Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 May 1.
Published in final edited form as: Biol Psychiatry. 2023 Dec 25;95(9):828–838. doi: 10.1016/j.biopsych.2023.12.019

Triple interactions between the environment, brain, and behavior in children: An ABCD study

Dongmei Zhi 1,#, Rongtao Jiang 2,#, Godfrey Pearlson 3, Zening Fu 4, Shile Qi 5, Weizheng Yan 4, Aichen Feng 6,7, Ming Xu 6,7, Vince Calhoun 4,*, Jing Sui 1,4,*
PMCID: PMC11006588  NIHMSID: NIHMS1955914  PMID: 38151182

Abstract

Background:

Environmental exposures play a crucial role in shaping children’s behavioral development. However, the mechanisms by which these exposures interact with brain functional connectivity and influence behavior remain unexplored.

Methods:

We investigated the comprehensive environment-brain-behavior triple interactions through rigorous association, prediction, and mediation analyses, while adjusting for multiple confounders. Particularly, we examined the predictive power of brain functional network connectivity (FNC) and 41 environmental exposures for 23 behaviors related to cognitive ability and mental health in 7655 children selected from the ABCD study at both baseline and follow-up time.

Results:

FNC demonstrated more predictability for cognitive abilities than mental health, with cross-validation from UK Biobank study (n= 20852), highlighting the importance of thalamus and hippocampus in longitudinal prediction, while FNC+environment demonstrated more predictive power than FNC in both cross-sectional and longitudinal prediction of all behaviors, especially for mental health (r = 0.32~0.63). We found family and neighborhood exposures were common critical environmental influencers on cognitive ability and mental health, which can be mediated by FNC significantly. Healthy perinatal development was a unique protective factor for higher cognitive ability, whereas sleep problems, family conflicts and adverse school environments specifically increase risk of mental health.

Conclusions:

This work revealed comprehensive environment-brain-behavior triple interactions based on ABCD study, identified cognitive control and default mode networks as the most predictive functional networks for a wide repertoire of behaviors, and underscored the long-lasting impact of critical environmental exposures on childhood development, in which sleep problem was the most prominent factor affecting the mental health.

Keywords: Environmental exposure, ABCD, cognition, mental health, functional network connectivity, individualized prediction, mediation analysis

INTRODUCTION

Adolescence has long been known as an important time for developing cognitive skills and is also a period when most mental disorders initially manifest (1). Moreover, the children’s brain undergoes a growth spurt in structural and functional maturation, building a foundation for behavioral outcomes (2). Despite the well-known fact that environmental exposures play a critical role in influencing behaviors, we have very limited understanding of how these exposures interact with the brain and in turn shape our behaviors, especially during adolescence (3).

Recent years have seen rapid growth of interest in examining the neural basis through which environmental exposures can have enduring effects on behaviors (4–6). Nevertheless, research in this context has been dominated by brain structural features in children (4, 7), with less focus on brain functional measures. This is partly due to the easier access and feasibility of structural magnetic resonance imaging (MRI) scans, though functional features have shown more predictive power for high-order cognition (8). As a measure that can well characterize individual variability (9, 10), functional network connectivity (FNC) quantifies the temporal statistical dependencies between functional activation among different brain networks, providing insightful correlations between heterogeneous personal behaviors with the brain (11). The limited evidence available on relationships between brain, environment, and behaviors (5, 12) calls for a more comprehensive investigation into how different brain functional networks may connect or even mediate associations between diverse environmental exposures and a wide range of childhood behaviors.

Furthermore, previous environment-behavior association studies often focus on a single behavioral domain (e.g., cognitive ability) or only a few environmental domains (4, 6, 13), thus not capturing the broader relationships between various factors. For example, sufficient sleep (14) and high socioeconomic status (4) may improve cognitive ability, while long-term exposure to air pollution (15) and low family income (5) can have the opposite impact. Perinatal factors such as birth weight and postnatal growth have been associated with only cognitive ability but not mental health (16), whereas higher community noise impaired both (17). Additionally, Alnæs et al. revealed three brain anatomical covariation patterns of perinatal complications, sociocognitive stratification, and urbanicity (18), and Modabbernia et al. found that socioeconomic circumstances, perinatal events, and cognition had the most reliable covariations with multiple brain measures (12). However, a comprehensive exploration of environment-brain-behavior triple interactions is still lacking, which may provide constructive insights into risky and protective environmental exposures for children’s brain and behavioral development.

Beyond association analysis, by revealing the predictive power of FNC patterns and environmental exposures on multiple behaviors thereby establishing their utility in longitudinal individual evaluation, we may identify imaging signatures with promising translational impact that could be missed by existing studies (19, 20). We applied NeuroMark (9), a fully automated independent component analysis (ICA) framework to the resting-state functional MRI (fMRI) data to generate the data-adaptive FNC patterns that serve as the predictive brain signatures. Further exploration of their mediating effects from environmental exposures to cognition and mental health may facilitate the elucidation of the neural underpinnings of positive and negative developmental trajectories in children.

Specifically, we included 7655 typically developing children from the ABCD study with 41 environmental exposures across 5 domains (spanning perinatal, family, school, neighborhood and individual lifestyle (12)), and 23 behaviors of two types—10 cognitive abilities and 13 mental health measures at both baseline and longitudinally. Fig. 1 display the whole research design including 4 steps (more details in Fig. S1):

Fig. 1. Research design.

Fig. 1.

Open in a new tab

Environment-brain-behavior triplet correlations 1) between environmental exposures and whole-brain functional network connectivity (FNC); and 2) between environmental exposures and multi-behaviors (baseline and 1-year, 2-year follow-up). 3) Building behavior (baseline and follow-up) prediction models using FNCs, environmental exposures, and their combinations respectively, identifying the most predictive functional network modules and environmental exposures. 4) Examining the mediating effect of the identified predictive FNCs for significant associations between environmental exposures and two types of behaviors.

  1. Identify the FNCs most susceptible to environmental influences and determine the environmental exposures impacting most FNCs by association analysis.

  2. Build environment-behavior association maps and identify the shared and unique environmental exposures affecting cognitive abilities and mental health, at both baseline and longitudinally.

  3. Characterize dominating functional networks, FNC signatures, and critical environmental exposures that support individual-level prediction of cognitive abilities and mental health (baseline and longitudinally), which were also externally validated via the UK Biobank data.

  4. Examine whether and to what extent this ‘predictome’ (21) FNC signatures mediates the environment-behavior associations.

METHODS AND MATERIALS

Participants from the ABCD study

This study used data from a population-based sample of 9–10-year-olds from 21 U.S. study sites in the ongoing ABCD study (release 3.0), including neuroimaging and behavioral data collected at baseline and longitudinally (22). Informed written consent was obtained from children and their parents, with ethical approval from each research site’s Institutional Review Boards. The current study included 7655 participants after rigorous data quality control (Fig. S2) (23).

FMRI data acquisition and processing

Resting-state fMRI data from the ABCD study were acquired and preprocessed as previously described and detailed in the Supplementary file (22). The preprocessed data were decomposed into 53 subject-specific independent components (ICs, Fig. S3) and their corresponding time courses via a spatially constrained single-subject ICA method with the Neuromark_fMRI_1.0 template as spatial references (available in GIFT at http://trendscenter.org/software/gift) (9). Paired correlations of the ICs were calculated by Pearson correlation and transformed using Fisher-Z transformation, where the upper triangle elements of the FNC matrix (53 × 52/2 = 1378) were extracted for further analysis.

Phenotypic measures

We examined a total of 41 summarized environmental exposures as done by Modabbernia, et al. (12), consisting of 5 domains (Table S2): perinatal/early development events (n = 13), life events/lifestyle (n = 7), family characteristics (n = 12), neighborhood (n = 6) and school environments (n = 3). For cognitive abilities, we used scores from a well-validated National Institute of Health Toolbox (n = 10) at baseline and 2-year follow-up (24). For mental health, we used the parent-reported Child Behavior Checklist (CBCL, n=11) and additional two metrics (Prodromal Symptoms and Subsyndromal Mania) at baseline, and 1-year follow-up (25), which were selected via balancing the sample size and information completeness (Table S3, S5, S6).

Baseline and longitudinal association analysis between environment and brain, behavior

Linear mixed-effect models (LMM) were adopted to examine the associations between 41 environmental exposures and 1378 whole-brain FNC pairs, 10 cognitive abilities, and 13 mental health measures at baseline (Fig. 1). Specifically, each FNC edge, cognitive ability or mental health was modeled as the dependent variable, environmental exposures and the nuisance covariates were modeled as fixed effects, while the family structure nested within sites was modeled as random effects (13). The nuisance covariates include age, sex, body mass index (BMI), puberty, ethnicity, handedness, and mean frame-wise displacement (FD, for the FNC analyses only), where sex, ethnicity and handedness were coded as dummy variables. The correlation r-value, t-statistic, and effect size Cohen’s d were estimated for each LMM model to reflect the association of specific environmental exposure with the dependent variable. The same analytical framework was employed to investigate the longitudinal associations between environmental exposures and follow-up cognitive abilities and mental health measures separately, but additionally included the baseline outcomes as a covariate.

Baseline and longitudinal behavior prediction using FNC and environmental exposures

Beyond brain-behavior correlations, to identify key predictive FNC signatures that support individual-level prediction of behaviors, we built FNC-based predictive model using partial least squares regression (PLSR) for each cognitive ability and mental health metric. 10-fold nested cross-validation with 200 random loops was utilized to avoid circularity bias. Model performance was assessed by averaging Pearson’s correlation and coefficient of determination (COD) across 200 repetitions between observed and predicted scores for all subjects (11, 26). We then evaluated the most average contributing FNC at edge, node and network levels across all repetitions (27, 28). Furthermore, to examine the promotion degree of environmental exposures on behavior prediction, we constructed independent prediction models using only FNC, only environment and FNC + environment. The most critical environmental exposures contributing to predictions were estimated and compared between cognitive abilities and mental health.

To examine the predictability of follow-up cognitive abilities and mental health by FNC, environment, or FNC+ environment at baseline, we implemented the above predictive procedure, where the baseline outcomes and confounding variables in the association analysis were set as covariates in the predictive model (29).

Cross-validation using the UK Biobank dataset

To further validate the generalizability of the FNC-based behavior prediction, 20,852 participants were selected from UK Biobank for cross-dataset validation, who have both FNC and fluid intelligence after rigorous quality control (Fig. S4, S5). The same predictive procedure for fluid intelligence was constructed within and across the ABCD and UK Biobank datasets.

Mediation Analysis

Standard mediation analysis in R toolbox (30) was used to examine whether and to what extent the ‘predictome’ FNC signatures mediated the significant environment-behavior associations. To distinguish connectivity features with positive and negative contributions in the predictive models (31), the predictive FNCs were separated into positive-weighted and negative-weighted connectivity sets. A standard three-variable path model was implemented for every mediation analysis (32) adjusting for the same confounding variables in the association analysis, where the predictor was an environmental exposure, the outcome variable was one cognitive ability or mental health measure, and the mediator was either positive-weighted or negative-weighted FNC.

Multiple-comparison correction

We performed FDR correction (q < 0.05) to determine significant environment-brain associations, and the p-value threshold was 2.3×10−4 with a total of 1378*41 tests. Bonferroni correction was used to determine significant environment-behavior associations, and the p-value threshold was 1.0×10−3 with a total of 41 tests for each behavior. For prediction, we performed 10000 permutation tests for the prediction accuracy, and the p-value threshold was 1.0×10−4. Moreover, we used 95% bias-corrected confidence intervals with 10000 bootstrap tests in the mediation analysis, and the p-value threshold was 0.05. When reporting p-values, the uncorrected p-values were reported.

RESULTS

Subcortical network was the most vulnerable to environmental exposures

As shown in Fig 2A, “family income” (rabs = 0.05~0.09) and “caregiver education” (rabs = 0.04~0.07) are top 2 ranked exposures influencing more FNC numbers, which manifest similar FNC architectures, especially the cross-module connections. Meanwhile, thalamus (rabs = 0.04~0.08) stands out as the brain region with the most environmental susceptibility, followed by precuneus (rabs = 0.04~0.08) and superior temporal gyrus (rabs = 0.04~0.09, Fig. 2B). Specifically, environmental exposures influence most on FNC connections in thalamus–temporal gyrus and thalamus-postcentral gyrus (Fig. 2C). From the perspective of function network module, subcortical network stands out with the highest vulnerability to environment, mainly in domains of family, perinatal and neighborhood exposures (Fig. 2D).

