GRAPHICAL ABSTRACT

INTRODUCTION

The global obesity epidemic remains one of the foremost public health challenges, with approximately 890 million adults classified as obese in 2022, according to World Health Organization reports [1,2]. The prevalence of adult obesity has more than doubled since 1990, while childhood obesity has quadrupled during the same period [2]. This condition is recognized as a metabolic disorder requiring treatment because it results from complex interactions among genetic, environmental, and behavioral factors [3]. Beyond weight gain, obesity markedly increases the risk of chronic diseases, including type 2 diabetes associated with insulin resistance [4], cardiovascular diseases linked to metabolic syndrome [5], non-alcoholic fatty liver disease that may progress to cirrhosis [6], and cancers driven by chronic inflammation [7].

Although lifestyle interventions such as dietary modification and exercise remain the cornerstone of obesity management, they are often ineffective in sustaining long-term weight reduction [8]. Physiological adaptations commonly counteract continued weight loss, as the body defends against energy deficits by reducing metabolic rate or increasing appetite, especially under caloric restriction [9]. These biological barriers have encouraged clinicians to integrate pharmacological interventions with conventional approaches. Such adjunct therapies target appetite regulation, nutrient absorption, and energy metabolism, thereby improving prospects for long-term weight control [10].

Glucagon-like peptide-1 receptor agonists (GLP-1RAs) have emerged as a promising therapeutic option, offering both glycemic control and substantial weight loss. Clinical trials have reported mean weight reductions of 12% to 22% [11,12]. GZR18, a next-generation, long-acting GLP-1RA, was developed to enhance patient adherence through once-weekly subcutaneous administration. It exerts its therapeutic effects by promoting glucose-dependent insulin secretion, delaying gastric emptying, and suppressing appetite. However, individual responses to GLP-1RA therapy vary widely, ranging from less than 5% to over 20% weight loss [13–15]. This heterogeneity was evident in the phase II trial of GZR18 among Chinese adults with overweight or obesity, where the mean weight change was −17.78% at week 30, but individual outcomes ranged from −10% to −26% [9,12, 16,17]. Such variability underscores the urgent need for biomarkers that can predict treatment response and guide personalized optimization strategies.

Omics technologies have transformed obesity research by identifying biomarkers of metabolic dysfunction through genomic, transcriptomic, epigenomic, proteomic, and metabolomic approaches [18,19]. However, single-platform analyses often fail to capture pathway interactions and may lack reproducibility. In contrast, integrated multi-omics analyses—combining proteomic, metabolomic, and lipidomic data with transcriptomic flux information—provide deeper insights into the physiologic systems underlying obesity [20–23].

Current evidence on pharmacotherapy for obesity remains limited in its multi-omics characterization of GLP-1RAs, as most studies assess baseline predictors and treatment responses independently. This separation hinders the identification of dual- purpose biomarkers essential for precision dosing. To address this gap, we conducted longitudinal multi-omics profiling before and during the phase II GZR18 trial, analyzing 221 plasma samples collected from 25 participants (n=25) across nine time points (four samples could not be tested due to hemolysis). Our objective was to determine whether circulating molecules could simultaneously predict therapeutic sensitivity at baseline and monitor treatment response dynamics. These findings introduce a novel paradigm for obesity management, integrating predictive molecular profiling with real-time biomarker monitoring and pathway-based personalized intervention.

METHODS

Ethical approval, study design, and patient cohort

This investigative substudy was derived from the parent phase II multicenter randomized controlled trial, prospectively registered with the Chinese Clinical Trial Registry under identifier CTR20231695 (http://www.chinadrugtrials.org.cn/clinicaltrials.searchlist.dhtml). The parent trial aimed to evaluate the efficacy and safety of GZR18, a novel long-acting GLP-1RA, in Chinese adult patients with obesity. All procedures were approved by the Beijing Friendship Hospital Institutional Review Board and Ethics Committee (2023-P2-221-02), and written informed consent was obtained from each participant. Out of the total 338 subjects enrolled across multiple centers in the parent trial, we identified 25 participants (n=25) from Beijing Friendship Hospital who completed the entire 30-week treatment regimen and provided all nine scheduled plasma samples. These participants also provided extended consent for additional biomarker measurements. The cohort represented the full spectrum of dose regimens: 12 mg biweekly (n=4), 18 mg biweekly (n=3), 18 mg weekly (n=1), 24 mg biweekly (n=2), 24 mg weekly (n=4), 48 mg biweekly (n=9), and placebo (n=2), thereby encompassing the complete range of therapeutic responses observed in the main trial.

Eligible participants met the inclusion criteria of the parent study, consisting of adults aged 18–75 years with obesity (body mass index [BMI] ≥28 kg/m²) or overweight (24≤ BMI <28 kg/m²) and at least one comorbidity (e.g., prediabetes, hypertension, or non-alcoholic fatty liver disease). Exclusion criteria included prior use of GLP-1RAs, use of medications influencing body weight, history of bariatric surgery, unmanaged cardiometabolic, hepatic, or renal impairment, malignancy, or psychiatric disorders—conditions that could potentially interfere with study outcomes. Peripheral blood samples were collected at baseline (week 0) and at 4-week intervals through week 33, processed into plasma by centrifugation, aliquoted, and stored at −80°C [24] for subsequent multi-omics analyses, including proteomics, metabolomics, and lipidomics. Therapeutic response was assessed through BMI trajectory modeling.

Multi-omics profiling and data acquisition

Multi-omics profiling was performed on 221 plasma samples collected from 25 participants at nine time points, encompassing proteomics, metabolomics, and lipidomics. All analyses were conducted using high-resolution, high-throughput mass spectrometry (MS) platforms under stringent quality control (QC) protocols.

