Poster No:
2036
Submission Type:
Abstract Submission
Authors:
Johannes Mohn1, Martin Bauer1, Fiona O'Donovan1, Katharina Pittner1, Nora Byington2, Gracie Grimsrud2, Jerod Rasmussen3, Damien Fair4, Elisabeth Binder5, Sybille Winter1, Sonja Entringer1, Oscar Miranda Dominguez2, Claudia Buss1, Christine Heim1
Institutions:
1Charité – Universitätsmedizin Berlin, Berlin, Germany, 2University of Minnesota, Minneapolis, MN, 3University of California, Irvine, Irvine, CA, 4Masonic Institute for the Developing Brain, University of Minnesota Medical School, Minneapolis, MN, 5Max-Planck-Institute of Psychiatry, Munich, Bavaria
First Author:
Johannes Mohn
Charité – Universitätsmedizin Berlin
Berlin, Germany
Co-Author(s):
Martin Bauer
Charité – Universitätsmedizin Berlin
Berlin, Germany
Damien Fair
Masonic Institute for the Developing Brain, University of Minnesota Medical School
Minneapolis, MN
Claudia Buss
Charité – Universitätsmedizin Berlin
Berlin, Germany
Introduction:
Cognitive abilities rely on distributed brain networks, but it remains unclear which network connections support cognitive development at different ages and predict individual differences in children. Recent analyses of large-scale data show that the reliable and generalizable estimation of brain-cognition associations requires thousands of individuals (Marek et al., 2022). These sample sizes are hard to obtain in developmental neuroimaging cohorts. Inspired by approaches in genomics, the new framework of polyneuro risk scores (PNRS) leverages large-scale population-based studies to derive brain feature scores for application in smaller samples. This approach can derive such associations out of different brain feature types, including resting-state functional connectivity (RSFC). Interestingly, a recent study using RSFC data in the Adolescent Brain Cognitive Development (ABCD) study established that a weighted sum of network edges explained roughly 21% variance in general cognitive ability and 5% variance in memory (Byington et al., 2023). We tested the generalizability and sensitivity of the PNRS method to the prediction of cognitive abilities in a pediatric sample enriched with cases of childhood maltreatment.
Methods:
RSFC data of 111 children aged 6-13 from the Berlin Kids2Health study were successfully preprocessed in harmonization with ABCD. All children completed cognitive testing with a battery of tasks tapping into executive function, memory, verbal comprehension, and IQ. 38 children had experienced maltreatment as assessed through extensive psychiatric assessment. We calculated PNRS for general cognitive ability and for memory based on association strengths (beta weights) obtained from ABCD (discovery sample, N = 6507, ages 9-10). Correlation analyses tested the associations of the scores with cognitive outcomes and dimensions of maltreatment.
Results:
PNRS for general cognitive ability explained 17% and 16% of variance in verbal and nonverbal IQ scores, respectively, but were not predictive of motor ability or conflict monitoring. This is comparable to prediction levels of these scores in ABCD. PNRS for memory were also significantly but less strongly associated with IQ scores and predicted performance on a working memory task (r = 0.31). PNRS for memory were negatively associated with childhood trauma severity (r = -0.37, p = 0.02), but not chronicity (r = -0.15, p = 0.13) and PNRS for cognitive ability showed similar trends, but the results did not reach significance in this small sample (severity, r = -0.31, p = 0.05; chronicity, r = -0.12, p = 0.21).
Conclusions:
We found that PNRS derived from discovery in a population-based study captured brain-wide feature correspondence to cognitive abilities in an independent smaller target sample that differed in age, cultural background, and cognitive testing protocol. These findings support the generalizability of the PNRS method to the robust prediction of cognitive abilities in children for the use in smaller samples. PNRS for memory further captured trauma-associated variation in the multivariate network profile supporting memory functions. These findings lay a basis for further studies aiming to understand individual differences in functional brain organization that can arise in the context of developmental risk factors.
Disorders of the Nervous System:
Neurodevelopmental/ Early Life (eg. ADHD, autism) 2
Modeling and Analysis Methods:
Task-Independent and Resting-State Analysis 1
Keywords:
Cognition
Development
Memory
Systems
Trauma
1|2Indicates the priority used for review
Provide references using author date format
Byington, N., et al. (2023). 'Polyneuro risk scores capture widely distributed connectivity patterns of cognition.' Developmental Cognitive Neuroscience, 60, 101231.
Marek, S. et al. (2022). 'Reproducible brain-wide association studies require thousands of individuals.' Nature, 603(7902), 654–660.