Development and validation of multi-modal transformer models for preeclampsia using clinical, metabolomic, and laboratory data

Preeclampsia (PE) constitutes a major contributor to maternal–fetal morbidity and mortality, yet prospective identification remains challenging due to limited predictive accuracy of conventional screening modalities. To address this gap, we developed multi-modal transformer models for PE subtypes. This retrospective multicenter cohort study enrolled 491 PE cases and 1,162 normotensive controls across six hospitals in China. Plasma metabolomics, placental growth factor (PlGF), and 40 routine early-gestation laboratory parameters were measured. Subtype-specific models for early-onset PE (EPE) and late-onset PE (LPE) were built using multi-modal transformer models. Distinctive metabolic signatures were identified, comprising 5 and 7 differentially expressed metabolites for EPE and LPE, respectively. On external validation, the multi-modal fusion model achieved an AUC of 0.915 (95% CI 0.862–0.968) for EPE. At a 90% specificity threshold, it yielded sensitivity of 72.2% and a negative predictive value of 97.2%. The clinical factors, metabolomics model attained an AUC of 0.832 (95% CI 0.75–0.914), yielding 57.7% sensitivity with 90% specificity. The multi-modal transformer model showed robust external performance for EPE, supporting its potential to complement current screening. In contrast, LPE performance was modest and requires further improvement. Given the retrospective design, prospective validation is essential before clinical implementation.

Authors

Institutions

Publication Details

Journal
European journal of medical research
Published
2026-11-11
DOI
https://doi.org/10.1186/s40001-026-05467-6
Primary Topic
Pregnancy and preeclampsia studies
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Development and validation of multi-modal transformer models for preeclampsia using clinical, metabolomic, and laboratory data

冯穗华, Shuyuan Li, Zhixu Qiu, Ruyun Gao et al.
European journal of medical research
Pregnancy and preeclampsia studies
article

Development and validation of multi-modal transformer models for preeclampsia using clinical, metabolomic, and laboratory data

冯穗华, Shuyuan Li, Zhixu Qiu, Ruyun Gao, 张苏芬, Gefei Xiao, Guixue Hou, Zhiguang Zhao, Xiao Zhang, Liang Lin, Qiang Zhao, Fengxiang Wei, Jie Qin, Wenqiu Xu, Xin Wang, Songchang Chen, Jianguo Zhang, Lanlan Zhang, Lijian Zhao, Rui Zhang, Shang Huang, Shaohan Cheng, Xu Han, Haiquan Hou, Jia Li, Wudong He
article en

Abstract

Preeclampsia (PE) constitutes a major contributor to maternal–fetal morbidity and mortality, yet prospective identification remains challenging due to limited predictive accuracy of conventional screening modalities. To address this gap, we developed multi-modal transformer models for PE subtypes. This retrospective multicenter cohort study enrolled 491 PE cases and 1,162 normotensive controls across six hospitals in China. Plasma metabolomics, placental growth factor (PlGF), and 40 routine early-gestation laboratory parameters were measured. Subtype-specific models for early-onset PE (EPE) and late-onset PE (LPE) were built using multi-modal transformer models. Distinctive metabolic signatures were identified, comprising 5 and 7 differentially expressed metabolites for EPE and LPE, respectively. On external validation, the multi-modal fusion model achieved an AUC of 0.915 (95% CI 0.862–0.968) for EPE. At a 90% specificity threshold, it yielded sensitivity of 72.2% and a negative predictive value of 97.2%. The clinical factors, metabolomics model attained an AUC of 0.832 (95% CI 0.75–0.914), yielding 57.7% sensitivity with 90% specificity. The multi-modal transformer model showed robust external performance for EPE, supporting its potential to complement current screening. In contrast, LPE performance was modest and requires further improvement. Given the retrospective design, prospective validation is essential before clinical implementation.

European journal of medical research
BGI Group (China) (CN), Hebei Medical University (CN), Guangdong Medical College (CN), Shanghai Jiao Tong University (CN), Harbin Institute of Technology (CN), Fudan University (CN), Shenzhen Maternity and Child Healthcare Hospital (CN), International Peace Maternity & Child Health Hospital (CN), Suzhou Municipal Hospital (CN), Shantou University Medical College (CN), Zhuhai Hospital of Integrated Traditional Chinese and Western Medicine (CN), Obstetrics and Gynecology Hospital of Fudan University (CN), Jiangmen Central Hospital (CN), University of Chinese Academy of Sciences (CN), BGI Genomics (CN), Nanjing Medical University (CN)
Pregnancy and preeclampsia studies
DOIOpen Access PDFOpenAlex
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon