Investigating Brain-Behavior Architecture Using Graph-Based Network Analysis: Preliminary Insights from Resting-State fMRI

Understanding how the brain supports various cognitive and behavioral functions requires moving beyond isolated domains to a multimodal network-level approach. This preliminary study uses graph theory analysis and resting-state fMRI data from 379 healthy adults (Human Connectivity Project version S1200) to investigate how large-scale brain network topology relates to seven cognitive and behavioral domains: Mental Health (MTL), Externalizing problems (EXT), High-level Cognitive Functions (HCF), Basic Cognitive Functions (BCF), Substance use/abuse (SUB), Delay Discounting Task (DDT), and Pain (PAI). Using the Schaefer 200-region parcellation and seven resting-state networks (DMN, FPN, VAN, DAN, LIMB, SMN, VIS), we calculated single node and network-level metrics across multiple thresholds and modeled nonlinear brain-behavior relationships using Generalized Additive Models (GAMs). Our results revealed domain-specific signatures of integration and segregation. Adaptive functions such as high and basic cognitive functions were associated with greater modularity in the fronto-parietal network (FPN) and segregation of the default mode network (DMN). On the other hand, maladaptive domains such as substance use and impulsivity were linked to increased integration in sensorimotor (SMN) and limbic networks (LIMB). Single-node analysis further identified key hubs whose topological features aligned with distinct cognitive and behavioral profiles. Our findings emphasize the advantage of a network neuroscience approach to investigating the complex and dynamic nature of human cognition and behavior. This preliminary study aims to advance our current understanding of the brain's cognitive and behavior architecture in healthy individuals.