Non-invasive detection biomarker for intrauterine condition of fetus with gestational diabetes mellitus and application thereof

Abstract

The application discloses a non-invasive detection biomarker for prompting intrauterine condition of a fetus of gestational diabetes mellitus and application thereof, the biomarker is a metabolite that can be detected from maternal blood, and the metabolite comprises betaine, taurine, pyruvic acid and D-mannose. The application also provides application of the biomarker in preparation of a detection product, detection of the concentration of the biomarker in peripheral blood of a pregnant woman, input of the concentration data into a pre-constructed fetus state evaluation model, construction of the model based on a fetus anchored cross-omics training set, and the training set containing paired maternal metabolome data and fetus / placenta molecular omics data, and evaluation of the possibility of a fetus existing metabolic abnormality risk according to a model output value. The application first realizes non-invasive, accurate and specific evaluation of the intrauterine metabolic health state of a fetus of a GDM pregnant woman through maternal blood, and makes up for the defects that the prior art can only diagnose maternal diseases and cannot evaluate fetus involvement.

Description

Technical Field

[0001] This invention relates to the field of biomedical detection technology, specifically to a non-invasive biomarker for indicating the intrauterine condition of a fetus with gestational diabetes mellitus (GDM), and its application in the preparation of related detection products and evaluation methods. Background Technology

[0002] Gestational diabetes mellitus (GDM) is an abnormal glucose metabolism state detected during pregnancy. Its prevalence is rising globally, and it has adverse effects on the short-term and long-term health of both mother and child. In addition to increasing the risk of perinatal complications for the mother, GDM can lead to fetal developmental abnormalities and increase the risk of long-term obesity, insulin resistance, and metabolic syndrome in offspring. Currently, the gold standard for clinical diagnosis of GDM is an oral glucose tolerance test (OGTT) performed between 24 and 28 weeks of gestation. However, this method has a relatively late detection window, a cumbersome procedure, and poor subject compliance, thus limiting its application in early screening and precise intervention. More importantly, the OGTT only reflects the mother's current glucose metabolism status and cannot provide crucial information on whether the fetus has been adversely affected by maternal hyperglycemia, i.e., it cannot assess the fetus's intrauterine metabolic health.

[0003] Currently, there is a lack of precise early diagnostic strategies for fetal metabolic abnormalities, and there is a lack of validated technologies that can non-invasively capture fetal physiological changes through maternal blood. In addition, most existing studies focus only on maternal indicators at a single mic level, or report sporadically on changes in the placenta and umbilical cord blood, lacking an integrated analytical framework that clearly anchors fetal abnormalities and systematically maps them to detectable molecules in maternal blood.

[0004] As is well known, the exchange of substances and signals between mother and fetus proceeds along the axis of "maternal peripheral blood → placenta → umbilical cord → umbilical blood," a process characterized by bidirectional and hierarchical regulation. Umbilical blood, as a key medium for this exchange, can sensitively reflect the fetal internal environment and metabolic stress state. At the maternal-fetal interface, human umbilical vein endothelial cells (HUVECs) possess barrier and signal integration functions; their transcriptome is highly sensitive to hyperglycemia, lipid imbalance, and inflammatory stimuli, capturing the molecular characteristics of placental endothelial remodeling. In the fetal-maternal direction, changes in the placental microenvironment and fetal endothelial phenotype affect maternal circulation by regulating signal transduction. These changes are reflected in the umbilical blood proteome, manifesting as reprogramming of pathways related to lipoprotein transport, complement-coagulation cascades, and extracellular matrix / adhesion. Simultaneously, soluble factors and extracellular vesicles / exosomes released from the placenta enter maternal circulation, ultimately forming quantifiable molecular characteristics in the maternal peripheral blood metabolome, thus reflecting fetal abnormalities. Conversely, from the mother to the fetus, the homeostasis of maternal glucose, lipids, and amino acids is sensed and regulated through the transporter and receptor network in the syncytiotrophoblast and placental vascular bed, thereby affecting transplacental flux and fetal exposure levels, ultimately shaping the intrauterine environment and fetal organ development.

[0005] Therefore, there is an urgent need for an integrated strategy that can anchor to the fetal side and map fetal abnormalities back to detectable molecules in maternal peripheral blood, in order to achieve non-invasive identification and early risk stratification of fetal metabolic disorders. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a non-invasive biomarker for assessing the intrauterine condition of the fetus in gestational diabetes mellitus (GDM) and its application. This biomarker enables non-invasive and specific assessment of the intrauterine metabolic health status of the fetus in pregnant women with GDM through maternal blood.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: This invention provides a non-invasive biomarker for detecting the intrauterine condition of a fetus with gestational diabetes mellitus, wherein the biomarker is one or more of betaine, taurine, pyruvate, and D-mannose.

