|Ph.D Student||Soudry Daniel|
|Subject||Neurons: from Biophysics to Functionality|
|Department||Department of Electrical Engineering||Supervisor||Professor Ron Meir|
|Full Thesis text|
It is commonly believed that neurons constitute the basic “circuit components” in the cortex, and that these neurons mainly communicate through voltage “spikes” called Action Potentials (APs) transmitted through “synapses”. Therefore, a quantitative characterization of the biophysical and functional aspects of the AP generation mechanism is of fundamental importance. This mechanism is based on “excitable” interactions between the voltage on the membrane of the cell and the protein pores that populate that membrane, the so-called “ion channels”.
Biophysical models of neurons, termed Conductance-Based Models (CBMs), often neglect both the inherent stochasticity in the ion channel dynamics, as well as various slow biophysical processes. However, recent experiments demonstrated that even with a simple periodical pulse stimulation (with a fixed amplitude), synaptically isolated cortical neurons can fire in a highly irregular manner and display complex temporal patterns over multiple timescales (“1/f behavior”) - from seconds to days. This raises the question as to whether the intrinsic ion channel “noise” in the neuronal dynamics and the slow processes should be included in the CBM to generate this type of behavior, and if so how.
In order to investigate this question, we first develop new numerical techniques to simulate populations of stochastic ion channels efficiently, and also novel analytical methods that allow us to understand and precisely tune general CBMs. For example, we show that if the input to the neuron matches the sparse “spiky” nature of the output (as in the experiment), we can derive the spiking Input-Output (I/O) relation with relatively few assumptions. This I/O, based on biophysically meaningful parameters, is well described by an `engineering-style' block diagram of a linear system with feedback. Such a linear I/O allows the utilization of well known statistical tools to derive all second order statistics, construct linear optimal estimators and perform parameter identification. These results hold numerically, even in some cases where our assumptions break down.
We use this relation to reproduce and interpret the irregular and complex 1/f response observed in isolated neurons stimulated with sparse inputs. We show, under such sparse stimulation, that the intrinsic ion channel noise should have a significant impact on the neurons` response. Moreover, we show that the neuronal response to sparse inputs can be decomposed into contributions from its long history of internal noise (days) and its “short” (few minutes) history of inputs. Thus, we can quantify memory and noise in the neuronal response for such sparse inputs.