Proteomics analysis

High-abundance plasma proteins were depleted using the Pierce™ Top 14 Abundant Protein Depletion Kit (Thermo Scientific, Waltham, MA, USA). The depleted samples were then subjected to filter-aided sample preparation-based tryptic digestion. Peptide separation was achieved using the Vanquish Neo nano-LC system, and detection was performed with an Astral DIA mass spectrometer (Thermo Scientific). Spectra were processed using DIA-NN (v1.8) with trypsin specificity and fixed carbamidomethyl (C) modification, as well as variable modifications for oxidation (M) and acetylation (N-term). Proteins were identified at a false discovery rate (FDR) of <1% [25]. The proteomic analysis yielded relative quantification of protein abundance, expressed as log₂-transformed intensity values. This method reliably identifies differentially abundant proteins between groups but does not provide absolute molar concentrations.

Metabolomics analysis

Sample preparation and untargeted metabolite profiling followed the protocol described by Dunn et al. [26] with minor modifications. A Vanquish Flex ultra performance liquid chromatography system coupled with a Q Exactive HF-X mass spectrometer (Thermo Scientific) was operated in both positive and negative ion modes. Chromatographic separation was performed on a Waters ACQUITY BEH C18 column (2.1×100 mm, 1.7 μm) (Waters Corp., Milford, MA, USA) using a gradient of 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B). Metabolites were detected and identified using MS-DIAL (v5.1.230912; https://systemsomicslab.github.io/compms/msdial/main.html) through searches against multiple databases, including MassBank of North America (MoNA), Global Natural Products Social Molecular Networking (GNPS), Human Metabolome Database (HMDB), and the Mass Spectrometry Data Independent Analysis Library (MS-DIAL) reference library, supplemented by an extensive in-house spectral database.

Lipidomics analysis

Lipid extraction was performed following the procedure described by Yin et al. [27], using a butanol–methanol system with additional internal standards. Subsequent analyses employed a SCIEX TripleTOF 5600+ mass spectrometer (Quadrupole-Quadrupole-Linear Ion Trap [QqQ-LIT] configuration) (SCIEX, Marlborough, MA, USA). Chromatographic separation was carried out on a Thermo Accucore C18 column (2.1×150 mm, 2.6 μm) using a mobile phase gradient of 10 mM ammonium formate in water (A) and 10 mM ammonium formate in isopropanol/acetonitrile (B). Peak extraction and downstream analyses were performed using Skyline version 21.1 (MacCoss Lab, University of Washington, Seattle, WA, USA) and R version 4.2.1 (R Foundation for Statistical Computing, Vienna, Austria).

Data preprocessing

Quantitative data from all three omics layers were log₂-transformed and internally normalized. Feature intensity distributions were adjusted using median normalization. Features with more than 30% missing values were excluded, and the remaining missing values were imputed using the K-nearest neighbor algorithm. This preprocessing pipeline produced consistent, high-quality datasets suitable for downstream analyses.

Quality control and integration

QC analyses were conducted independently within each omics platform to assess technical reproducibility and consistency. Instrument performance and processing stability were evaluated using pooled QC samples created by combining aliquots of study samples. Internal standards with known spiked concentrations were included to evaluate normalization efficiency and instrument response. Data quality was assessed through multiple approaches, including density plots (distribution analysis), principal component analysis (PCA) for outlier detection and batch identification, and Pearson correlation coefficient (PCC) heatmaps for sample correlation analysis. These evaluations confirmed that technical bias was minimal. Only features meeting QC thresholds (≥70% quantification across samples) were retained and integrated for subsequent analyses.

Statistical analysis

Clinical outcome modeling

The therapeutic response to GZR18 was quantified using two BMI-based clinical measures over time: (1) the BMI slope (Z), representing treatment sensitivity and calculated using the Theil-Sen estimator across nine time points, and (2) the percent change in BMI from baseline to week 30 (%BMI), representing absolute efficacy. Individual BMI trajectories were plotted and compared across treatment regimens [28,29].

Biomarker analysis

To identify baseline molecular predictors of treatment sensitivity, individual linear regression models were fitted to each omics feature (proteins, metabolites, and lipids). Parallel regression models were used to simultaneously test predictive (baseline) and monitoring (longitudinal) biomarkers. The predictive model was defined as:

Z=a×X1+b×X2+c×X3+d

and for %BMI change (Y):

Y=a×X1+b×X2+c×X3+d

where X₁=baseline abundance, X₂=dose, X₃=dosing frequency.

Associations between X₁ and were chosen at P<0.05, and the direction of the coefficient was used to determine therapeutic outcomes. Both models were evaluated for residual normality, heteroscedasticity, and influence diagnostics. Features that were significant in both models were designated as dual-response biomarkers and subsequently annotated and mapped in molecular networks. The primary focus included (1) direct predictive and monitoring comparison modeling; (2) validation of diagnostic candidates; and (3) multimodal visualization of molecular–clinical relationships.

Pathway enrichment and network analysis

To determine the biological relevance of potential biomarkers, gene set enrichment analysis (GSEA) was performed using Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, integrating both predictive and monitoring biomarker sets. Each feature was ranked based on the product of its regression coefficient and −log₁₀ (P value), and enrichment analysis was conducted using clusterProfiler [30]. Adjusted P values <0.05 were considered statistically significant.

KEGG pathway annotations were used to map candidate biomarkers and integrate them with STRING protein–protein interaction data to construct multi-omics networks. Predictive and monitoring biomarkers generated distinct networks, and glycogenin 1 (GYG1) and its direct interactors were analyzed within STRING-based subnetworks. Visualization of enriched features was performed through molecular interaction networks, and hub features were evaluated based on node-level enrichment metrics.