Fig. 2. Summary of associations between whole-brain FNC with 41 environmental exposures.

Fig. 2.

Open in a new tab

(FDR corrected, p < 0.05). (A) Ranking of environmental factors influencing more FNC pairs, and the top two FNC patterns. (B) Ranking of FNC nodes associated with more exposures, and the top 5 nodes (independent component, ICs). (C) Mapping of the number of environmental exposures correlated with each FNC, and within each network module, where the top FNC nodes are illustrated. (D) Distribution of environmental exposures correlated with each FNC module. Abbreviation: SCN, subcortical network; SMN, somatomotor network; CCN, cognitive control network; DMN, default mode network; CBN, cerebellum network.

Critical environmental exposures associated with behavior at baseline and longitudinally

We next tested the environment–behavior associations to unveil the diversity in the impact of environmental exposures on cognitive ability and mental health at both baseline and longitudinally. At baseline (Fig. 3A), nine environmental exposures were commonly correlated with 80% or more behaviors, in which “family income”, “caregiver education”, “caregiver marital status”, “neighborhood security”, and “area deprivation index” correlate with nearly all behaviors significantly, which belong to family and neighborhood domains. In contrast, eleven environmental exposures were linked to more mental health problems while two were linked to more cognitive abilities. Specifically, “months breastfed” and “delayed verbal development” were uniquely associated with cognitive abilities, while “sleep problems”, “family conflict parents”, “school environment”, “secondary caregiver warmth”, “maternal substance use” and “screen use during weekdays” uniquely linked to psychiatric problems. Detailed association maps for the two most representative behaviors, Cognition Total Composite and CBCL Total Problems (Fig. 3B), and other behaviors were illustrated in Fig. S6, S7. Notably, these significant associations did not change appreciably after controlling for the participants’ diagnosis status of mental disorders (Table S8, S9).

Fig. 3. Environment-behavior association.

Fig. 3.

Open in a new tab

The summarized correlation mapping between 5 domains of environmental exposures and 23 behaviors, at both baseline (A) and longitudinally (C), where the triangle, square and round black dots denote the mental health-specific, cognition-specific and shared environmental exposures respectively. The bar above denotes the number of environmental exposures significantly correlated with each of the 23 behavioral items (Bonferroni corrected, p < 0.05). (B) Environmental exposures significantly associated with Cognition Total Composite and Total Problems CBCL Syndrome respectively, where the top 10 most associated exposures were annotated, and the dots outside the dotted line represent significant correlations passing the Bonferroni correction (p < 0.05), with Cohen’s d was displayed.

When summarizing the findings at the environmental level, we observed that three psychiatric assessments, i.e., “CBCL Total Problems”, “Withdrawn/Depressed Syndrome”, and “Social Problems” were significantly associated with the most environmental exposures (31 out of 41). Similarly, the “Cognition Total”, “Oral Reading”, “Crystallized Composite” were linked to the most environmental factors, instead, “Pattern Comparison” and “Flanker Inhibitory Control and Attention” were linked to the least number of exposures.

Most importantly, at follow-up (Fig. 3C), it is remarkable that “sleep problems” showed prominent associations with all mental health problems. Five baseline environmental exposures were commonly correlated with more than five follow-up behaviors, primarily falling in family and neighborhood domains, i.e., “family income”, “caregiver education”, “severe financial difficulty”, “caregiver marital status”, and “area deprivation index”. In contrast, five exposures uniquely linked to more 1-year-later mental health problems, especially “sleep problems”, “family conflict”, “parental psychopathology”, and “number of people living”, while two exposures specifically linked to more 2-year-later cognitive abilities, i.e., “delayed verbal development” and “month breastfed”. Interestingly, school domains and multiple pregnancy measures link to almost all baseline mental problems, especially 1-year-later “Prodromal Symptom”. Overall, “Prodromal Symptom”, “Sub-syndromal Mania”, and “Withdrawn/Depressed Syndrome” were three follow-up mental problems most associated with baseline exposures, and so did the “Picture Vocabulary” in cognitive ability.

The dominating role of environmental exposures in predicting multiple behaviors

Results demonstrated that all 10 cognitive abilities and 13 mental health measures can be significantly predicted by whole-brain FNC (p < 1.0×10−4, 10000 permutation tests, Table S10), where the prediction accuracies for most cognitive abilities were much higher than those for mental health. Similar results were revealed by the connectome-based predictive modeling (31) and random forest (33) models (Fig. S10). Specifically, Cognition Total Composite showed the highest predictability among all behaviors (r = 0.39, COD = 0.14, Fig. 4A, B, C). Notably, these predictions remained significant even controlling for multiple covariates, and data harmonization via ComBat (34) (Fig. 4B).

Fig. 4.

Fig. 4.

Open in a new tab

The prediction results for behaviors. Prediction of Cognition Total Composite based on FNC only (A) and after adjusting for covariates (B) including site, harmonization by Combat, mean frame displacement (FD), pubertal stage, age, sex, BMI, and handedness, across 200 repetitions of 10-fold cross-validation. (C) The top 1% predictive connections and network modules for Cognition Total after averaging across 2000 cross-validation rounds. (D) Cross-validation between the ABCD and UK Biobank datasets for prediction of fluid intelligence using FNC. Comparison of prediction accuracy using only FNC, only environment, or their combination for (E)10 types of cognitive abilities and (F)13 types of mental measures. (G) The summarized top contributing FNC modules and environmental exposures for predicting cognition or mental health. *, p < 0.05, **, p < 0.01, ***, p < 0.001, FDR corrected. SCN, subcortical network; SMN, sensorimotor network; VIS, visual network; CCN, cognitive control network; DMN, default mode network; CBN, cerebellum network.

