Network comparisons and their applications in connectomics

Niharika Shimona D'Souza, Archana Venkataraman

Abstract: Over the past decade, there has been a growing emphasis on neuroscience to analyze the human brain from the perspective of complex networks. Here, connectomics, or the study of underlying connectivity patterns in the brain, has provided several fundamental insights into its intrinsic organization. However, hypothesis-driven discovery in this domain is extremely challenging due to the high data dimensionality, environmental confounds, and considerable interindividual variability. In light of these challenges, this chapter provides a deep dive into network-based analyses in connectomics. Specifically, we will introduce the idea of “network comparison” from two complementary perspectives: capturing group differences using simple graph-theoretic measures and mechanistic network models that integrate another layer of complexity to parse connectomics data. Next, we examine data-driven approaches to network comparison that can capture intersubject variability. These approaches span classical machine learning, geometric techniques, and deep learning. We conclude with a brief snapshot on extending network comparison concepts to dynamic network analysis, as a prelude to later chapters of this book.