Definition of response groups for comparative analysis

Participants were stratified into two response groups: good responders (GR, n=17), defined as achieving ≥10% BMI reduction at week 30, and poor responders (PR, n=8), defined as achieving <10% reduction. This binary classification was chosen to ensure distinct response phenotypes for comparative analysis. Baseline clinical and multi-omics characteristics were compared between groups using independent t tests and chi-square tests.

RESULTS

Clinical response and BMI trajectories

Among the 25 obese patients treated with GZR18 (weeks 0–30), higher-dose biweekly regimens (48 mg) produced significantly steeper reductions in BMI (Z<–0.4) compared with placebo (Z=–0.22±0.03; P<0.001). The BMI trajectories displayed nonlinear kinetics of weight loss, with maximal reduction observed after week 12 (Fig. 1A). Analysis of response distribution showed that 68% of patients exhibited rapid responses (Z<–0.4), whereas 32% demonstrated moderate responses (Z=–0.25 to −0.39). The 95% confidence interval closely aligned with the mean response estimate (Fig. 1B). Per-patient analysis revealed a strong correlation between dose and dosing frequency (P< 0.001). These findings confirm a clear dose-dependent relationship while highlighting interindividual variability, suggesting that additional molecular pathways may influence treatment response (Fig. 1).

Multi-omics profiling and quality control

For each of the 25 participants, 221 blood samples were analyzed in triplicate using high-throughput MS, quantifying 4,014 proteins, 670 metabolites, and 787 lipids. Quantitative datasets across the three omics layers were evaluated using probability distributions, PCA, and PCC matrices. The proteomics, metabolomics, and lipidomics QC results are shown in Supplemental Figs. S1–S3. Each figure includes four panels: (Supplemental Figs. S1A, S2A, S3A) quantification statistics confirming minimal molecular heterogeneity among samples, (Supplemental Figs. S1B, S2B, S3B) distribution plots demonstrating normal data distribution, (Supplemental Figs. S1C, S2C, S3C) PCA plots showing the absence of technical artifacts, and (Supplemental Figs. S1D, S2D, S3D) PCC heatmaps indicating high reproducibility. Collectively, these analyses confirm that the datasets were consistent and technically robust for downstream molecular investigations. The precision and reliability of these high-quality datasets enable accurate identification of molecular determinants of obesity and therapeutic response (Supplemental Figs. S1–S3).

Integrated biomarker signatures for prediction and monitoring of obesity drug response

The clinical data demonstrated progressive BMI reduction throughout GZR18 treatment, confirming its efficacy trend. Baseline multi-omics analysis using linear regression revealed significant correlations between molecular features and BMI slope (Z). Volcano plots identified key predictive biomarkers across omics layers. In lipidomics, phosphatidylinositol PI (18:2-18:2) and phosphatidylethanolamine PE (P-18:1/16:0) were positively associated with treatment response, whereas PE (18:1-20:3) and PE (P-16:0/16:0) were negatively associated. Metabolomic analysis identified oxoglutaric acid and 2-oxoarginine as positive predictors, while methylmalonylcarnitine showed a negative correlation. Proteomic profiling revealed glycogenin 1 (GYG1), glucose-6-phosphate isomerase (GPI), and polymeric immunoglobulin receptor (PIGR) as positive correlates of therapeutic success (Fig. 2A).

Longitudinal linear regression analysis identified additional markers for tracking drug efficacy in relation to BMI changes from baseline. Lipidomic results indicated strong associations with phosphatidylcholine PC (18:0-20:3), PC (P-16:0/18:1), and PC (O-18:1/18:1), showing both positive and negative correlations with BMI change. Metabolomic profiling revealed pyrimethamine, PC 36:2, and PC (18:1-18:1) as metabolites linked to BMI reduction. Proteomic analysis identified insulin like growth factor binding protein 2 (IGFBP2), alcohol dehydrogenase 4 (ADH4), and engulfment and cell motility domain-containing protein 2 (ELMOD2) as potential efficacy markers (Fig. 2B). Collectively, these biomarkers were significantly associated with BMI response (P<0.05) at both baseline and longitudinal time points, suggesting their dual roles in prediction and treatment monitoring (Fig. 2).

Clinical and molecular comparison of good versus poor responders

To assess heterogeneity in GZR18 treatment outcomes, participants were stratified into GR (≥10% BMI reduction, n=17) and PR (<10% BMI reduction, n=8). Baseline clinical characteristics were comparable between groups, with no significant differences in age, sex distribution, or baseline BMI (Table 1). However, GR exhibited a significantly greater mean BMI reduction (−12.6%±1.8%) than PR (−3.2%±1.1%, P<0.001), confirming successful stratification.

Comparative multi-omics profiling revealed that several baseline biomarkers were significantly elevated in GR compared with PR. This included GYG1, which showed a 1.35-fold higher relative abundance (P=0.003) in GR. These values represent relative log₂-intensity units from MS; absolute quantification will be required in future validation to establish clinically relevant concentration thresholds. Other elevated biomarkers included PI (18:2-18:2) and oxoglutaric acid. Additionally, IGFBP2 and PC (18:0-20:3) demonstrated greater dynamic changes during treatment, suggesting that these features serve not only as predictive markers but also as effective longitudinal indicators of therapeutic response (Table 2).

Identification and validation of GYG1 as a dual-response biomarker

In this analysis, GYG1 was identified as a dual-response biomarker significantly correlated (P<0.05) with both baseline prediction (BMI slope Z) and longitudinal monitoring (%BMI change) of therapeutic response. Two-dimensional significance mapping (Fig. 3A) revealed several proteins, metabolites, and lipids associated with both models, among which GYG1 showed the strongest overall correlation.