Obviously, environmental exposures themselves can achieve much higher prediction accuracy than using only FNCs for most cognitive abilities (rmax = 0.47, COD = 0.22) and all mental health (rmax = 0.63, COD = 0.40, Table S10) with p < 1.0×10−5 (10,000 permutations). Noted that their combination further improved the prediction accuracy of multiple behaviors (Fig. 4), whose accuracy were significantly higher than using only FNC (p < 0.001, FDR corrected), and such an improvement was more remarkable in mental health (r = 0.32~0.63, Δr = 0.17~0.53, Fig. 4E) than in cognitive ability (r = 0.12~0.47, Δr = 0.00~0.08, Fig. 4F) at p < 1.0×10−5 (10,000 permutations), suggesting the prominent role of environmental exposures on development risks of adolescent mental health.

Fig. 4G summarized the most contributing functional network modules and environmental exposures for either 10 cognitive abilities or 13 mental health measures. Results showed that FNCs in CCN-CCN, DMN-DMN, and DMN-CCN showed the most predictive power for cognitive abilities, while FNCs in SCN-SCN, SMN-SCN, and CCN-CCN contributed most to mental health. This suggests that CCN-CCN within-network connections are shared crucial predictors for both domains. In contrast, FNCs within DMN contributed more to cognition, while FNCs within SCN primarily contributed to mental health.

Furthermore, “caregiver education”, “caregiver marital status”, “delayed verbal development”, “family income”, and “school environment” were the top 5 environmental exposures for cognitive prediction; whereas “family conflict parent/youth”, “severe financial difficulty”, “sleep problems”, and “maternal medical conditions” were the top 5 predictive environmental adversities for mental health, which are highly overlapped with those most associated exposures in Fig. 3A. For each metric, the prediction results, FNC signatures and critical environmental exposures were provided in Table S10, Fig. S11–S13, with high stability of the predictive weights (Fig. S14).

Cross-validation using the UK Biobank dataset

The fluid intelligence can be significantly predicted both within ABCD (r = 0.25, COD = 0.05) and UK Biobank (r = 0.26, COD = 0.06) datasets using FNC (Fig. 4D). More importantly, when directly applied the FNC-based prediction model trained on ABCD to UK Biobank, significant predictions can still be achieved (r = 0.08, p < 1.0 × 10−30), and vice versa (r = 0.09, p < 1.0×10−15).

Longitudinal behavior prediction using FNC and environmental exposures

Results showed that four 2-year-later cognitive abilities can be significantly predicted with r > 0.13 (p < 1.0×10−5, Table 1) using only FNC, especially the Picture Vocabulary (r = 0.27) and Crystallized Composite (r = 0.22). However, Prodromal Symptom was the most predictable (r = 0.12, p < 1.0×10−5) in 1-year-later mental health. Similar to baseline prediction, longitudinal prediction accuracy with only environment, or FNC + environment was much higher than only FNC, especially for mental health. Specifically, the prediction accuracy for “Picture Vocabulary” increased most in all cognitive abilities, from r = 0.27 to r = 0.44; while the “Thought Syndrome” increased most in all mental health, i.e., from r = 0.05 to r = 0.45.

Table 1.

Longitudinal prediction of follow-up behaviors using FNC and environmental exposures.

Only FNC Only Environment FNC + Environment Most predictive features
Prediction accuracy R P R P R P Top 5 fMRI ICs Top 5 exposures
Thought 0.05 ± 0.01 < 10−3 0.46 ± 0.01 <10−5* 0.45 ± 0.01 <10−5* Caudate
Precuneus
Superior frontal gyrus
Postcentral gyrus
Left postcentral gyrus		 Family conflict parents
Number of people living
School environment
Maternal medical cond.
Sleep problems
TotProb 0.04 ± 0.01 < 0.05 0.44 ± 0.01 <10−5* 0.42 ± 0.01 <10−5* Hippocampus
Precuneus
Middle temporal gyrus
Right inferior frontal gyrus
Middle cingulate cortex		 Family conflict parents
Number of people living
School environment
Maternal medical cond.
Sleep problems
Attention 0.06 ± 0.01 <10−4* 0.36 ± 0.01 <10−5* 0.37 ± 0.01 <10−5* Middle temporal gyrus
Middle cingulate cortex
Middle temporal gyrus
Left postcentral gyrus
Inferior parietal lobule		 Family conflict parents
Number of people living
School environment
Delayed verbal development
Maternal medical cond.
RuleBreak 0.05 ± 0.01 < 0.01 0.35 ± 0.01 <10−5* 0.35 ± 0.01 <10−5* Thalamus
Caudate
Hippocampus
Middle temporal gyrus
Sub/hypo-thalamus		 Family conflict parents
Family conflict youth
Severe financial difficulty
Caregiver marital status
Family income
Prodromal 0.12 ± 0.01 <10−5* 0.29 ± 0.01 <10−5* 0.28 ± 0.01 <10−5* Caudate
Precuneus
Superior temporal gyrus
Middle occipital gyrus
Cerebellum		 Family conflict youth
Severe financial difficulty
Family income
Caregiver education
Caregiver marital status
WithDep 0.04 ± 0.01 < 0.05 0.29 ± 0.01 <10−5* 0.28 ± 0.01 <10−5* Precuneus
Precentral gyrus
Supplementary motor
Right inferior frontal gyrus
Paracentral lobule		 Family conflict parents
Number of people living
Days of physical activity
Severe financial difficulty
Sleep problems
PicVocab 0.27 ± 0.01 <10−5* 0.43 ± 0.02 <10−5* 0.44 ± 0.02 <10−5* Hippocampus
Precuneus
Superior med frontal gyrus
Superior parietal lobule
Cerebellum		 Delayed verbal development
Severe financial difficulty
Family income
Caregiver education
Caregiver marital status
Crystal 0.22 ± 0.01 <10−5* 0.38 ± 0.03 <10−5* 0.37 ± 0.03 <10−5* Thalamus
Hippocampus
Putamen
Superior med frontal gyrus
Inferior frontal gyrus		 Delayed verbal development
School environment
Family income
Caregiver education
Caregiver marital status
Reading 0.13 ± 0.01 <10−5* 0.24 ± 0.03 <10−5* 0.24 ± 0.03 <10−5* Thalamus
Hippocampus
Superior parietal lobule
Posterior cingulate cortex
Superior temporal gyrus		 Delayed verbal development
School environment
Severe financial difficulty
Positive school involvement
Caregiver education
Picture 0.13 ± 0.01 <10−5* 0.20 ± 0.01 <10−5* 0.21 ± 0.01 <10−5* Precuneus
Superior temporal gyrus
Paracentral lobule
Middle occipital gyrus
Cerebellum		 School environment
School disengagement
Severe financial difficulty
Family income
Caregiver marital status
Open in a new tab

Note: The longitudinal prediction accuracy by only functional network connectivity (FNC), only environment, and combination of FNC and environmental exposures were represented by the mean + standard deviation across 200 repetitions of ten-fold cross-validation, and the top five independent components (ICs), and environmental exposures were displayed. The star indicates that the longitudinal prediction is significant across all 200 repetitions (p < 1×10−4).

