Polygenic Risk and White Matter Improve Deep Learning Prediction of Youth Depression and Suicidality
Poster No:
1161
Submission Type:
Late-Breaking Abstract Submission
Authors:
Jung Youn Min1, Heehwan Wang2, Yoonjung Joo3, Bo-Gyeom Kim4, Gakyung Kim5, Eunji Lee4, Jiook Cha6
Institutions:
1Department of Brain and Cognitive Sciences, Seoul National University, Seoul, Republic of Korea, Seoul, Seoul, 2Graduate School of Artificial Intelligence, Seoul National University, Seoul, Seoul, 3Department of Digital Health, Samsung Advanced Institute for Health Sciences & Technology (SAIHST, Seoul, Seoul, 4Department of Psychology, Seoul National University, Seoul, Seoul, 5Department of Brain and Cognitive Sciences, Seoul National University, Seoul, Seoul, 6Seoul National University, Seoul, Korea, Republic of
First Author:
Jung Youn Min
Department of Brain and Cognitive Sciences, Seoul National University, Seoul, Republic of Korea
Seoul, Seoul
Department of Brain and Cognitive Sciences, Seoul National University, Seoul, Republic of Korea
Seoul, Seoul
Co-Author(s):
Yoonjung Joo
Department of Digital Health, Samsung Advanced Institute for Health Sciences & Technology (SAIHST
Seoul, Seoul
Department of Digital Health, Samsung Advanced Institute for Health Sciences & Technology (SAIHST
Seoul, Seoul
Late Breaking Reviewer(s):
Eduardo Garza-Villarreal, M.D., Ph.D.
Universidad Nacional Autónoma de México campus Juriquilla
Juriquilla, Querétaro
Universidad Nacional Autónoma de México campus Juriquilla
Juriquilla, Querétaro
Introduction:
Youth depression is a major global health concern, marked by high recurrence, chronicity, and reduced treatment efficacy (Fleisher & Katz, 2001; Organization, 2017; Parker et al., 2003). Polygenic risk and heritability contribute to depression severity and persistence (Franić et al., 2010; Harder et al., 2022). Genetic vulnerabilities particularly affect white matter (Elliott et al., 2018; Kochunov et al., 2015), a key substrate for long-range connectivity and depression risk (Shen et al., 2021; Vulser et al., 2018). Given these gene-brain interactions, we hypothesized that integrating genetic and white matter data would improve risk stratification. However, most studies examine genetic and neuroimaging factors separately, potentially overlooking their interactive effects on depression risk. Additionally, existing studies have largely focused on Caucasian populations, limiting cross-ethnic applicability.
We propose a deep learning framework integrating genetic risk with white matter imaging (track-weighted fractional anisotropy, TW-FA) to predict depression and related suicidal risk in children. Using a multi-ethnic youth cohort, we conduct (1) cross-sectional Major Depressive Disorder (MDD) classification, (2) prospective prediction of depressive symptoms, and (3) external validation in an independent Korean adolescent dataset. By explicitly modeling gene-brain interactions, our findings refine neurobiological models of depression and provide a framework for early screening in youth mental health.
We propose a deep learning framework integrating genetic risk with white matter imaging (track-weighted fractional anisotropy, TW-FA) to predict depression and related suicidal risk in children. Using a multi-ethnic youth cohort, we conduct (1) cross-sectional Major Depressive Disorder (MDD) classification, (2) prospective prediction of depressive symptoms, and (3) external validation in an independent Korean adolescent dataset. By explicitly modeling gene-brain interactions, our findings refine neurobiological models of depression and provide a framework for early screening in youth mental health.
Methods:
We analyzed data from the Adolescent Brain and Cognitive Development (ABCD) study and a Korean cohort, including 5,248 ABCD and 108 Korean participants after quality control. Polygenic risk scores were derived from previously validated calculations (Joo et al., 2024). TW-FA maps were generated from diffusion MRI and preprocessed using MRtrix3. The ABCD cohort was split into three groups for PGS-based pretraining, using a 3D-CNN to model white matter–PGS associations. The pretrained model was fine-tuned using 266 propensity-matched participants in held-out sets through 10-fold cross-validation. Zero-shot 2-year MDD prediction evaluated generalization to unseen cases (N=118), ensuring robustness. Depression-related white matter regions were identified via Explainable AI, including Integrated Gradients and SmoothGrad.
