SYNERGY-VAE: An explainable generative deep learning framework for discovering depression subgroups from multimodal population health data
SYNERGY-VAE: Ein erklärbares generatives Deep-Learning-Framework zur Entdeckung von Depressions-Untergruppen aus multimodalen Bevölkerungs-Gesundheitsdaten
PLOS Digital Health · 24.09.2026 · 5 (9) · e0001719
DOI: 10.1371/journal.pdig.0001719 ↗Auf einen Blick
SYNERGY-VAE identifizierte in NHANES-Daten 3 latente Cluster mit unterschiedlichen Depressionsprävalenzen. Das beste Cluster-Modell erreichte mit XGBoost eine AUC von 0,839.
Fragestellung
Lässt sich mit multimodalen NHANES-Daten eine transparente Vorhersage von Depressionsrisiko und latente Untergruppen identifizieren?
Methode
Variational Autoencoder mit 5 NHANES-Modi (demografisch, Ernährung, Untersuchung, Labor, Fragebogen); Clustering im Latenzraum; 70/30 Train-Test-Split; Modelle mit Top-30-PFI-Features.
Ergebnisse
3 latente Cluster mit Depressionsprävalenzen von 6,8 % bis 10,9 %. Cluster-Modelle waren besser als gepoolte Modelle; bestes Ergebnis in Cluster 2 mit XGBoost: AUC 0,839 (95 %-KI 0,804–0,874).
Schlussfolgerung der Autor:innen
SYNERGY-VAE zeigt nach Angaben der Autor:innen das Potenzial generativer, erklärbarer Deep-Learning-Verfahren zur Identifikation latenter Gesundheitsphänotypen für Precision Mental Health.
Kernaussagen
- 3 Cluster mit unterschiedlichen Depressionsprävalenzen
- Bestes Modell: XGBoost in Cluster 2, AUC 0,839
- Feature-Wichtigkeit variierte zwischen den Clustern
Deutsche Zusammenfassung: KI-gestützt aus dem Original-Abstract erstellt; alle Zahlen wurden automatisch mit dem Original abgeglichen. Maßgeblich ist der Original-Abstract.
Kurz gefasst
Fachartikel in der Zeitschrift PLOS Digital Health, online erschienen am 24.09.2026. Verfasst von Divya Sharma und 5 weiteren Autor:innen. Beteiligte Einrichtung: Department of Mathematics and Statistics, York University, Toronto, Ontario, Canada. Die Arbeit nennt 39 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Themen: Depression
Abstract (Originalsprache)
Depression represents a complex, multifactorial disorder influenced by the interplay of biological, behavioral, and social determinants. While extensive health surveys, including the National Health and Nutrition Examination Survey (NHANES), gather comprehensive multidomain data, existing studies tend to analyze these domains independently, overlooking the potential to identify integrated, data-driven patterns of depression risk. We have developed SYNERGY-VAE, an interpretable and multimodal deep learning framework that is designed to identify latent health subgroups and facilitate the transparent prediction of depression risk, utilizing NHANES-derived multimodal data from a large analytic cohort that spans five distinct domains from 2005 to 2018. SYNERGY-VAE utilizes a variational autoencoder to learn a shared latent representation derived from five modalities of NHANES, specifically demographic, dietary, examination, laboratory, and questionnaire data. Clustering within this latent space revealed subpopulations exhibiting distinct health signatures. To enhance interpretability, we triangulated insights utilizing encoder weights, standardized mean differences, and permutation feature importance (PFI). Machine learning classifiers were trained within each cluster employing the top 30 PFI-ranked features, with performance evaluated through the receiver operating characteristic area under the curve (AUC) in a 70/30 train-test split. Three latent clusters emerged, each demonstrating markedly different observed prevalence rates of depression within the analytic sample, ranging from 6.8% to 10.9%. Cluster-specific models consistently surpassed pooled models in performance, with the highest predictive accuracy identified in Cluster 2 utilizing XGBoost (AUC = 0.839, 95% CI:0.804–0.874). The importance of features varied across clusters, indicating unique depression risk profiles specific to each subgroup. SYNERGY-VAE demonstrates the power of generative, explainable deep learning for the identification of latent health phenotypes, thereby furthering the objectives of precision mental health. By integrating high-dimensional, multimodal data and facilitating transparent subgroup discovery, our model may inform future stratified and context-aware screening research and contribute to precision psychiatry-oriented research in large health survey datasets.
Autor:innen und Institutionen
- Department of Mathematics and Statistics, York University, Toronto, Ontario, Canada.
- Biostatistics Department, Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada.
- Biostatistics Department, Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada.
- AI for Mental Health Program, St. Michael's Hospital, Unity Health Toronto, Toronto, Ontario, Canada.
- Department of Electrical and Computer Engineering, Toronto Metropolitan University, Toronto, Ontario, Canada.
- Department of Psychiatry, University of Toronto, Toronto, Ontario, Canada.
ORCID-Profile: Divya Sharma
Zitieren
APA
Sharma, D., Rueda, A., Lin, Q., Meshkat, S., Perivolaris, A., & Bhat, V. (2026). SYNERGY-VAE: An explainable generative deep learning framework for discovering depression subgroups from multimodal population health data. PLOS Digital Health, 5(9), e0001719. https://doi.org/10.1371/journal.pdig.0001719
Vancouver
Sharma D, Rueda A, Lin Q, Meshkat S, Perivolaris A, Bhat V. SYNERGY-VAE: An explainable generative deep learning framework for discovering depression subgroups from multimodal population health data. PLOS Digital Health. 2026;5(9):e0001719. doi:10.1371/journal.pdig.0001719
BibTeX
@article{Sharma2026,
author = {Sharma, Divya and Rueda, Alice and Lin, Qiaowei and Meshkat, Shakila and Perivolaris, Argyrios and Bhat, Venkat},
title = {SYNERGY-VAE: An explainable generative deep learning framework for discovering depression subgroups from multimodal population health data},
journal = {PLOS Digital Health},
year = {2026},
volume = {5},
number = {9},
pages = {e0001719},
doi = {10.1371/journal.pdig.0001719}
}
EUVIMED