Validation analysis demonstrated a significant negative correlation (P<0.05) between baseline GYG1 levels and treatmentrelated BMI reduction (Fig. 3B). Stratification by median baseline GYG1 expression indicated greater weight loss in participants with higher expression levels across all doses (12–48 mg) and dosing schedules within both 1- and 2-week cycles (P<0.01) (Fig. 3C). This pattern remained stable throughout the treatment period. These findings collectively support GYG1 as a robust biomarker for both predicting and monitoring therapeutic outcomes (Fig. 3).

GSEA enrichment analysis of candidate biomarkers

GSEA identified strong correlations between key metabolic pathways involved in obesity treatment and the candidate biomarkers (Fig. 4). Predictive biomarkers associated with BMI reduction (Z) were significantly enriched in adipocytokine signaling and glucagon signaling pathways. In contrast, monitoring biomarkers that tracked longitudinal BMI changes were enriched in insulin signaling and starch/sucrose metabolism pathways. These biomarkers were also implicated in additional obesity-related signaling cascades, including sphingolipid signaling and ATP-binding cassette (ABC) transporter pathways, which further underscores their roles in metabolic regulation.

Multi-omics network analysis identified GYG1 as the principal intersection connecting predictive and monitoring biomarkers, underscoring its key role in mediating therapeutic efficacy. Each pathway demonstrated statistical significance (FDR-adjusted P<0.05), and normalized enrichment score confirmed their biological relevance. Collectively, these results provide mechanistic insight into how blood-based proteins, metabolites, and lipids influence both therapeutic and metabolic responses to obesity treatment (Fig. 4).

Regulatory network analysis

Multi-omics network modeling identified GYG1 as a central hub bridging predictive and monitoring biosignatures, highlighting its pivotal role in metabolic remodeling during therapy. Mitogenactivated protein kinase (MAPK) signaling and protein biosynthesis pathways were strongly associated with predictive biomarkers (BMI reduction slope) (Fig. 5A). Longitudinally monitored biomarkers formed dynamic modules reflecting temporal treatment-related changes (Fig. 5B). Substantial overlap was observed between the biological mechanisms underlying these networks, suggesting shared pathways in the initiation and maintenance of treatment effects. Significance mapping (−log₁₀ P) revealed co-regulated functional clusters, including upregulated lipid mobilization and downregulated lipid storage processes characteristic of obesity metabolism. The integrated network topology effectively mapped biomarker–pathway interactions involved in the therapeutic response mechanism (Fig. 5).

In-depth functional analysis of GYG1

Network-based functional analyses identified GYG1 as an emergent metabolic regulator positioned at the core of treatment efficacy. Its dual predictive and monitoring roles underscore its pivotal contribution to therapeutic success. GYG1 was significantly correlated (coefficient=2.0) with both starch/sucrose metabolism and insulin signaling pathways, mediated through direct interactions with glycogen synthase 1 (GYS1), amylo-alpha- 1,6-glucosidase, 4-alpha-glucanotransferase (AGL), glycogen phosphorylase, brain form (PYGB), glycogen branching enzyme 1 (GBE1), and uridine diphosphate glucose pyrophosphorylase 2 (UGP2) (Fig. 6A). Metabolomic network analyses showed that essential intermediates—citric acid, fumaric acid, succinic acid, and lactic acid—were functionally organized within the regulatory network. Regulatory proteins such as glycogen- binding protein (GBP) and 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 1 (PFKAB1) displayed synchronized behavior (Fig. 6B). Pathway modeling further demonstrated that GYG1 modulates latent signaling within glucagon and classic metabolic circuits, aligning with both short-term response prediction and long-term treatment monitoring. Collectively, these findings establish GYG1 as a master metabolic regulator and a potential clinical biomarker for obesity pharmacotherapy, owing to its multifunctional interactions and integrative role across molecular systems (Fig. 6).

DISCUSSION

The present study advances understanding of obesity treatment, particularly regarding the long-acting GLP-1 receptor agonist (GZR18), through a comprehensive multi-omics approach. It addresses a critical gap in obesity pharmacotherapy research by identifying biomarkers capable of both predicting and monitoring therapeutic response. Using an advanced longitudinal design that integrates proteomic, metabolomic, and lipidomic datasets, this study reveals not only the variable nature of GZR18 responsiveness but also key molecular features essential for individualized management. This integrated multi-omics perspective provides valuable insights that may guide future biomarkerdriven strategies in obesity treatment.

A major finding of this study is the substantial interindividual variability observed in GZR18 response, consistent with prior research on GLP-1RAs. The BMI slope was significantly steeper in participants receiving the high-dose (48 mg) and twiceweekly GZR18 regimens compared with the placebo group (P<0.001). Despite this overall dose-dependent trend, marked heterogeneity in weight reduction was observed, ranging from minimal (<5%) to substantial (>20%) loss. These findings emphasize that obesity treatment outcomes are not uniform, as patients exhibit differential responses to identical regimens due to multifactorial biological influences. Such interpatient variability is not unique to GZR18 but represents a well-documented challenge in obesity pharmacotherapy [31]. Previous studies on GLP-1RAs have also reported the coexistence of responders and nonresponders, suggesting that individual biological determinants, including genetics, metabolic rate, and gut microbiota, may either facilitate or hinder therapeutic efficacy [32].

The present study employed an advanced multi-omics profiling approach to elucidate the complex metabolic landscape underlying obesity and its treatment response. Analyses across proteomic, metabolomic, and lipidomic layers identified specific biomarkers associated with both baseline therapeutic sensitivity and dynamic response trajectories. For example, in lipidomics, baseline PI (18:2-18:2) and PE (P-18:1/16:0) levels showed positive correlations with treatment response. In metabolomics, oxoglutaric acid and 2-oxoarginine emerged as positive predictors, while in proteomics, GYG1, GPI, and PIGR were identified as potential predictors of GZR18 efficacy. This integrative multi-omics strategy represents a significant methodological advance over traditional single-platform analyses, which often fail to capture the interconnected molecular pathways driving obesity pathophysiology [33].

