From biomarker discovery to prediction models: clinical translation of multi-omics for early prediction of gestational diabetes mellitus
- 1Clinical Laboratory, Shanghai Putuo District Maternity and Infant Hospital
Abstract
Gestational diabetes mellitus (GDM) affects roughly one in seven pregnancies worldwide, yet diagnosis by oral glucose tolerance testing at 24–28 weeks of gestation leaves only a narrow window for intervention. The demonstrated benefit of treating early-diagnosed GDM has intensified interest in first-trimester prediction. This narrative review synthesises recent advances in early-pregnancy biomarkers for GDM, shifting the emphasis from biomarker discovery to clinical translation. Conventional lipid indicators and composite indices provide moderate predictive power at minimal cost (odds ratios 1.4–1.7; AUCs 0.75–0.82). Metabolomic and lipidomic studies have identified recurrent signatures — phospholipid remodelling, glycerolipid accumulation, fatty acid and acylcarnitine excess, branched-chain amino acids and bile acids — that map onto insulin resistance, lipotoxicity and hepatic steatosis, and omics-enhanced models report derivation-stage AUCs of approximately 0.83–0.95 with apparent incremental value over clinical factors alone, although these estimates derive predominantly from small cohorts that have rarely undergone external validation. Gut-microbiome features, thyroid parameters and environmental exposures further widen the aetiological frame. Translation remains limited, however, by small derivation cohorts, scarce external validation, analytical heterogeneity and the absence of trials testing biomarker-guided early intervention. We propose a pragmatic roadmap towards clinically deployable first-trimester prediction, with particular attention to high-burden Asian populations.
Related
- AI-driven multi-omics biomarkers in precision breast cancer therapeutics: from discovery to clinical translation
- Early prediction of gestational diabetes mellitus using deep learning applied to routinely collected clinical data.
- Patient-Derived Organoids in Functional Precision Oncology: Tumor Modeling, Drug-Response Prediction, and Clinical Translation.
- A machine learning-based predictive model for gestational diabetes mellitus based on first trimester clinical indicators
- HELIOS: A spatial omics-informed multi-organ cell atlas enables AI-based tumour microenvironment profiling and biomarker discovery from routine histopathology
BibTeX
@article{yan2026from,
title = {{From biomarker discovery to prediction models: clinical translation of multi-omics for early prediction of gestational diabetes mellitus}},
author = {Huibo Yan},
journal = {Frontiers in Endocrinology},
year = {2026},
month = oct,
doi = {10.3389/fendo.2026.2007266},
url = {https://www.frontiersin.org/articles/10.3389/fendo.2026.2007266/full},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - From biomarker discovery to prediction models: clinical translation of multi-omics for early prediction of gestational diabetes mellitus
AU - Huibo Yan
PY - 2026
DA - 2026/10/22
JO - Frontiers in Endocrinology
DO - 10.3389/fendo.2026.2007266
UR - https://www.frontiersin.org/articles/10.3389/fendo.2026.2007266/full
AB - Gestational diabetes mellitus (GDM) affects roughly one in seven pregnancies worldwide, yet diagnosis by oral glucose tolerance testing at 24–28 weeks of gestation leaves only a narrow window for intervention. The demonstrated benefit of treating early-diagnosed GDM has intensified interest in first-trimester prediction. This narrative review synthesises recent advances in early-pregnancy biomarkers for GDM, shifting the emphasis from biomarker discovery to clinical translation. Conventional lipid indicators and composite indices provide moderate predictive power at minimal cost (odds ratios 1.4–1.7; AUCs 0.75–0.82). Metabolomic and lipidomic studies have identified recurrent signatures — phospholipid remodelling, glycerolipid accumulation, fatty acid and acylcarnitine excess, branched-chain amino acids and bile acids — that map onto insulin resistance, lipotoxicity and hepatic steatosis, and omics-enhanced models report derivation-stage AUCs of approximately 0.83–0.95 with apparent incremental value over clinical factors alone, although these estimates derive predominantly from small cohorts that have rarely undergone external validation. Gut-microbiome features, thyroid parameters and environmental exposures further widen the aetiological frame. Translation remains limited, however, by small derivation cohorts, scarce external validation, analytical heterogeneity and the absence of trials testing biomarker-guided early intervention. We propose a pragmatic roadmap towards clinically deployable first-trimester prediction, with particular attention to high-burden Asian populations.
SN - 1664-2392
ER -