Functional connectivity in cultured cortical networks during development: Comparison between correlation and information theory-based algorithms

Goal of this work is to present a general approach to estimate functional connectivity in in vitro cortical networks coupled to Micro-Electrode Array (MEAs). Specifically, we developed and optimized a Partial Correlation (PC) based algorithm and we compared it to Cross Correlation (CC) and Transfer Entropy (TE) methods. First, we applied the algorithms to simulated networks with different average connectivity degrees. Second, we used a specific validation procedure based on the accuracy coefficient (ACC) to evaluate the algorithm's performances and we found Partial Correlation to be the best method to infer functional connections from spiking activity of in vitro cortical networks. Finally, we used PC to estimate connectivity during development (i.e., from 2nd to 4th week) from recordings of cortical networks coupled to MEAs.