The top five most predictive IC nodes and environmental exposures were listed in Table 1, where the hippocampus, precuneus, thalamus and caudate played a prominent role in the longitudinal behavior prediction. Consistent with the baseline critical exposures, we found that family conflict and sleep problems uniquely contributed most to mental health, while delayed verbal development was specifically prominent for cognitive abilities in the longitudinal prediction. Particularly, number of people living played a more important role in longitudinal prediction than cross-sectional prediction of mental health, as increased number of people living at baseline can decrease mental health problems 1 year later. Moreover, family income and severe financial difficulty were shared high-contributing predictors for longitudinal cognition and mental health.

Mediation analysis

Results demonstrated that positive-weighted predictive FNCs mediated more cognitive abilities, whereas negative-weighted predictive FNCs mediated more mental health (Fig. 5). Specifically, Crystallized Composite, Cognition Total, and Picture Vocabulary can be mediated by positive-weighted FNCs from the most number of environmental exposures. Similar condition exists for Prodromal symptom and Attention problem in mental health, but by negative-weighted FNCs. In terms of common and unique exposures linked to two types of behaviors, the shared influential exposures were primarily in domains of family and neighborhood. In contrast, “sleep problems” and “family conflict parents” specifically affected mental health, whereas “maternal age”, “months breastfed” and “delayed verbal development” specifically affected cognitive abilities, mainly in the domain of “perinatal/early development”. The significant mediation results for all cognitive abilities and mental health were provided in Table S11, S12 and Fig. S15, S16.

Fig. 5. The mediating effects.

Fig. 5.

Open in a new tab

We further tested whether the identified predictive FNCs can mediate associations from environment to multiple behaviors. For each of (A) cognitive and (B) mental health measures, the radar maps showed the number of environmental exposures that can be significantly mediated by FNC (10,000 bootstrap tests at p < 0.05). Similarly, for each environmental exposure, radar maps display how many (C) cognitive metrics or (D) mental health measures can be significantly mediated by FNCs, where top factors are highlighted in bold texts. (E) The common and specific environmental influencers between cognitive abilities and mental health in mediations, the more mediation involved, the bigger text size.

DISCUSSION

This work revealed comprehensive environment-brain-behavior triple interactions within a large longitudinal sample of typically developing children. We found common critical environmental exposures that have substantial and long-lasting importance on cognitive ability and mental health, which mainly fall in family and neighborhood domains, especially the family domain linked most to children’s brain function (Fig. 2D). This aligned with the concept that high socioeconomic status (family income and caregiver education) reflects conducive home learning environment and high-quality parent-child interactions (35), while such long-term stimulation may effectively support children’s functional brain development (36), and be linked to behavioral development (37, 38). Extending previous studies, we underscore the importance of the “area deprivation index”, a measure of neighborhood-level socioeconomic status (39) and “neighborhood security”, which can be changed practically by improving public environmental sanitation or enhancing children’s safety education to reduce risk factors rapidly for certain mental health.

For unique influencers, healthier perinatal exposures such as longer month breastfed and earlier verbal development promise better cognitive abilities in adolescents, which also shows strong lasting links to the children’s brain connectome in all network modules. This may be attributed that infant brain is marked by rapid development of neurons and explosive growth of cortex, thus are highly vulnerable to environmental exposures (40), suggesting that healthy perinatal development is irreplaceable protective factor for evolving cognitive function. Whereas more sleep problems, family conflict and adverse school environment were especially link to increased risk of mental health (Fig. 3), at both baseline and longitudinally. Previous studies have found that sleep problems were associated with high emergence of depressive problems (6, 13, 14). Similarly, family conflict was identified to be one of the most robust risk factors for suicidality (41), which resonated with our results linking family conflict with the emergence of baseline or follow-up mental health problems.

More importantly, environmental exposures play a much more dominating role than brain connectivity in predicting all behaviors, especially in longitudinal prediction; among which sleep problems emerge as the most prominent factor affecting all mental health. In longitudinal prediction (Table 1), we again found “family income” is the shared key predictor; “family conflict” and “sleep problems” uniquely contributed most for predicting 1-year-later mental health, while delayed verbal development specifically works for 2-year-later cognitive abilities prediction, highlighting their respectively dominating role in the two types of behavioral development (6, 42). Interestingly, unlike the association results, we discovered more “number of people living” promised less follow-up mental health problems in middle childhood, suggesting that larger number of household members, especially older siblings, is a protective factor for reducing risks of psychopathology due to more family member communication and interaction (43). Furthermore, school environment appears in top-5 predictors in 6 out of 10 follow-up behaviors, implicating that constructing positive school environment could be one of the most effective public health interventions for reducing psychiatric risk and improving cognitive ability in practice (44).

Intriguingly, when looking into the predominant exposures, we found cognitive abilities in middle childhood were influenced most by environmental exposures that are relatively fixed such as parents’ education, perinatal exposures and family income. However, lifestyle and school environmental exposures that are flexible and changeable in childhood influenced mental health substantially. Particularly, more sleep problems, family conflict and adverse school environment increase the risk for baseline and 1-year-later mental health. Note that these exposures can be modified immediately by improving self-sleep habits, providing harmonious family relationships, or creating a positive school environment.