Furthermore, this study identified biomarkers capable of both predicting and monitoring therapeutic response, including PC (18:0-20:3), PC (P-16:0/18:1), and IGFBP2. These findings are particularly significant because they enable real-time assessment of treatment efficacy through minimally invasive plasma profiling. The discovery of dual-function biomarkers offers a powerful clinical advantage, allowing obesity treatment to be tailored to each patient’s molecular signature [21]. Such precision-guided approaches could help clinicians optimize dosing regimens and identify early signs of treatment resistance.

A central discovery of this study is the identification of GYG1 as a dual-response biomarker significantly associated with both baseline prediction and longitudinal monitoring of therapeutic success. GYG1 emerged as a principal hub within the biomarker interaction network, exhibiting extensive connectivity with metabolic pathways—particularly those involving insulin signaling and starch/sucrose metabolism. Mechanistically, GYG1 encodes glycogenin-1, the primer enzyme for glycogen biosynthesis that catalyzes the autoglucosylation of a tyrosine residue to initiate glycogen synthase activity. This reaction is a critical step in glycogen storage and glucose mobilization during periods of energy demand [34,35]. Dysregulation of glycogen metabolism has been strongly implicated in metabolic disorders such as obesity, insulin resistance, and type 2 diabetes [36,37]. Mechanistic studies suggest that impaired glycogen synthesis and degradation contribute to systemic insulin resistance and defective glucose homeostasis [36]. Although direct human studies on GYG1 are limited, genomic and transcriptomic data indicate associations between GYG1 variants and BMI, as well as altered glycogen-related enzyme expression in obese or insulin-resistant states [38]. These findings support a potential role for GYG1 in regulating glycogen turnover, energy storage efficiency, and insulin sensitivity—central elements of obesity pathogenesis. From a translational standpoint, GYG1 represents a promising candidate biomarker for both predictive and monitoring purposes: it may identify likely responders before therapy initiation and enable longitudinal tracking of treatment efficacy. Such dual-purpose biomarkers are increasingly recognized as essential tools in precision medicine for guiding patient stratification, dosing adjustments, and treatment intensification strategies [39,40]. Validation of GYG1 in larger, independent cohorts incorporating multi-omics datasets and clinical outcomes will be crucial for confirming its clinical applicability. Furthermore, its inclusion in multi-marker predictive panels may enhance diagnostic precision and therapeutic personalization, paving the way for biomarker-based precision management of obesity.

Furthermore, the stratification of patients into GR and PR provided critical validation for our biomarker findings. Despite the absence of significant differences in baseline clinical characteristics such as BMI or age, GR exhibited significantly higher baseline levels of GYG1, PI (18:2-18:2), and oxoglutaric acid. This finding underscores that molecular predisposition to treatment success—rather than overt clinical phenotype—is the primary determinant of the heterogeneous response to GZR18. The distinct trajectories of monitoring biomarkers such as IGFBP2 further reinforce their value in tracking therapeutic efficacy. While our proteomic analysis offered robust relative quantification to identify these differences, future studies employing absolute quantification methods (e.g., immunoassays) will be essential to translate these discoveries into clinically actionable concentration thresholds. Collectively, these results support the feasibility of implementing a precision-medicine approach in obesity pharmacotherapy.

Pathway enrichment analysis played a crucial role in narrowing down the biological processes involved in obesity treatment. Predictive biomarkers were predominantly enriched in adipocytokine and glucagon signaling pathways, whereas monitoring biomarkers were strongly associated with insulin signaling and alanine, aspartate, and glutamate metabolism. These findings are consistent with the established mechanisms of metabolic dysregulation in obesity, wherein insulin resistance and dyslipidemia are known to drive disease progression [41,42]. Moreover, network analysis revealed GYG1 as a central node linking these interconnected pathways, emphasizing its regulatory role in metabolic modulation following GZR18 therapy. The integration of multi-omics data with network-based modeling enabled the identification of critical molecular hubs—such as GYG1—that play pivotal roles in mediating therapeutic response. This approach not only provides mechanistic insights into the pharmacodynamics of GZR18 but also underscores the multifactorial complexity of obesity as a metabolic disorder. Such pathway-level understanding may guide future therapeutic strategies, including interventions targeting specific metabolic routes to increase efficacy [43].

This research has several implications for the clinical management and pharmacotherapy of obesity. First, dual-response biomarkers such as GYG1 warrant validation in larger, ethnically diverse cohorts to confirm their generalizability across populations. Second, a detailed exploration of molecular interactions between predictive and monitoring biomarkers may reveal new pharmacologic targets for optimizing the therapeutic efficacy of GLP-1RAs and other anti-obesity agents. Third, the integration of multi-omics data with machine learning could refine predictive modeling and enable more accurate, patient-centered care.

In summary, this study demonstrates the transformative potential of multi-omics methodologies for advancing precision medicine in obesity. It contributes substantially to the personalization of obesity management by identifying biomarkers that both predict and monitor individual treatment responses. The dual-functionality of biomarkers such as GYG1, together with pathway-level therapeutic optimization, enhances understanding of the biological mechanisms underlying obesity and its pharmacologic interventions [44].

This study has several limitations. The relatively small sample size from a single center, although suitable for an intensive longitudinal multi-omics design, limits generalizability and necessitates validation in larger and more diverse populations. Moreover, the multi-omics profiling used here provided relative quantification of biomarkers; clinical translation will require absolute quantification techniques (e.g., immunoassays) to determine precise serum concentrations. Finally, although the mechanistic insights presented are supported by pathway enrichment and network analyses, functional validation in experimental models will be essential to confirm causality and elucidate downstream effects.