Results:
The PGS-pretrained 3D-CNN using TW-FA significantly outperformed unimodal baselines in both cross-sectional and 2-year follow-up MDD prediction. For cross-sectional classification, the pretrained model achieved an AUROC of 0.62±0.121, improving by 24% over PGS-only models, 5.8% over TW-FA-only models. In 2-year follow-up MDD prediction, the pretrained model (AUROC = 0.61±0.011) significantly outperformed PGS-only models by 8.5%, TW-FA-only models by 3.6% (p < 0.05). For MDD with suicidal attempts, the PGS-pretrained model (AUROC 0.665±0.071) outperformed non-pretrained models. Explainable AI identified low FA in key tracts, including the superior longitudinal fasciculus, cingulum, corticospinal tract, and corpus callosum, as major MDD risk features. Additionally, the pretrained model demonstrated strong generalization in an independent Korean adolescent sample, achieving an AUROC of 0.673±0.154, surpassing the from-scratch model by 26.5%.
Conclusions:
Our study shows that PGS-based pretraining enhances neuroimaging-driven classification and prognosis of youth depression and suicidality, surpassing unimodal models. By linking genetic predisposition to white matter features, we offer a scalable framework for early risk identification. The superior longitudinal fasciculus, corpus callosum, and cingulum, were identified as core pathways of MDD pathophysiology. The model's strong performance in an independent Korean cohort highlights its cross-population applicability. Our findings highlight the potential of integrating genetics and neuroimaging to enhance early depression detection, supporting its application in personalized psychiatry.
Disorders of the Nervous System:
Psychiatric (eg. Depression, Anxiety, Schizophrenia) 2
Genetics:
Genetics Other
Modeling and Analysis Methods:
Classification and Predictive Modeling 1
Diffusion MRI Modeling and Analysis
Multivariate Approaches
Keywords:
Affective Disorders
Computational Neuroscience
Data analysis
Machine Learning
PEDIATRIC
Psychiatric Disorders
White Matter
WHITE MATTER IMAGING - DTI, HARDI, DSI, ETC
1|2Indicates the priority used for review
·(A,B) Performance comparison in MDD classification (C,D) Performance comparison in 2y-follow-upl MDD prediction (E,F) AUROC comparison in Korean cohort (G) Voxel wise t-test of XAI map (p<0.001,FDR)
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Provide references using APA citation style.
Fleisher, W. P., & Katz, L. Y. (2001). Early onset major depressive disorder. Paediatrics & Child Health, 6(7), 444-448.
Franić, S., Middeldorp, C. M., Dolan, C. V., Ligthart, L., & Boomsma, D. I. (2010). Childhood and adolescent anxiety and depression: beyond heritability. Journal of the American Academy of Child & Adolescent Psychiatry, 49(8), 820-829.
Harder, A., Nguyen, T.-D., Pasman, J. A., Mosing, M. A., Hägg, S., & Lu, Y. (2022). Genetics of age-at-onset in major depression. Translational Psychiatry, 12(1), 124.
Joo, Y. Y., Lee, E., Kim, B.-G., Kim, G., Seo, J., & Cha, J. (2024). Polygenic architecture of brain structure and function, behaviors, and psychopathologies in children. bioRxiv, 2024.2005.2022.595444. https://doi.org/10.1101/2024.05.22.595444
Kochunov, P., Jahanshad, N., Marcus, D., Winkler, A., Sprooten, E., Nichols, T. E., Wright, S. N., Hong, L. E., Patel, B., & Behrens, T. (2015). Heritability of fractional anisotropy in human white matter: a comparison of Human Connectome Project and ENIGMA-DTI data. Neuroimage, 111, 300-311.
Organization, W. H. (2017). Depression and other common mental disorders: global health estimates.
Parker, G., Roy, K., Hadzi-Pavlovic, D., Mitchell, P., & Wilhelm, K. (2003). Distinguishing early and late onset non-melancholic unipolar depression. Journal of Affective Disorders, 74(2), 131-138.
Shen, X., MacSweeney, N., Chan, S. W., Barbu, M. C., Adams, M. J., Lawrie, S. M., Romaniuk, L., McIntosh, A. M., & Whalley, H. C. (2021). Brain structural associations with depression in a large early adolescent sample (the ABCD study®). EClinicalMedicine, 42.
Vulser, H., Paillère Martinot, M.-L., Artiges, E., Miranda, R., Penttilä, J., Grimmer, Y., van Noort, B. M., Stringaris, A., Struve, M., & Fadai, T. (2018). Early variations in white matter microstructure and depression outcome in adolescents with subthreshold depression. American Journal of Psychiatry, 175(12), 1255-1264.