A 2026 Singapore birth-cohort preprint found a conditional childhood-anxiety signal: the cord-blood DNA methylation component predicted less age-13 anxiety only in boys in the low DMN-limbic trajectory group from ages 4.5 to 10.5 years. The paper is useful for mapping developmental risk biology, but it is not a screening test and it does not show that changing methylation or brain-network development prevents anxiety.
Research Highlights
- Conditional anxiety signal: in 97 children, the DNA methylation-by-sex-by-DMN-limbic trajectory interaction predicted age-13 anxiety symptoms (B = 0.56, SE = 0.24, p = .022).
- Male-only simple slope: the methylation component predicted less anxiety only in boys in the low DMN-limbic trajectory group (B = -0.16, SE = 0.25, p = .021).
- Female pathway unresolved: methylation was not associated with anxiety in girls at low, mean, or high DMN-limbic trajectory levels (all p > .38), even though the study expected the association to be clearer in girls.
- DAN comparison was null: dorsal-attention/limbic network integration did not significantly moderate the methylation-anxiety link (B = 0.56, SE = 0.41, p = .184).
- Evidence strength is exploratory: only 97 of 331 cord-blood dyads entered the final analytic sample, and anxiety was measured at 1 age-13 time point.
DNA methylation is an epigenetic chemical mark that can influence how strongly genes are expressed without changing the DNA sequence itself. Parks et al. focused on 126 cytosine-guanine sites in stress-related genes from umbilical cord blood, then reduced that high-dimensional methylation pattern into a single principal-component score.
Network integration means that regions from different large-scale brain networks are grouped into shared functional communities, suggesting coordinated activity at a system level. In this study, the key network pair was the default mode network (DMN), which supports self-referential thinking and internally focused thought, with the limbic network, which includes emotion-processing regions such as orbitofrontal and anterior temporal areas.
97 Children Had Birth Methylation, Childhood fMRI, and Age-13 Anxiety Data
Parks et al. used the Growing Up in Singapore Towards healthy Outcomes (GUSTO) cohort, a longitudinal birth cohort that followed children from pregnancy into adolescence. Of 331 mother-child dyads with cord-blood samples, 97 children had at least 2 good-quality resting-state functional magnetic resonance imaging scans between ages 4.5 and 10.5 years plus anxiety assessment at age 13.
Resting-state fMRI measures spontaneous brain-activity correlations while a person is not performing a task. The researchers used those scans to estimate each child’s trajectory of DMN-limbic and dorsal-attention/limbic integration across childhood.
The anxiety outcome was the Multidimensional Anxiety Scale for Children, Second Edition (MASC-2), a 50-item self-report measure covering physical symptoms, generalized anxiety, harm avoidance, social anxiety, obsessive-compulsive symptoms, and separation anxiety or phobias. Higher raw total scores indicated more severe anxiety symptoms.
- Sample: 57 girls and 40 boys entered the final analytic model.
- Neuroimaging window: resting-state fMRI scans were collected at ages 4.5, 6, 7.5, and 10.5 years.
- Epigenetic input: 126 stress-related CpG sites were assayed from cord blood and summarized through principal component analysis.
- Covariates: models adjusted for child ethnicity, gestational age, maternal age, maternal education, prenatal maternal depression/anxiety symptoms, and scan motion parameters.
The Main Result Was Sex-Specific and Opposite the Expected Direction
The primary model found a significant 3-way interaction among cord-blood DNA methylation, biological sex, and DMN-limbic integration trajectory predicting age-13 anxiety symptoms. The coefficient was B = 0.56, SE = 0.24, t = 2.33, p = .022.
The simple slopes narrowed the result: the stress-gene methylation component at birth was associated with less anxiety only among boys in the low DMN-limbic trajectory group.
That result was partly contrary to the prediction. Parks et al. expected stronger methylation-anxiety associations in females, given the higher prevalence of anxiety disorders in girls and women and prior evidence of sex differences in emotion-processing networks. Instead, the only significant simple slope appeared in a male subgroup.
DAN-Limbic Integration Did Not Carry the Same Anxiety Signal
The comparison network was the dorsal attention network (DAN), a system involved in externally directed, goal-focused attention. The paper tested whether DAN-limbic integration trajectories moderated the same methylation-anxiety association.
That model did not pass the threshold for statistical significance. The DAN-limbic 3-way interaction was B = 0.56, SE = 0.41, t = 1.13, p = .184. In practical terms, the study did not support the idea that childhood trajectories of attention-emotion integration explained the methylation link to age-13 anxiety.
Interpretation: the signal pointed more toward internally focused self-referential/emotional circuitry than toward externally directed attentional control. That is biologically plausible because the DMN has been linked to worry, rumination, and internally directed threat processing, but the present study cannot prove that DMN-limbic development caused lower or higher anxiety.
Adjacent Anxiety Studies Support the Network Frame but Not a Screening Claim
Pilkay et al. reported that trauma context exerted intergenerational effects on child mental health through DNA methylation, with child anxiety-related behaviors among the outcomes. That paper helps place Parks et al. inside a small but growing literature on early-life methylation and later anxiety risk, including sex-specific patterning.
Network-level anxiety evidence points in the same broad direction. Guo et al. found that adolescents with first-episode, drug-naive generalized anxiety disorder had disruptions in functional and structural networks. Makovac et al. used graph theory and argued that generalized anxiety abnormalities extended beyond the familiar amygdala-prefrontal circuit into broader network coordination.
