Fisher Information as the Intrinsic Metric of Conductance-Based Neuronal Noise

It is generally accepted that sensory neurons transduce “information” from their environment into biologically relevant signals, allowing for adaptation and body homeostasis. However, it has not been clear what sort of information metric is generated by, or flows through, the neuron. In this work, we stated this problem based on the physical foundations of a conductance-based model of neuronal excitability, starting from the mother equations of an ensemble of N identical ion channels. For any number and nature of channel states, we show that the Fisher information in the output (membrane conductance, potential or current) about the input (the stimulus) arises as the intrinsic information metric of the gating dynamics of the ensemble. We found that the Fisher information is a primary component of the fluctuations of the neuronal membrane conductance, along with a parameter accounting for how deterministic or stochastic the ensemble is (a function of N). We also obtain a relation for the rate of transitions of the neuron, from a given configuration with n open channels to another, which is the primary origin of the intrinsic noise in the absence of artificial stimulation. This result establishes an important analytical correlation between the fluctuations of the neuron output signal, the information it carries (in Fisher’s terms), its physical and kinetic properties, and the corresponding stimuli. This might constitute a theoretical foundation for: i) designing of strategies for stochastic resonance-based computing procedures inspired by the neural intrinsic noise, and ii) modulating neural function by tuning its intrinsic Fisher information or the expected value of conductance fluctuations, with that of a binomially distributed external noise. For channels with inactive states, the model suggests the simultaneous use of two different signals of such noise. Such proposals constitute a new paradigm, which qualitatively differs from the traditional Gaussian white or Lévy noises.