[0008] The present invention also provides an application of the above-mentioned biomarkers in the preparation of a detection product, the detection product being used to assess the intrauterine metabolic health status of GDM fetuses; The applications include: The concentration of the biomarker was determined based on peripheral blood samples from pregnant women; Based on the concentration of the biomarker, and in conjunction with a fetal status assessment model associated with the biomarker, the likelihood of GDM-related metabolic abnormalities in the pregnant woman and the fetus in the womb is assessed.

[0009] Furthermore, the fetal status assessment model is associated with the biomarker, and the concentration of the biomarker is used as the input of the fetal status assessment model. The fetal status assessment model outputs risk assessment results for the pregnant woman to be tested having GDM and fetal metabolic status assessment results.

[0010] Furthermore, the fetal status assessment model is a multi-omics integrated assessment model constructed based on a fetal-anchored training set; the training set contains multi-omics data from paired samples from GDM pregnant women and healthy control pregnant women; the paired samples include: maternal peripheral blood samples, placental interface tissue or cell samples from the same pregnancy, and fetal umbilical cord blood samples; the multi-omics data includes at least: metabolomic data from maternal peripheral blood samples, transcriptomic or proteomic data from placental interface tissue or cell samples, and proteomic or metabolomic data from fetal umbilical cord blood samples.

[0011] Furthermore, the construction of the multi-omics integrated evaluation model includes: preprocessing and standardizing the multi-omics data of the training set; using a cross-omics data integration algorithm to mine the association network between maternal metabolomics data, placental interface molecular omics data, and fetal circulation molecular omics data; and based on the association network, screening out a set of maternal metabolic biomarkers and their feature weights that are consistent with the direction of molecular changes on the fetal side in the GDM state and have strong statistical association.

[0012] Furthermore, the cross-omics data integration algorithm includes a bidirectional orthogonal partial least squares model; the mining of the association network includes calculating the correlation between maternal metabolites and fetal gene or protein expression levels, and identifying metabolite-molecule pairs with consistent orientation and correlation strength exceeding a preset threshold.

[0013] Furthermore, the association network includes at least one of the following metabolite-molecule pairs: betaine with SCARB1 positive correlation pair, taurine with GPX8 negative correlation pair, pyruvate with GALK2 positive correlation pair, and D-mannose with AKR1B1 positive correlation pair.

[0014] Furthermore, the assessment results output by the fetal status assessment model are presented in the form of GDM risk probability, risk level, and scores indicating the degree of abnormalities in fetal lipid metabolism, oxidative stress, and glucose metabolism.

[0015] Furthermore, placental interface tissue or cell samples include human umbilical vein endothelial cells, placental villus tissue, or trophoblast cells; placental interface molecular omics data are transcriptomic data from HUVECs; and fetal circulation molecular omics data are proteomic data from umbilical cord blood.

[0016] The research approach of this invention is as follows: A multi-omics framework anchored to the fetus is employed. This framework begins with HUVEC transcriptomics analysis via RNA sequencing and umbilical cord blood proteomics quantified through data-independent collection, followed by maternal plasma metabolomics analysis via ultra-high performance liquid chromatography-high resolution mass spectrometry. Differential analysis, pathway enrichment, and interaction network mapping were performed. Subsequently, cross-omics transcript-metabolite and protein-metabolite associations were constructed to identify convergent pathways and pivotal nodes, thereby obtaining an integrated map for prioritizing candidate biomarkers.

[0017] Experimental results demonstrate that this framework identifies unique metabolic reprogramming in maternal plasma reflecting fetal lateral transcriptional and proteomic abnormalities. HUVEC transcriptomics revealed alterations in inflammation and metabolic signaling, while umbilical cord blood proteomics highlighted changes in lipid transport, complement-coagulation, and adhesion pathways. Maternal metabolomics revealed systemic shifts in amino acid, lipid, and nucleotide metabolism. Cross-omics correlation analysis and bidirectional orthogonal partial least squares modeling demonstrated consistent covariance across the three levels, such as betaine–SCARB1, D-mannose–AKR1B1, pyruvate–GALK2, and taurine–GPX8, linking maternal metabolites to fetal molecular states. Ultimately, betaine, taurine, pyruvate, and D-mannose were preferentially identified as core maternal blood biomarkers, demonstrating strong discriminative ability for GDM and potential for early non-invasive risk stratification of fetal metabolic abnormalities.