In conclusion, this study establishes a robust multi-omics framework for obesity pharmacotherapy, identifying GYG1 as a clinically actionable dual-response biomarker involved in glycogen and starch/sucrose metabolism. We delineate phase-specific mechanisms in which MAPK activation initiates therapeutic response, while lipid mobilization sustains long-term efficacy. Through integrated proteomic, metabolomic, and lipidomic profiling, the study defines validated biomarker panels that support precision dosing and real-time monitoring. These findings advance personalized obesity management by pinpointing therapeutic targets, interventions, and actionable pathways identified through network analysis. Future validation in larger, multiethnic populations and incorporation into real-time clinical monitoring systems are recommended.

Supplementary Material

Article information

CONFLICTS OF INTEREST

No potential conflict of interest relevant to this article was reported.

ACKNOWLEDGMENTS

This research is funded by the National Science and Technology Major Projects (2018-ZX09711003-014-001). The authors would like to acknowledge Mr. Tian Xie, Dr. Anshun He, Dr. Wei Chen, and Hangzhou Cosmos Wisdom Biotech Co., Ltd. for their valuable contributions and technical support to this study.

AUTHOR CONTRIBUTIONS

Conception or design: R.D. Acquisition, analysis, or interpretation of data: L.L., M.H., O.G., R.D. Drafting the work or revising: L.L., M.H., O.G., R.D. Final approval of the manuscript: L.L., M.H., O.G., R.D.

Fig. 1

Heterogeneous body mass index (BMI) responses to GZR18 treatment. (A) Longitudinal trajectories (weeks 0–30) demonstrate dose-dependent BMI reduction, colored by slope Z and shaped by dosing interval, with locally estimated scatterplot smoothing (LOESS) curves and 95% confidence intervals (CIs). Facets represent dosage groups. (B) Bimodal response distribution (x-axis: slope Z; dot size: −log₁₀ P; error bars: 95% CI). Color denotes dosing frequency.

Fig. 2

Multi-omics signatures of obesity drug response. (A) Baseline predictors. Volcano plots of week 0 molecular features (lipidomics, metabolomics, proteomics) regressed against body mass index (BMI) response slope (Z). Axes: regression coefficient (x) and −log₁₀ (P value) (y). Dotted line indicates P=0.05. (B) Longitudinal monitoring. Associations between dynamic omics profiles and percent BMI change across nine time points (n=221). Covariates include dose and dosing frequency. PI, phosphatidylinositol; PE, phosphatidylethanolamine; PC, phosphatidylcholine; Neg, negative; NS, not significant; Pos, positive; GPI, glucose-6-phosphate isomerase; PIGR, polymeric immunoglobulin receptor; GYG1, glycogenin 1; SCYL2, SCY1 like pseudokinase 2; NLAPE, N-acyl phosphatidylethanolamine; IGFBP2, insulin like growth factor binding protein 2; ADH4, alcohol dehydrogenase 4; TMED5, transmembrane emp24 domain containing 5; ELMOD2, engulfment and cell motility domain containing 2.

Fig. 3

Dual-response biomarker identification and validation. (A) Two-dimensional significance plot of biomarkers associated with both therapeutic prediction (y-axis: body mass index [BMI] slope Z) and longitudinal monitoring (x-axis: % BMI change). Each dot represents a molecular feature (protein, metabolite, or lipid) with P<0.05 in both models; color indicates effect direction (blue=positive [Pos], red=negative [Neg]). (B) Scatter plot of baseline glycogenin 1 (GYG1) abundance versus % BMI change (P<0.05). (C) BMI trajectories stratified by baseline GYG1 expression (median split), showing greater weight loss in high-expression groups across all doses.

Fig. 4

Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of predictive and monitoring biomarkers. (A) Gene set enrichment analysis (GSEA) results for predictive biomarkers associated with body mass index (BMI) reduction slope (Z). (B) GSEA results for monitoring biomarkers associated with longitudinal BMI changes. Features were ranked by regression coefficient × −log₁₀ (P value) and analyzed against KEGG pathways using clusterProfiler. Plots display running enrichment scores (ES; y-axis) versus ranked features (x-axis), with feature weights (log₂ fold change [fc]) shown as upper bar plots. Significant pathways (adjusted P<0.05) are labeled with normalized enrichment scores (NES) and ES values. MAPK, mitogen-activated protein kinas; CoA, coenzyme A; Rap1, Rasassociated protein 1.

Fig. 5

Multi-omics interaction networks of predictive and monitoring biomarkers. (A) Network of predictive biomarkers (body mass index [BMI] reduction slope). (B) Network of monitoring biomarkers (longitudinal BMI change). Networks integrate omics features with Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways from gene set enrichment analysis (GSEA) (Fig. 4), highlighting mitogen-activated protein kinase (MAPK) signaling and protein biosynthesis modules in (A) and dynamic metabolic clusters in (B). Nodes encode regulation direction (red=upregulated, green=downregulated), molecule type (shape), and statistical significance (−log₁₀ P). Edges denote interactions (STRING/KEGG). Central hubs (e.g., glycogenin 1 [GYG1]) and shared modules emphasize mechanistic overlap.

