09.03.2013 Views

Contents - Max-Planck-Institut für Physik komplexer Systeme

Contents - Max-Planck-Institut für Physik komplexer Systeme

Contents - Max-Planck-Institut für Physik komplexer Systeme

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Main results.<br />

probability density [kHz]<br />

1. In the case of deterministic adaptation we find<br />

that the ISI density can be well approximated<br />

by an inverse Gaussian probability density. This<br />

density is equal to the first-passage-time density<br />

of a Brownian motion with a drift corresponding<br />

to a base current that is reduced by the adaptation<br />

(Fig. 2A). Furthermore, we have derived an<br />

analytical formula for the serial correlation coefficient<br />

ρn, which shows that neighboring ISIs are<br />

negatively correlated (Fig. 3). This characteristic<br />

feature has been numerically found in other<br />

neuron models with deterministic adaptation and<br />

fast fluctuations (e.g. [2, 3]) and has been experimentally<br />

observed for a variety of neurons (for<br />

reviews see [4, 5]).<br />

2. A stochastic adaptation current can be approximated<br />

by a long-correlated colored noise process<br />

with effective parameters (reduced values<br />

of mean, noise variance, correlation time). As a<br />

consequence, the ISI densities become strongly<br />

skewed and peaked compared to an inverse<br />

Gaussian distribution (Fig. 2B) Another striking<br />

difference is, that the ISI correlations are positive<br />

and not negative as in the case of fast noise<br />

(Fig. 3). We have provided analytical formulas for<br />

the skewness, kurtosis and serial correlation coefficient<br />

of the ISIs that clearly explain these properties.<br />

0.20<br />

0.15<br />

0.10<br />

0.05<br />

0.00<br />

0 5 10 15 20 25 30<br />

A<br />

simulation<br />

inverse Gaussian<br />

0.20<br />

0.15<br />

0.10<br />

0.05<br />

0.00<br />

0 5 10 15 20 25 30<br />

interspike interval [ms]<br />

B<br />

channel model<br />

theory<br />

inv. Gaussian<br />

Figure 2: (A) Interspike interval (ISI) density for the case of deterministic<br />

adaptation. The gray bars display the result from a simulation of<br />

the model, the dashed line represents the inverse Gaussian distribution<br />

obtained from the theory. (B) ISI density in the case of stochastic<br />

adaptation. An inverse Gaussian distribution (dashed line) with the<br />

same mean ISI and the same variance does not fit the ISI density obtained<br />

from the simulation of the model with adaptation channels<br />

(gray bars). Circles depict simulation results of the diffusion approximation<br />

(Gaussian approximation) of the channel noise, the solid line<br />

shows our theoretical approximation.<br />

Conclusions. We have put forward several novel analytical<br />

results that characterize the non-renewal spiking<br />

behavior of neurons with noise and stochastic or<br />

deterministic adaptation. Our analysis has revealed<br />

a striking difference of the ISI statistics depending on<br />

whether slow adaptation noise or fast current noise is<br />

dominating. These findings suggest an indirect way to<br />

determine the dominant source of noise on the basis<br />

of the ISI statistics. As an application, we have used<br />

our theoretical results to understand the origin of neuronal<br />

response variability of auditory receptor cells of<br />

locusts (collaboration with J. Benda, Munich). Preliminary<br />

analysis indicates that slow adaptation currents<br />

may indeed contribute to neuronal variability of this<br />

primary sensory neuron.<br />

serial correlation coefficient<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0.0<br />

determ. adaptation<br />

stoch. adaptation<br />

theory<br />

0.2<br />

0 2 4 6 8 10<br />

lag<br />

Figure 3: Serial correlation coefficient ρn as a function of the lag n.<br />

In the case of deterministic adaptation interspike intervals (ISIs) are<br />

negatively correlated, whereas in the case of stochastic adaptation<br />

ISIs become positively correlated.<br />

[1] T. Schwalger, K. Fisch, J. Benda, and B. Lindner. PLoS Comp.<br />

Biol., 6(12):e1001026, 2010.<br />

[2] M. J. Chacron, K. Pakdaman, and A. Longtin. Neural Comp.,<br />

15:253, 2003.<br />

[3] Y.-H. Liu and X.-J. Wang. J. Comp. Neurosci., 10:25, 2001.<br />

[4] O. Avila-Akerberg and M. J. Chacron. Experimental Brain Research,<br />

2011.<br />

[5] F. Farkhooi, M. F. Strube-Bloss, and M. P. Nawrot. Phys. Rev. E,<br />

79(2):021905–10, 2009.<br />

2.9. How Stochastic Adaptation Currents Shape Neuronal Firing Statistics 59

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!