Sex-specific neural evidence also fits the caution. Padgaonkar et al. found that youth internalizing symptoms were associated with different amygdala connectivity patterns in males and females. That makes it reasonable to test sex-by-network models, but it also warns against treating a single subgroup finding as a universal anxiety biomarker.
Why Higher Methylation Could Look Protective in Boys
The male-only protective slope can sound backwards if methylation is treated as a simple risk mark. It is not that simple. Methylation can increase or decrease transcription depending on genomic location, and the Parks et al. component combined many stress-related loci into a single score.
The paper’s principal component loaded mostly positively on loci mapped to genes involved in neuroendocrine and stress regulation, including NCOR2, CRHBP, HSD11B1, and HTR2A. Higher scores therefore reflected proportionally higher methylation across those loci, not a single gene switch.
Two interpretations are plausible:
- Stress-system calibration: higher methylation across this component may index early biological calibration of stress-response pathways rather than damage.
- Developmental-context dependence: the same methylation pattern may have different meaning depending on how self-referential and limbic networks integrate across childhood.
The second point is the useful part of the paper. Epigenetic risk markers may be weak in isolation because their behavioral meaning depends on brain-development context, sex, and timing of the outcome.
The Evidence Is Better for Hypothesis Generation Than Prediction
Evidence-strength note: this is a preprint analysis of an observational birth cohort. It can support a sex-specific developmental-risk hypothesis, but it cannot support clinical prediction, diagnosis, intervention selection, or causal claims about methylation changing anxiety risk.
The final analytic sample was also selective. Only 97 of the original 331 dyads with cord-blood methylation samples had the required combination of usable fMRI scans and age-13 anxiety assessment. Included and excluded dyads differed significantly in maternal education (X2 = 9.91, p = .042), which does not invalidate the result but does narrow confidence in generalizability.
Several additional caveats matter:
- Single anxiety endpoint: anxiety was measured at age 13, so the study did not model when symptoms emerged or whether they persisted.
- Subgroup dependence: the clearest result was 1 male/low-DMN-limbic subgroup slope, while most other simple slopes were not significant.
- Peripheral tissue: cord-blood methylation is not brain methylation, even when the selected genes are biologically relevant to stress regulation.
- Preprint status: peer review may change model details, language, or interpretation.
Practical Read: Conditional Biology, Not a Childhood Anxiety Test
The practical use of Parks et al. is conceptual. It argues that early anxiety risk is unlikely to be captured by a single methylation score, a single scan, or a simple sex difference. Risk may emerge from the interaction of stress biology at birth, sex-specific development, and how emotion-related brain networks organize across childhood.
The wrong read is that higher methylation is good, boys are protected, or DMN-limbic integration should be pushed in a particular direction. The study does not test intervention, and the protective-looking slope appeared only under a specific developmental condition.
A better read is narrower: childhood anxiety biology may require interaction models. The strongest future designs would test larger cohorts, repeated anxiety measures, independent replication, and models that compare whether methylation adds predictive value beyond family history, early temperament, adversity exposure, and ordinary symptom screening.
Questions About Childhood Anxiety, DNA Methylation, and Brain Networks
Did this study identify a clinical biomarker for childhood anxiety?
No. The result was a conditional association in 97 children, not a validated screening biomarker. It should not be used to classify individual children as high- or low-risk.
What does DMN-limbic integration mean in plain language?
It describes how strongly self-referential brain systems and emotion-processing systems appear to participate in shared functional communities across childhood. It is a network-organization measure, not a direct readout of anxiety.
Was the finding stronger in girls?
No. The paper expected stronger associations in females, but methylation was not significantly associated with anxiety in girls at low, mean, or high DMN-limbic trajectory levels. The only significant simple slope appeared in boys with slower DMN-limbic integration increases.
Why does the preprint status matter?
Preprints are public scientific manuscripts that have not completed peer review. The data may be useful, but model choices, interpretation, and wording can change before journal publication.
References
- Parks KC, Uy J, Buthmann J, Tao X, Tan AP, Gotlib I. Sex Differences in the Interaction of Epigenetic Risk and Trajectories of Default Mode Limbic Network Integration Predicting Childhood Anxiety. Research Square. 2026. https://doi.org/10.21203/rs.3.rs-9286569/v1
- Pilkay S, Riffer A, Carroll A. Trauma context exerts intergenerational effects on child mental health via DNA methylation. Epigenetics. 2024;19(1):2333654. https://doi.org/10.1080/15592294.2024.2333654
- Guo X, Yang F, Fan L, et al. Disruption of functional and structural networks in first-episode, drug-naive adolescents with generalized anxiety disorder. Journal of Affective Disorders. 2021;284:229–237. https://doi.org/10.1016/j.jad.2021.01.088
- Makovac E, Mancini M, Fagioli S, et al. Network abnormalities in generalized anxiety pervade beyond the amygdala-prefrontal cortex circuit: insights from graph theory. Psychiatry Research: Neuroimaging. 2018;281:107–116. https://doi.org/10.1016/j.pscychresns.2018.09.006
- Padgaonkar NT, Lawrence KE, Hernandez LM, Green SA, Galvan A, Dapretto M. Sex Differences in Internalizing Symptoms and Amygdala Functional Connectivity in Neurotypical Youth. Developmental Cognitive Neuroscience. 2020;44:100797. https://doi.org/10.1016/j.dcn.2020.100797
- Makovac E, Meeten F, Watson DR, et al. Alterations in amygdala-prefrontal functional connectivity account for excessive worry and autonomic dysregulation in generalized anxiety disorder. Biological Psychiatry. 2016;80(10):786–795. https://doi.org/10.1016/j.biopsych.2015.10.013