Neural Circuit Inference from Function to Structure

Curr Biol. 2017 Jan 23;27(2):189-198. doi: 10.1016/j.cub.2016.11.040. Epub 2017 Jan 5.

Abstract

Advances in technology are opening new windows on the structural connectivity and functional dynamics of brain circuits. Quantitative frameworks are needed that integrate these data from anatomy and physiology. Here, we present a modeling approach that creates such a link. The goal is to infer the structure of a neural circuit from sparse neural recordings, using partial knowledge of its anatomy as a regularizing constraint. We recorded visual responses from the output neurons of the retina, the ganglion cells. We then generated a systematic sequence of circuit models that represents retinal neurons and connections and fitted them to the experimental data. The optimal models faithfully recapitulated the ganglion cell outputs. More importantly, they made predictions about dynamics and connectivity among unobserved neurons internal to the circuit, and these were subsequently confirmed by experiment. This circuit inference framework promises to facilitate the integration and understanding of big data in neuroscience.

Keywords: bipolar cells; brain circuit; circuit model; computational neuroscience; ganglion cells; machine learning; neural code; neurophysiology; retina; vision.

MeSH terms

  • Action Potentials
  • Animals
  • Models, Neurological*
  • Neurons / chemistry
  • Neurons / cytology
  • Neurons / physiology*
  • Retinal Ganglion Cells / chemistry
  • Retinal Ganglion Cells / physiology*
  • Urodela / anatomy & histology*
  • Urodela / physiology*