Multivariate directed connectivity analysis (MDCA) enables decoding of directed interactions in neural and social networks

Abstract

AbstractMany scientific fields currently face the daunting task of studying the dynamics of complex networks. For example, while we know that the rich mental phenomena of humans and other animals are mediated by complex systems of neural circuits in the brain, the mechanistic links between these biological networks and the functions that they mediate are poorly understood. Here we present a novel class of methods, termed multivariate directed connectivity analysis, to investigate network dynamics via patterns of directed interactions between network nodes. We validate these methods using simulated data and apply them to three real-world datasets, two neuroscientific and one investigating the 2016 US presidential candidates’ influence on the social media service Twitter. We find that these methods enable novel understanding of how information processing is distributed across networks. The methods are generally applicable to the study of dynamic information networks in biological, computational, and other fields of research.