Constructing functional graphs and recurrent neural network models of neural dynamics using calcium imaging data in Python

Abstract

Leveraging the power of computational analyses and modeling could assist with the study of neural dynamics. Complex analyses like graph theory could enable biomarker discovery in psychiatric illnesses, and recurrent neural network models could be used to reveal mechanisms underlying behavior and cognition. More broadly, both graph theory and recurrent neural networks could be used to understand the topology of functional networks in microcircuits within the brain. This research aims to generate software tools that neuroscientists can readily use to assist experimental analyses of neural data while remaining extensible for more complex analyses. Two Python packages being developed towards this goal are shared: cagraph and carnn. These packages support graph theory analysis and recurrent neural network modeling of neural systems that have been recorded using calcium imaging.