In contrast, brain connectome can mainly predict cognitive abilities and baseline mental health, but shows weaker links to 1-year-later mental health prediction, with the exception of prodromal syndrome, which may be due to the fact that mental health is measured through parental observation and consequently may be influenced by participants’ subjective feelings (25). FNCs in DMN and CCN showed the most predictive power for cognitive ability, especially the important role of DMN in longitudinal prediction (Fig. 4, Table 1). This is not surprising since DMN and CCN have been considered widely involved in different aspects of cognition (45–47). In comparison, FNCs in SCN contributed the most to mental health. Abnormalities in DMN have been consistently revealed to be implicated in adult psychiatric disorders (48). Our results further revealed the critical role of SCN in mental health problems in adolescents. The SCN has been implicated in impulsivity, attention deficits, and emotional regulation (49, 50). One interesting finding is that hippocampus and thalamus manifest as prominent brain nodes in predicting 2-year-later comprehensive reading and crystal intelligence. Specifically, hippocampus plays a crucial role in long-term episode memory, which can mediate behaviors that allow learning to take place (51), so as to contribute most to the follow-up cognitive ability. Thalamus that conveys the subcortical-cortical information (52) acts as a bridge between sensory perception and cognition (53), involved in wide deficits in behaviors. Moreover, the predictive FNCs significantly mediate the environment-behavior associations, implying the potential environment-brain-behavior plasticity loops (4).

There are several limitations in this study. First, there may exist potential collinearity among environmental exposures or behavior outcomes, deserving further investigation via exploratory factor analysis in future (54), however, this is not the emphasis of the current study. Second, the study is primarily based on association analysis and does not allow for causal inferences about the environment-brain-behavior relationships without further validation using randomized controlled trials. Nevertheless, it offers a critical first step for future studies to examine the neurobiological mechanisms underlying behaviors. In addition, FNC matrices for adolescents from the ABCD datasets were estimated using the NeuroMark template derived from adults, which may underestimate the divergence of spatial network distribution configurations between adolescents and adults (55–58). Nevertheless, this concern is lessened given the generalization of models between the ABCD and UKB datasets, and the applicability of the NeuroMark template across different age groups (59–61). Individualized atlas developed for different age groups deserve further exploration. Furthermore, it is worth noting that ethnicity was only considered as a covariate. Future research can establish ethnicity-specific models to examine the disparities of the environment-brain-behavior relationships across different ethnicities.

Collectively, the present study unveiled comprehensive environment-brain-behavior triple interactions based on ABCD study at both baseline and longitudinally, identified CCN, DMN and SCN as the most predictive functional networks for a wide repertoire of behaviors, emphasized the long-term importance of critical environmental exposures to promote brain and behavioral development in children, especially the attainable targets with family conflict, sleep quality, school and neighborhood environments to promote the healthy development of adolescents.

Supplementary Material

2
3

KEY RESOURCES TABLE.

Resource Type Specific Reagent or Resource Source or Reference Identifiers Additional Information
Add additional rows as needed for each resource type Include species and sex when applicable. Include name of manufacturer, company, repository, individual, or research lab. Include PMID or DOI for references; use “this paper” if new. Include catalog numbers, stock numbers, database IDs or accession numbers, and/or RRIDs. RRIDs are highly encouraged; search for RRIDs at https://scicrunch.org/resources. Include any additional information or notes if necessary.
Antibody N/A
Bacterial or Viral Strain N/A
Biological Sample N/A
Cell Line N/A
Chemical Compound or Drug N/A
Commercial Assay Or Kit N/A
Deposited Data; Public Database Human studies; Both male and female The ABCD study (https://abcdstudy.org);
The UK Biobank study.
Genetic Reagent N/A
Organism/Strain N/A
Peptide, Recombinant Protein N/A
Recombinant DNA N/A
Sequence-Based Reagent N/A
Software; Algorithm GIFT toolbox;
R software;
MATLAB 2017		 GIFT toolbox: https://trendscenter.org/software/gift/;
R software:https://www.r-project.org/;
MATLAB 2017:https://www.mathworks.com/products/matlab.html.
Transfected Construct N/A
Other
Open in a new tab

Acknowledgments

This work was supported by the Natural Science Foundation of China (62373062, 82022035), the China Postdoctoral Science Foundation (2022M710434), the National Institutes of Health grants (R01EB005846, R01MH117107, R01MH118695) and National Science Foundation (2112455).

Footnotes

Competing interests

The authors report no biomedical financial interests or potential conflicts of interest.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Data and code availability

Neuroimaging and behavioral data from the ABCD dataset are obtained from https://nda.nih.gov/abcd with the approval of the ABCD consortium. The data from UK Biobank used in this study can be publicly accessible via their standard data access procedure at https://www.ukbiobank.ac.uk/. Matlab and R scripts written to perform most of the analyses are available from the authors upon request.

