Multi-omics analysis of obesity and glucose dysregulation including prediabetes and Type 2 diabetes in the KORA cohort

Niu, Jiefei (2026): Multi-omics analysis of obesity and glucose dysregulation including prediabetes and Type 2 diabetes in the KORA cohort. Dissertation, LMU München: Faculty of Medicine

Abstract

With the rising global burden of obesity and T2D, there is an urgent and growing need for more refined and effective strategies to support early disease detection, personalized risk assessment, and a more detailed mechanistic understanding of these highly prevalent and interrelated metabolic disorders. Both obesity and T2D contribute significantly to global morbidity and mortality, being linked to a wide range of complications, including cardiovascular disease, liver dysfunction, hepatic complications, and inflammation complications. Despite substantial epidemiological evidence linking these conditions, the biological mechanisms that underlie their development and progression remain only partially understood, particularly at the molecular level. To address this knowledge gap, this cumulative doctoral thesis, based on two peer-reviewed publications, adopts a multi-omics approach that integrates high-throughput proteomic, glycomic, and genomic profiling to investigate circulating molecular biomarkers associated with metabolic dysfunction. The research is based on extensive, well-characterized data from the KORA study, one of Germany’s most comprehensive population-based cohorts with long-term follow-up and detailed clinical, biochemical, and genetic information. The first study focused on identifying proteomic signatures linked to obesity and aimed to uncover protein-level alterations associated with excess adiposity and to explore their potential as diagnostic and mechanistic markers. The second study centered on glycomics, assessing longitudinal changes in the plasma N-glycome using paired data in the KORA F4 and FF4 waves and aimed to investigate whether specific N-glycans are classified as glycemic deterioration over time, including the transition from normoglycemia to prediabetes or T2D. Paper I investigated the plasma proteome in 2,045 participants (564 obese and 1481 non-obese) from the KORA FF4 cohort (2013–2014) using untargeted LC–MS/MS in DDA mode. After rigorous quality control and normalization, 809 proteins were retained for downstream analysis. A combination of multivariable logistic regression in the basic model (age, sex) and Priority-Lasso feature selection identified 16 proteins significantly associated with obesity, defined as BMI ≥30 kg/m². These proteins were enriched in lipid metabolism, immune modulation, and inflammatory pathways. Four types of machine learning models were built on the selected sixteen proteins and demonstrated strong classification accuracy (AUCs ranged from 0.791 to 0.820), outperforming models regarding classical clinical risk factors alone (AUCs ). For model explainability in RF, the top proteins were AFM, CRP, and LGALS3BP. Correlation analyses showed that all sixteen selected proteins were significantly associated with obesity complication risk factors such as BMI, waist circumference, blood pressure, etc. To evaluate causal relationships between these significant proteins and BMI, two-sample Mendelian randomization (MR) analyses were performed using GWAS summary statistics for BMI and external pQTL datasets. These analyses provided evidence that BMI may causally influence circulating levels of AFM, CRP, and CFH, suggesting that these proteins may act as downstream effectors of obesityrelated metabolic disturbances. Paper II focused on longitudinal changes in the plasma N-glycome and their association with glycemic progression. A total of 500 KORA participants with paired samples from the F4 (2006–2008) and FF4 (2013–2014) surveys were included, among whom 250 progressed from normoglycemia to prediabetes or T2D. After excluding missing variables, 473 participants remained for analysis (normoglycemia = 242, prediabetes/T2D = 231). Plasma N-glycans were measured using HILIC-UPLC-FLR. 39 glycan peaks and 16 derived traits representing glycosylation features were included in the analysis. Nineteen glycans were significantly associated with glycemic deterioration in the basic model, such as GP32 and GP22. These glycans were also strongly correlated with core T2D-related traits. Among them, GP32 emerged as the most robust marker, showing consistent associations across all stages of progression. Classification models that integrated glycan profiles with clinical variables achieved an AUC of 0.895, outperforming models based on either component alone. To further explore mechanistic insights, genome-wide genotyping from KORA S4 was incorporated using KORA FF4 (n = 442) for 13 glycans and 6 derived traits associated with prediabetes/T2D. Glycan-QTL mapping revealed strong genetic regulation of key glycan traits, identifying associations between GP32 and the sialylation-related gene ST3GAL4, as well as between GP34 and the fucosylation enzyme gene FUT8. These genes are known to regulate terminal glycan structures involved in immune modulation and inflammation, both of which are central to early T2D pathogenesis. MR analysis suggested causal roles for glycans such as GP19 in T2D, BMI, and HbA1c, while GP19, a high-mannose glycan likely originating from ApoB-100, showed inverse associations with T2D, BMI, and HbA1c, potentially reflecting protective anti-inflammatory effects in early metabolic imbalance. Collectively, this thesis provides new molecular insights into the biological processes that connect obesity and early glycemic deterioration. By leveraging high-throughput proteomic and glycomic technologies, combined with genetic analyses and causal inference, these two studies demonstrate how circulating biomarkers can reflect disease-relevant biological states and may be harnessed to improve individual risk prediction. Moreover, the incorporation of genome-wide data in glycomics underscores the value of integrating multi-omics layers to uncover regulatory mechanisms underlying complex traits. These findings contribute to the growing field of omics-driven epidemiology and support the use of molecular biomarkers for early detection, risk stratification, and personalized prevention strategies in metabolic disease. While each study addresses a distinct clinical phenotype—obesity in Paper I and prediabetes/T2D progression in Paper II—the shared methodological framework and population-based data provide a coherent perspective on metabolic deterioration. Future directions include experimental validation of candidate markers, replication in diverse populations, and translation into risk assessment tools. Ultimately, this work reinforces the value of omics-informed epidemiology in unraveling the molecular architecture of metabolic diseases and sets the stage for more personalized and mechanism-based prevention strategies.

• Item Type: Theses (Dissertation, LMU Munich)
• Keywords: multiomics, proteomics, obesity, glycomics, type 2 diabetes, prediabetes
• Subjects: 600 Technology, Medicine; 600 Technology, Medicine > 610 Medical sciences and medicine
• Faculties: Faculty of Medicine
• Language: English
• Date of oral examination: 14. April 2026
• 1. Referee: Peters, Annette
• Signature of the printed copy: 0700/UMD 22781