Cognitive development reflects children’s multidimensional environments and individual-specific patterns of functional brain network organization

Poster Session F - Tuesday, April 16, 2024, 8:00 – 10:00 am EDT, Sheraton Hall ABC

Arielle Keller¹ (arielleskeller@gmail.com), Tyler Moore¹, Audrey Luo¹, Elina Visoki², Mārtinš Gatavinš², Alisha Shetty¹, Zaixu Cui³, Yong Fan¹, Eric Feczko⁴, Audrey Houghton⁴, Hongming Li¹, Allyson Mackey¹, Oscar Miranda-Dominguez⁴, Adam Pines⁵, Russell Shinohara¹, Kevin Sun¹, Damien Fair⁴, Theodore Satterthwaite¹, Ran Barzilay²; ¹University of Pennsylvania, ²Children's Hospital of Philadelphia, ³Chinese Institute for Brain Research, ⁴University of Minnesota, ⁵Stanford University

Cognitive neurodevelopment does not take place in a vacuum, but enmeshed in a complex web of environmental exposures and experiences (exposome). However, it remains challenging to comprehensively characterize the many interconnected features of individual environments in large samples. Moreover, most neuroimaging studies use group atlases to define functional brain networks, masking inter-individual variation in spatial organization (functional topography). Here, we leverage cutting-edge computational approaches to investigate how individual differences in cognition emerge during childhood by characterizing both multidimensional environments and individual-specific patterns of functional brain network organization. We conduct pre-registered analyses in discovery (n=5,139) and replication (n=5,137) samples of youth from the longitudinal Adolescent Brain Cognitive Development Study. We quantify the exposome using multilevel exploratory factor analysis with bifactor rotation across 354 child-report, parent-report and geocoded features. We identified seventeen personalized functional networks (PFNs) using non-negative matrix factorization. Linear mixed effects models and cross-validated ridge regressions related the exposome to PFN functional topography and cognition, accounting for age, biological sex, family (siblings), site, and head motion. Exposome scores were associated with current cognition (discovery: ꞵs=0.12-0.50, ps-bonf<.001; replication: ꞵs=0.15-0.48, ps-bonf<.001) and cognition measured two years later (discovery: ꞵs=0.08-0.24, ps-bonf<.001; replication: ꞵs=0.11-0.22, ps-bonf<.001). The exposome was reflected in PFN topography (actual vs. predicted exposome: discovery: r=0.440, p<0.001; replication: r=0.462, p<0.001). Models trained on the exposome more accurately and parsimoniously predicted current and future cognition (rs=0.42-0.46, ps<.001, AICs -454.28-248.82) than models trained on rich personalized neuroimaging data (rs=0.41-0.45, ps<.001, AICs = 2.02x106). These results highlight the importance of childhood environments in cognitive neurodevelopment.

Topic Area: EXECUTIVE PROCESSES: Development &aging