References

  • 1.Gargano LM, Locke S, Li J, Farfel MRJPr (2018): Behavior problems in adolescence and subsequent mental health in early adulthood: results from the World Trade Center Health Registry Cohort. Pediatric Research. 84:205–209. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.James SL, Abate D, Abate KH, Abay SM, Abbafati C, Abbasi N, et al. (2018): Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017. The Lancet. 392:1789–1858. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Rosenzweig MRJDn (2003): Effects of differential experience on the brain and behavior. Developmental Neuropsychology. 24:523–540. [DOI] [PubMed] [Google Scholar]
  • 4.Noble KG, Houston SM, Brito NH, Bartsch H, Kan E, Kuperman JM, et al. (2015): Family income, parental education and brain structure in children and adolescents. Nature Neuroscience. 18:773–778. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Marshall AT, Betts S, Kan EC, McConnell R, Lanphear BP, Sowell ERJNm (2020): Association of lead-exposure risk and family income with childhood brain outcomes. Nature Medicine. 26:91–97. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Gong W, Rolls ET, Du J, Feng J, Cheng WJNC (2021): Brain structure is linked to the association between family environment and behavioral problems in children in the ABCD study. Nature Communications. 12:1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Ge R, Sassi R, Yatham L, Frangou S (2022): Neuroimaging profiling identifies distinct brain maturational subtypes of youth with mood and anxiety disorders. bioRxiv. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Chen J, Tam A, Kebets V, Orban C, Ooi LQR, Asplund CL, et al. (2022): Shared and unique brain network features predict cognitive, personality, and mental health scores in the ABCD study. Nat Commun. 13:2217. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Du Y, Fu Z, Sui J, Gao S, Xing Y, Lin D, et al. (2020): NeuroMark: An automated and adaptive ICA based pipeline to identify reproducible fMRI markers of brain disorders. NeuroImage: Clinical. 28:102375. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Smith SM, Beckmann CF, Andersson J, Auerbach EJ, Bijsterbosch J, Douaud G, et al. (2013): Resting-state fMRI in the human connectome project. Neuroimage. 80:144–168. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Li J, Bzdok D, Chen J, Tam A, Ooi LQR, Holmes AJ, et al. (2022): Cross-ethnicity/race generalization failure of behavioral prediction from resting-state functional connectivity. Science Advances. 8:eabj1812. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Modabbernia A, Janiri D, Doucet GE, Reichenberg A, Frangou SJBp (2021): Multivariate patterns of brain-behavior-environment associations in the adolescent brain and cognitive development study. Biological Psychiatry. 89:510–520. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Cheng W, Rolls E, Gong W, Du J, Zhang J, Zhang X-Y, et al. (2021): Sleep duration, brain structure, and psychiatric and cognitive problems in children. Molecular Psychiatry. 26:3992–4003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Yang FN, Xie W, Wang Z (2022): Effects of sleep duration on neurocognitive development in early adolescents in the USA: a propensity score matched, longitudinal, observational study. The Lancet Child & Adolescent Health. 6:705–712. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Clifford A, Lang L, Chen R, Anstey KJ, Seaton AJEr (2016): Exposure to air pollution and cognitive functioning across the life course–a systematic literature review. Environmental Research. 147:383–398. [DOI] [PubMed] [Google Scholar]
  • 16.Huang C, Martorell R, Ren A, Li ZJIjoe (2013): Cognition and behavioural development in early childhood: the role of birth weight and postnatal growth. International Journal of Epidemiology. 42:160–171. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Raess M, Brentani AVM, Flückiger B, de Campos BLdA, Fink G, Röösli MJEi (2022): Association between community noise and children’s cognitive and behavioral development: A prospective cohort study. Environment International. 158:106961. [DOI] [PubMed] [Google Scholar]
  • 18.Alnæs D, Kaufmann T, Marquand AF, Smith SM, Westlye LT (2020): Patterns of sociocognitive stratification and perinatal risk in the child brain. Proceedings of the National Academy of Sciences. 117:12419–12427. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Sui J, Jiang R, Bustillo J, Calhoun VJBp (2020): Neuroimaging-based individualized prediction of cognition and behavior for mental disorders and health: methods and promises. Biol Psychiat. 88:818–828. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Woo C-W, Chang LJ, Lindquist MA, Wager TDJNn (2017): Building better biomarkers: brain models in translational neuroimaging. Nature neuroscience. 20:365–377. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Rashid B, Calhoun V (2020): Towards a brain-based predictome of mental illness. Human brain mapping. 41:3468–3535. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Casey B, Cannonier T, Conley MI, Cohen AO, Barch DM, Heitzeg MM, et al. (2018): The adolescent brain cognitive development (ABCD) study: imaging acquisition across 21 sites. Developmental Cognitive Neuroscience. 32:43–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Qi T, Schaadt G, Friederici AD (2021): Associated functional network development and language abilities in children. NeuroImage. 242:118452. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Gershon RC, Wagster MV, Hendrie HC, Fox NA, Cook KF, Nowinski CJJN (2013): NIH toolbox for assessment of neurological and behavioral function. Neurology. 80:S2–S6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Achenbach T, Rescorla LJRCfC, Youth,, Families (2001): Manual for the ASEBA school-age forms & profiles: an integrated system of multi-informant assessment Burlington, VT: University of Vermont. Research Center for Children. 1617. [Google Scholar]
  • 26.Poldrack RA, Huckins G, Varoquaux GJJp (2020): Establishment of best practices for evidence for prediction: a review. JAMA Psychiatry. 77:534–540. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Jiang R, Woo C-W, Qi S, Wu J, Sui JJISPM (2022): Interpreting Brain Biomarkers: Challenges and solutions in interpreting machine learning-based predictive neuroimaging. 39:107–118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Jiang R, Scheinost D, Zuo N, Wu J, Qi S, Liang Q, et al. (2022): A Neuroimaging Signature of Cognitive Aging from Whole-Brain Functional Connectivity. Advanced Science. 9:2201621. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Reuben A, Caspi A, Belsky DW, Broadbent J, Harrington H, Sugden K, et al. (2017): Association of childhood blood lead levels with cognitive function and socioeconomic status at age 38 years and with IQ change and socioeconomic mobility between childhood and adulthood. Jama. 317:1244–1251. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Tingley D, Yamamoto T, Hirose K, Keele L, Imai K (2014): Mediation: R package for causal mediation analysis.
  • 31.Shen X, Finn ES, Scheinost D, Rosenberg MD, Chun MM, Papademetris X, et al. (2017): Using connectome-based predictive modeling to predict individual behavior from brain connectivity. nature protocols. 12:506–518. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Baron RM, Kenny DAJJop, psychology s (1986): The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology. 51:1173. [DOI] [PubMed] [Google Scholar]
