A Connectional Gradient of Individual Variability across Functional Connectome Edges

1506 

Abstract Submission 

Hang Yang¹, guowei wu^1,2, Yaoxin Li^1,3, Xiaoyu Xu^1,4, Yiyao Ma¹, Runsen Chen⁵, Adam Pines⁶, Ting Xu⁷, Valerie Sydnor⁸, Theodore Satterthwaite⁹, Zaixu Cui¹

¹Chinese Institute for Brain Research, Beijing, China, ²University of Chinese Academy of Sciences, Beijing, China, ³Michigan Neuroscience Institute, University of Michigan, Ann Arbor, MI, ⁴State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, China, ⁵Vanke School of Public Health, Tsinghua University, Beijing, China, ⁶Psychiatry and Behavioral Sciences, Stanford School of Medicine, Stanford University, Stanford, CA, ⁷Child Mind Institute, New York, NY, ⁸University of Pittsburgh, Pittsburgh, PA, ⁹UPenn, Philadelphia, PA

Hang Yang  
Chinese Institute for Brain Research
Beijing, China

Guowei Wu  
Chinese Institute for Brain Research|University of Chinese Academy of Sciences
Beijing, China|Beijing, China

Yaoxin Li  
Chinese Institute for Brain Research|Michigan Neuroscience Institute, University of Michigan
Beijing, China|Ann Arbor, MI

Xiaoyu Xu  
Chinese Institute for Brain Research|State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University
Beijing, China|Beijing, China

Yiyao Ma  
Chinese Institute for Brain Research
Beijing, China

Runsen Chen  
Vanke School of Public Health, Tsinghua University
Beijing, China

Adam Pines  
Psychiatry and Behavioral Sciences, Stanford School of Medicine, Stanford University
Stanford, CA

Ting Xu  
Child Mind Institute
New York, NY

Valerie Sydnor  
University of Pittsburgh
Pittsburgh, PA

Theodore Satterthwaite  
UPenn
Philadelphia, PA

Zaixu Cui  
Chinese Institute for Brain Research
Beijing, China

The regional functional connectivity (FC) variability exhibits a cortical gradient, increasing from the primary sensorimotor to higher-order association areas (Mueller et al. 2013; Sydnor et al. 2021). This inter-individual variability in FC matures throughout development (Tooley, Bassett, and Mackey 2021; Teeuw et al. 2019). However, the organization of individual FC variability at the edge level remains unclear. In this study, we identified a connectional gradient in the edge-level FC variability and investigated its maturation during youth.

We utilized the Human Connectome Project (HCP)-development (HCP-D, n = 415, aged 8–21) dataset (Somerville et al. 2018) and unrelated HCP-young adult (HCP-YA, n = 245, aged 22–35) dataset (Van Essen et al. 2013) to assess edge-level FC variability. The minimally preprocessed fMRI data were post-processed with the extensible connectivity pipelines (XCP-D), including nuisance regression (36P), filtering (0.01–0.08 Hz), and smoothing (FWHM = 6 mm). We obtained 45- and 58-min fMRI data from the HCP-D and HCP-YA datasets and equally divided them into 12 and 8 sessions, respectively. We extracted regional BOLD timeseries using the Schaefer atlas with 400 regions (Schaefer et al. 2018) and calculated FC between each pair of cortical regions using Pearson correlation. A linear mixed-effects model estimated inter- and intra-individual variability for each FC edge (Xu et al. 2022). The intra-class coefficient (ICC) measured adjusted inter-individual variability, considering intra-individual variability (Mueller et al. 2013). Finally, a 400×400 inter-individual variability matrix was generated, with each element representing the variation in connection strength for an edge across all participants.
Then, we explored the FC variability gradient maturation in youth by employing a sliding-window method (Vasa et al. 2018) to sort HCP-D participants by age (length = 50, step = 5), resulting in 74 groups (Fig. 2a). First, we assessed whether FC variability matured towards an adult-like pattern during youth. Using FC variability in the HCP-YA dataset as a reference, we estimated the alignment between each of the 74 HCP-D variability matrices and the reference using Spearman's rank correlations (Fig. 2b). We used a general additive model (GAM) to measure associations between variability gradient alignment and age while controlling for sex and in-scanner motion. Next, we introduced a gradient slope to quantify the connectional gradient through linear regression on the 21 ranked network-level variability values (Fig. 2d), and assessed the development of the variability gradient slope using the same GAM.

We observed high variability in association network connections in both HCP-YA and HCP-D datasets (Fig. 1a–f). We averaged FC variabilities at the network level and ranked them for each network separately, and defined this as the 'connectional variability gradient'. The ranking revealed decreasing individual variability from within-network edges to those between association networks and further to sensorimotor-association connections (Fig. 1g–h). Next, we examined the whole-brain connectome by ranking the FC variability of all network-level connections. Within-association network edges peaked on this gradient, while sensorimotor-other network edges were at the base (Fig. 1i–j). Additionally, we found a significant positive correlation between age and variability gradient alignment (partial R2 = 0.29, P = 1.23×10-6, Fig. 2c), and the slope of the network-level FC variability gradient significantly declined during youth (partial R2 = -0.82, P = 1.78×10-67, Fig. 2e).

We identified a connectional gradient in the edge-level FC variability which declined continuously along an axis from the edges among association networks to those that connect the sensorimotor and association networks. Furthermore, the connectional variability gradient matured into an adult-like pattern with a flatter gradient slope during youth.

Early life, Adolescence, Aging ²

Connectivity (eg. functional, effective, structural) ¹

ADULTS

Data analysis

Development

FUNCTIONAL MRI

MRI