Fig. 6

Glycogenin 1 (GYG1) interaction network and pathway analysis. (A) Protein–protein interaction network showing GYG1 and its five direct interactors (glycogen synthase 1 [GYS1], amylo-alpha-1,6-glucosidase, 4-alpha-glucanotransferase [AGL], glycogen phosphorylase, brain form [PYGB], glycogen branching enzyme 1 [GBE1], and uridine diphosphate glucose pyrophosphorylase 2 [UGP2]), annotated with interaction metrics. The accompanying heatmap depicts their coordinated pathway participation. (B) Quantitative comparison of GYG1 pathway contributions between predictive (bottom) and monitoring (top) models, displaying effect size and statistical significance. KEGG, Kyoto Encyclopedia of Genes and Genomes; GPI, glucose-6-phosphate isomerase; PRKAB1, protein kinase, AMP-activated, beta 1 non-catalytic subunit; PIK3R1, phosphoinositide-3-kinase regulatory subunit 1; PPP1CC, protein phosphatase 1 catalytic subunit gamma; FLOT1, flotillin 1; RHEB, Ras homolog, mTORC1 binding; PFKP, phosphofructokinase, platelet.

References

• 1. GBD 2021 Adolescent BMI Collaborators. Global, regional, and national prevalence of child and adolescent overweight and obesity, 1990–2021 with forecasts to 2050: a forecasting study for the Global Burden of Disease Study 2021. Lancet 2025;405:785–812.PubMedPMC
• 2. World Health Organization. Obesity and overweight [Internet] Geneva: WHO; 2025 [cited 2025 Dec 9]. Available from: https://www.who.int/news-room/fact-sheets/detail/obesity-and-overweight.
• 3. Ahmed SK, Mohammed RA. Obesity: prevalence, causes, consequences, management, preventive strategies and future research directions. Metabol Open 2025;27:100375.ArticlePubMedPMC
• 4. Masenga SK, Kabwe LS, Chakulya M, Kirabo A. Mechanisms of oxidative stress in metabolic syndrome. Int J Mol Sci 2023;24:7898.ArticlePubMedPMC
• 5. Dina C, Tit DM, Radu A, Bungau G, Radu AF. Obesity, dietary patterns, and cardiovascular disease: a narrative review of metabolic and molecular pathways. Curr Issues Mol Biol 2025;47:440.ArticlePubMedPMC
• 6. Valencia O, Lopez C, Vanegas-Duarte E, Fillizola C, Bejarano Ramirez DF, Cortes Mejia NA, et al. Risk factors related to the development of nonalcoholic fatty liver: a systematic review. Can J Gastroenterol Hepatol 2025;2025:9964486.ArticlePubMedPMCPDF
• 7. Sung H, Siegel RL, Torre LA, Pearson-Stuttard J, Islami F, Fedewa SA, et al. Global patterns in excess body weight and the associated cancer burden. CA Cancer J Clin 2019;69:88–112.ArticlePubMedPDF
• 8. Kong Y, Yang H, Nie R, Zhang X, Zuo F, Zhang H, et al. Obesity: pathophysiology and therapeutic interventions. Mol Biomed 2025;6:25.ArticlePubMedPMCPDF
• 9. Wilding JPH, Batterham RL, Calanna S, Davies M, Van Gaal LF, Lingvay I, et al. Once-weekly semaglutide in adults with overweight or obesity. N Engl J Med 2021;384:989–1002.ArticlePubMed
• 10. Müller TD, Bluher M, Tschop MH, DiMarchi RD. Anti-obesity drug discovery: advances and challenges. Nat Rev Drug Discov 2022;21:201–23.ArticlePubMedPDF
• 11. Thota-Kammili S, Tama E, Hurtado Andrade MD. Glucagon-like peptide 1 receptor agonists for obesity: efficacy, side effects, and risks. Curr Treat Options Gastroenterol 2024;22:145–54.ArticlePDF
• 12. Jastreboff AM, Aronne LJ, Ahmad NN, Wharton S, Connery L, Alves B, et al. Tirzepatide once weekly for the treatment of obesity. N Engl J Med 2022;387:205–16.ArticlePubMed
• 13. Gomez-Martin M, Canfell OJ, Chai LK, Jansson AK, Littlewood R, Sullivan C, et al. What is the role of primary prevention of obesity in an age of effective pharmaceuticals? Curr Obes Rep 2025;14:39.PubMedPMC
• 14. Chakhtoura M, Haber R, Ghezzawi M, Rhayem C, Tcheroyan R, Mantzoros CS. Pharmacotherapy of obesity: an update on the available medications and drugs under investigation. EClinicalMedicine 2023;58:101882.ArticlePubMedPMC
• 15. Schmitz SH, Saunders KH, Aronne LJ. Cutting-edge approaches to obesity management: the latest pharmacological options. Endocrinol Metab Clin North Am 2025;54:85–102.ArticlePubMed
• 16. Liu Y, Chen W, He X, He A, Zhao L, Xie T, et al. The safety, tolerability, pharmacokinetics and pharmacodynamics of GZR18 in healthy American and Chinese adult subjects. Diabetes Obes Metab 2025;27:2777–89.PubMed
• 17. Abstracts from the 42nd Annual Meeting of the Obesity Society at Obesityweek(R) November 3–6, 2024. Obesity (Silver Spring) 2024;32(Suppl 1):5–308.PubMed
• 18. Payab M, Tayanloo-Beik A, Falahzadeh K, Mousavi M, Salehi S, Djalalinia S, et al. Metabolomics prospect of obesity and metabolic syndrome; a systematic review. J Diabetes Metab Disord 2022;21:889–917.ArticlePubMedPDF
• 19. Aleksandrova K, Egea Rodrigues C, Floegel A, Ahrens W. Omics biomarkers in obesity: novel etiological insights and targets for precision prevention. Curr Obes Rep 2020;9:219–30.ArticlePubMedPMCPDF