  • 33.Breiman LJMl (2001): Random forests. 45:5–32. [Google Scholar]
  • 34.Fortin J-P, Parker D, Tunç B, Watanabe T, Elliott MA, Ruparel K, et al. (2017): Harmonization of multi-site diffusion tensor imaging data. Neuroimage. 161:149–170. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Duncan GJ, Magnuson KJWIRCS (2012): Socioeconomic status and cognitive functioning: moving from correlation to causation. Wiley Interdisciplinary Reviews: Cognitive Science. 3:377–386. [DOI] [PubMed] [Google Scholar]
  • 36.Rakesh D, Zalesky A, Whittle SJDcn (2021): Similar but distinct–Effects of different socioeconomic indicators on resting state functional connectivity: Findings from the Adolescent Brain Cognitive Development (ABCD) Study®. Developmental Cognitive Neuroscience. 51:101005. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Korous KM, Causadias JM, Bradley RH, Luthar SS, Levy R (2022): A systematic overview of meta-analyses on socioeconomic status, cognitive ability, and achievement: The need to focus on specific pathways. Psychological reports. 125:55–97. [DOI] [PubMed] [Google Scholar]
  • 38.Judd N, Sauce B, Wiedenhoeft J, Tromp J, Chaarani B, Schliep A, et al. (2020): Cognitive and brain development is independently influenced by socioeconomic status and polygenic scores for educational attainment. Proceedings of the National Academy of Sciences. 117:12411–12418. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Fan CC, Marshall A, Smolker H, Gonzalez MR, Tapert SF, Barch DM, et al. (2021): Adolescent Brain Cognitive Development (ABCD) study Linked External Data (LED): Protocol and practices for geocoding and assignment of environmental data. Developmental cognitive neuroscience. 52:101030. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Miguel PM, Pereira LO, Silveira PP, Meaney MJ (2019): Early environmental influences on the development of children’s brain structure and function. Developmental Medicine & Child Neurology. 61:1127–1133. [DOI] [PubMed] [Google Scholar]
  • 41.Janiri D, Doucet GE, Pompili M, Sani G, Luna B, Brent DA, et al. (2020): Risk and protective factors for childhood suicidality: a US population-based study. The lancet Psychiatry. 7:317–326. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Silva PA, Williams S, McGee RJDM, Neurology C (1987): A longitudinal study of children with developmental language delay at age three: later intelligence, reading and behaviour problems. 29:630–640. [DOI] [PubMed] [Google Scholar]
  • 43.Grinde B, Tambs K (2016): Effect of household size on mental problems in children: results from the Norwegian Mother and Child Cohort study. BMC psychology. 4:1–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Zalsman G, Hawton K, Wasserman D, van Heeringen K, Arensman E, Sarchiapone M, et al. (2016): Suicide prevention strategies revisited: 10-year systematic review. The lancet Psychiatry. 3:646–659. [DOI] [PubMed] [Google Scholar]
  • 45.Chen J, Tam A, Kebets V, Orban C, Ooi LQR, Asplund CL, et al. (2022): Shared and unique brain network features predict cognitive, personality, and mental health scores in the ABCD study. Nature communications. 13:1–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Rosenberg MD, Finn ES, Scheinost D, Papademetris X, Shen X, Constable RT, et al. (2016): A neuromarker of sustained attention from whole-brain functional connectivity. Nature neuroscience. 19:165–171. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Jiang R, Zuo N, Ford JM, Qi S, Zhi D, Zhuo C, et al. (2020): Task-induced brain connectivity promotes the detection of individual differences in brain-behavior relationships. NeuroImage. 207:116370. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Doucet GE, Janiri D, Howard R, O’Brien M, Andrews-Hanna JR, Frangou SJEP (2020): Transdiagnostic and disease-specific abnormalities in the default-mode network hubs in psychiatric disorders: A meta-analysis of resting-state functional imaging studies. 63:e57. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Rosenberg MD, Finn ES, Scheinost D, Papademetris X, Shen X, Constable RT, et al. (2016): A neuromarker of sustained attention from whole-brain functional connectivity. 19:165–171. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Seidman LJ, Valera EM, Makris NJBp (2005): Structural brain imaging of attention-deficit/hyperactivity disorder. 57:1263–1272. [DOI] [PubMed] [Google Scholar]
  • 51.Bird CM, Burgess N (2008): The hippocampus and memory: insights from spatial processing. Nature Reviews Neuroscience. 9:182–194. [DOI] [PubMed] [Google Scholar]
  • 52.Guillery R, Sherman SM (2002): Thalamic relay functions and their role in corticocortical communication: generalizations from the visual system. Neuron. 33:163–175. [DOI] [PubMed] [Google Scholar]
  • 53.Hwang K, Shine JM, Bruss J, Tranel D, Boes A (2021): Neuropsychological evidence of multi-domain network hubs in the human thalamus. Elife. 10:e69480. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Michelini G, Barch DM, Tian Y, Watson D, Klein DN, Kotov R (2019): Delineating and validating higher-order dimensions of psychopathology in the Adolescent Brain Cognitive Development (ABCD) study. Translational psychiatry. 9:1–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Dong H-M, Margulies DS, Zuo X-N, Holmes AJJPotNAoS (2021): Shifting gradients of macroscale cortical organization mark the transition from childhood to adolescence. 118:e2024448118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56.Power JD, Fair DA, Schlaggar BL, Petersen SEJN (2010): The development of human functional brain networks. 67:735–748. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57.Edde M, Leroux G, Altena E, Chanraud SJJonr (2021): Functional brain connectivity changes across the human life span: From fetal development to old age. 99:236–262. [DOI] [PubMed] [Google Scholar]
  • 58.Liu T, Wang L, Suo D, Zhang J, Wang K, Wang J, et al. (2022): Resting-state functional MRI of healthy adults: temporal dynamic brain coactivation patterns. 304:624–632. [DOI] [PubMed] [Google Scholar]
  • 59.Passiatore R, Antonucci LA, DeRamus TP, Fazio L, Stolfa G, Sportelli L, et al. (2023): Changes in patterns of age-related network connectivity are associated with risk for schizophrenia. 120:e2221533120. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 60.Agcaoglu O, Wilson TW, Wang Y-P, Stephen JM, Fu Z, Calhoun VDJJonm (2022): Altered resting fMRI spectral power in data-driven brain networks during development: A longitudinal study. 372:109537. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Abrol A, Fu Z, Du Y, Wilson TW, Wang YP, Stephen JM, et al. (2023): Developmental and aging resting functional magnetic resonance imaging brain state adaptations in adolescents and adults: A large N (> 47K) study. 44:2158–2175. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

2
NIHMS1955914-supplement-2.pdf (29.8MB, pdf)
3
NIHMS1955914-supplement-3.xlsx (136.6KB, xlsx)

Data Availability Statement

Neuroimaging and behavioral data from the ABCD dataset are obtained from https://nda.nih.gov/abcd with the approval of the ABCD consortium. The data from UK Biobank used in this study can be publicly accessible via their standard data access procedure at https://www.ukbiobank.ac.uk/. Matlab and R scripts written to perform most of the analyses are available from the authors upon request.

RESOURCES