Functional reconstruction of dyadic and triadic subgraphs in spiking neural networks

Abstract

Neural networks reconstructed from measurement data are known to exhibit various forms of nonrandom structures, including subgraph motifs and small worldedness. It has been suggested such nonrandom structures are critical for neural information-processing; however, it is unclear how the topological structure of anatomical networks influences the reconstruction of functional networks. To better understand the importance of such nonrandom structures, we study how dyadic and triadic subgraphs are preserved during the reconstruction. We use a model-free information-theoretic measure, transfer entropy, to quantify the directional associations of pairwise neuronal activity. We employ multiplex networks to compare how dyadic and triadic subgraphs differ from structural to functional networks, with particular attention to recurrent connections. We find that certain structural subgraphs have more influence on the topology of the functional network than others.