• 20. Uematsu S, Ohno S, Tanaka KY, Hatano A, Kokaji T, Ito Y, et al. Multi-omics-based label-free metabolic flux inference reveals obesity-associated dysregulatory mechanisms in liver glucose metabolism. iScience 2022;25:103787.ArticlePubMedPMC
• 21. Woldemariam S, Dorner TE, Wiesinger T, Stein KV. Multiomics approaches for precision obesity management: potentials and limitations of omics in precision prevention, treatment and risk reduction of obesity. Wien Klin Wochenschr 2023;135:113–24.PubMedPMC
• 22. Wang H, Zhuang Y, Hua R, Yao T, Lin K, Zhang Y, et al. Multi-omics exploration of obesity biomarkers in sedentary and weight loss cohorts. Phenomics 2025;5:137–53.ArticlePubMedPDF
• 23. Mir FA, Mall R, Ullah E, Iskandarani A, Cyprian F, Samra TA, et al. An integrated multi-omic approach demonstrates distinct molecular signatures between human obesity with and without metabolic complications: a case-control study. J Transl Med 2023;21:229.ArticlePubMedPMCPDF
• 24. Alfadda AA, Abdel Rahman AM, Benabdelkamel H, Al-Malki R, Alsuwayni B, Alhossan A, et al. Metabolomic effects of liraglutide therapy on the plasma metabolomic profile of patients with obesity. Metabolites 2024;14:500.ArticlePubMedPMC
• 25. Geyer PE, Kulak NA, Pichler G, Holdt LM, Teupser D, Mann M. Plasma proteome profiling to assess human health and disease. Cell Syst 2016;2:185–95.ArticlePubMed
• 26. Dunn WB, Broadhurst D, Begley P, Zelena E, Francis-Mc-Intyre S, Anderson N, et al. Procedures for large-scale metabolic profiling of serum and plasma using gas chromatography and liquid chromatography coupled to mass spectrometry. Nat Protoc 2011;6:1060–83.ArticlePubMedPDF
• 27. Yin R, Wang X, Li K, Yu K, Yang L. Lipidomic profiling reveals distinct differences in plasma lipid composition in overweight or obese adolescent students. BMC Endocr Disord 2021;21:201.ArticlePubMedPMCPDF
• 28. Sen PK. Estimates of the regression coefficient based on Kendall’s tau. J Am Stat Assoc 1968;63:1379–89.Article
• 29. Wilcox RR. Fundamentals of modern statistical methods: substantially improving power and accuracy; New York: Springer Science & Business Media; 2010.
• 30. Wu T, Hu E, Xu S, Chen M, Guo P, Dai Z, et al. ClusterProfiler 4.0: a universal enrichment tool for interpreting omics data. Innovation (Camb) 2021;2:100141.ArticlePubMedPMC
• 31. Ard J, Fitch A, Fruh S, Herman L. Weight loss and maintenance related to the mechanism of action of glucagon-like peptide 1 receptor agonists. Adv Ther 2021;38:2821–39.ArticlePubMedPMCPDF
• 32. Klen J, Dolzan V. Glucagon-like peptide-1 receptor agonists in the management of type 2 diabetes mellitus and obesity: the impact of pharmacological properties and genetic factors. Int J Mol Sci 2022;23:3451.ArticlePubMedPMC
• 33. Rodriguez-Munoz A, Motahari-Rad H, Martin-Chaves L, Benitez-Porres J, Rodriguez-Capitan J, Gonzalez-Jimenez A, et al. A systematic review of proteomics in obesity: unpacking the molecular puzzle. Curr Obes Rep 2024;13:403–38.ArticlePubMedPMCPDF
• 34. Cai Y, Liu P, Xu Y, Xia Y, Peng X, Zhao H, et al. Biomarkers of obesity-mediated insulin resistance: focus on microR-NAs. Diabetol Metab Syndr 2023;15:167.ArticlePubMedPMCPDF
• 35. Weng TH, Pien YC, Chen CJ, Chen PP, Tseng YT, Chen YC, et al. Human glycogenins maintain glucose homeostasis by regulating glycogen metabolism. Nat Commun 2025;16:6556.ArticlePubMedPMCPDF
• 36. Adeva-Andany MM, Gonzalez-Lucan M, Donapetry-Garcia C, Fernandez-Fernandez C, Ameneiros-Rodriguez E. Glycogen metabolism in humans. BBA Clin 2016;5:85–100.ArticlePubMedPMC
• 37. Petersen MC, Shulman GI. Mechanisms of insulin action and insulin resistance. Physiol Rev 2018;98:2133–223.ArticlePubMedPMC
• 38. Taneera J, Lang S, Sharma A, Fadista J, Zhou Y, Ahlqvist E, et al. A systems genetics approach identifies genes and pathways for type 2 diabetes in human islets. Cell Metab 2012;16:122–34.ArticlePubMed
• 39. FDA-NIH Biomarker Working Group. BEST (Biomarkers, EndpointS, and other Tools) resource Silver Spring; Food and Drug Administration (US): 2016 [cited 2025 Dec 9]. Available from: https://www.ncbi.nlm.nih.gov/books/NBK326791.
• 40. Califf RM. Biomarker definitions and their applications. Exp Biol Med (Maywood) 2018;243:213–21.ArticlePubMedPMCPDF
• 41. Nimptsch K, Konigorski S, Pischon T. Diagnosis of obesity and use of obesity biomarkers in science and clinical medicine. Metabolism 2019;92:61–70.ArticlePubMed
• 42. Yang M, Liu S, Zhang C. The related metabolic diseases and treatments of obesity. Healthcare (Basel) 2022;10:1616.ArticlePubMedPMC
• 43. Zhou G, Li S, Xia J. Network-based approaches for multi-omics integration. Methods Mol Biol 2020;2104:469–87.ArticlePubMed
• 44. Cifuentes L, Hurtado AMD, Eckel-Passow J, Acosta A. Precision medicine for obesity. Dig Dis Interv 2021;5:239–48.ArticlePubMedPMC

Figure & Data

References

Citations

Citations to this article as recorded by