Generating Quantitatively Accurate, but Computationally Concise, Models of Single Neurons - Computational Neuroscience: A Comprehensive Approach - page 195

7.2.6 Implications

To build a realistic biophysical model of an oxytocin cell or a vasopressin cell from
the bottom up, there are in principle a very large number of basic membrane prop-
erties to be incorporated. Apart from the Na + and K + conductances that underlie
the generation of the spike, there are a large number of Ca 2 + conductances and K +
conductances that appear to play specific, potentially important roles, and since sev-
eral of the latter are Ca 2 + -dependent, the intracellular Ca 2 + dynamics, involving
buffering and mobilisation of intracellular Ca 2 + stores also need to be included. The
disposition of these conductances in different cellular compartments is poorly un-
derstood, and other conductances, in particular to chloride, and non-specific cation
conductances, are also important, as may be the precise cellular topology. To build
a network model it would also be necessary to incorporate elements reflecting the
nature of stimulus-secretion coupling and the different underlying mechanisms at
dendrites and nerve endings; vasopressin cells, for instance, express different popu-
lations of Ca 2 + channels at the soma and nerve terminals.
On the other hand, we might take an approach that is consciously simplistic, in-
corporating progressively only those features of cells that are essential to explain
particular behaviors, and incorporating these in a minimalist way. We need to set
a verifiable objective: to develop computational models that mimic cells so closely,
that for specific defined attributes, they are essentially indistinguishable in their be-
havior from real cells. In the example shown hereon, we look at the normal discharge
patterning of oxytocin cells and seek a minimalist quantitatively accurate model of
this. To be a good model, the spike output of the model cell must be indistinguishable
from outputs of real cells by any statistical analysis that is applied.

7.3 Statistical methods to investigate the intrinsic

mechanisms underlying spike patterning

7.3.1 Selecting recordings for analysis

Before starting the analysis of the firing activity, we must select suitable recordings.
Stationarity is an essential prerequisite for the analysis to be meaningful. A series is
stationary if there are no systematic trends or rhythmic variations. We started with
stable, long recordings (up to 3h) of spontaneous activity from identified rat oxy-
tocin cells in vivo , and from these, stationary recordings, or long stationary stretches
from recordings that were not stationary throughout, were chosen. Stationarity was
checked with the help of bicubic splines , a series of smooth cubic curves fitted over
short stretches of activity, then joined with the same slope at the joints to form one
continuous curve. One way of looking at temporal patterns is by looking at the time
between consecutive spikes ( interspike intervals ). Interspike intervals provide infor-
mation about the relationship between spikes in a way easily accessible for statistical