescent

state (Figure 1). The timescale of adaptation affects the duration that a neuron spends in each state. In a realistic network the timescale of adaptation may be distributed across the population of LNs. We sought to determine if such variation can compromise the coloring-based dynamics of the inhibitory subnetwork. We simulated the dynamics of the network Ketanserin with broad variability consistent with the timescale of the Ca2+-dependent potassium current. A parameter τx was added to the timescale τm of the Ca2+-dependent potassium current (see equation for m in the section Ca2+-dependent potassium current IKCa in the Supplemental Information). Its value τx was picked from a uniform random distribution with values extending from −0.02 to 0.01 ( Figure 6D). The range of this distribution was adjusted to generate a large variation in the oscillatory switching frequency of two reciprocally coupled LNs. Switching between the active and quiescent state was slow (τx = 0.01) and increased dramatically for τx = –0.02 ( Figure 6E). We found that a wide variation in the Ca2+-dependent potassium current leads to nonuniformity in the duration of LN spike bursts across the duration of the odor presentation. However, despite this realistic variability, the dynamics of the LNs continues to follow the coloring of the network ( Figure 6F).