[0018] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention establishes a multi-omics analysis framework anchored to the fetus and spanning the fetal-maternal axis to characterize non-invasive markers of fetal metabolic alterations associated with GDM in maternal peripheral blood. Integrated multi-omics analysis revealed that molecular changes observed in HUVECs, cord blood, and maternal peripheral blood during GDM pregnancies were primarily enriched in biological pathways related to inflammatory stress response, lipid and amino acid metabolism, and redox homeostasis. Furthermore, this invention identified betaine, taurine, pyruvate, and D-mannose as potential circulating biomarkers in maternal blood that reliably reflect fetal metabolic status and exhibit strong discriminative properties. These metabolites are not only stably present in maternal circulation but also align with the changes observed in fetal cord blood and HUVECs, thus supporting the feasibility of a novel, non-invasive, and biologically interpretable method for early identification of fetal metabolic disorders and assessment of GDM-related risk.

[0019] (2) This invention identified four maternal circulating metabolites (betaine, taurine, pyruvate, and D-mannose) as candidate biomarkers closely related to fetal metabolic disorders, forming a complex metabolic profile reflecting fetal molecular abnormalities. These metabolites not only differentiate GDM from normal pregnancy at the metabolic level, but also share mechanistic consistency with the dysregulated pathways revealed by HUVEC transcriptomics and umbilical cord blood proteomics, thus establishing a translayered, biologically interpretable link within the maternal-fetal metabolic axis. This complex profile illustrates how placental molecular signals are projected into maternal circulation through metabolic pathways, providing a biologically verifiable basis for non-invasive assessment of fetal metabolic function.

[0020] (3) This invention establishes and validates a multi-omics integrated model along the "maternal peripheral blood-HUVECs-fetal circulation" axis, revealing the bidirectional coupling between maternal and fetal metabolic networks. The innovation of this framework lies in demonstrating that fetal molecular stress and metabolic abnormalities can be stably and non-invasively reflected through small molecule signals in maternal peripheral blood. Maternal metabolic profiles not only capture systemic insulin resistance and energy redistribution, but also interact dynamically with placental metabolic regulation and signal transduction. Attached Figure Description

[0021] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0022] Figure 1 For HUVEC transcriptional aberrations and their cross-omics coupling with maternal metabolites: (A) Volcano plot showing significantly differentially expressed genes (DEGs) (red, upregulated; green, downregulated; yellow, not significant). (B) Mapping DEGs to the STRING database to construct a protein-protein interaction network. Using the CytoHubba plugin in Cytoscape, nodes were scored using five algorithms (MCC, degree centrality, bottleneck centrality, proximity centrality, and betweenness centrality). The intersection of the top 10% of genes ranked by each algorithm yielded 19 hub genes. (C) Heatmap of the 19 hub genes (normalized by Z-score; samples were hierarchically clustered) showing distinguishable group patterns (color from green to red indicates expression levels from low to high). (D) KEGG enrichment analysis of DEGs.

[0023] Figure 2 For umbilical cord blood protein abnormalities and their cross-omics coupling with maternal metabolites: (A) Volcano plot. (B) Mapping DEPs to the STRING database to construct a PPI network, and using the CytoHubba plugin in Cytoscape, scoring them using five algorithms; taking the intersection of the top 10% of proteins from each algorithm, 17 pivot proteins were obtained. (C) Heatmap of the 17 pivot proteins; (D) KEGG enrichment analysis of DEPs.

[0024] Figure 3 An overview of the intergroup separation and differences in maternal peripheral blood metabolomics: PCA plots of clinical samples in (A) ESI+ and (B) ESI- modes show the clustering of the normal control group, GDM patient group, and mixed quality control samples. OPLS-DA plots in (C) ESI+ and (D) ESI- modes further demonstrate the significant distinction between the normal group and the GDM group. 200 OPLS-DA permutation tests performed in (E) ESI+ and (F) ESI- modes indicate no overfitting of the model, supporting its robustness. Volcano plots of the metabolomics profile in (G) ESI+ and (H) ESI- modes highlight significantly altered metabolites, showing the global distribution of up- and down-regulation characteristics. In the PCA and OPLS-DA plots, green dots represent normal pregnancies (n = 27), red dots represent GDM cases (n = 27), and yellow dots represent QC samples.

[0025] Figure 4 Heatmaps and VIP rankings for core differentially expressed metabolites: (A) Heatmap of the top 20 differentially expressed metabolites (DMs), showing high / low expression levels consistent with group, with colors ranging from green (low) to red (high). (B) VIP values ​​obtained from OPLS-DA in ESI+ and (C) ESI- modes, sorted from high to low, highlighting candidates with strong discriminative contributions; the right panel displays metabolite levels by group (red, upregulated; green, downregulated).

[0026] Figure 5 KEGG pathway enrichment for differentially expressed metabolites: KEGG over-enrichment analysis of maternal peripheral blood in (A) ESI+ and (B) ESI- modes. In the bubble plot, the X-axis represents the proportion of metabolites (number of metabolites hit in the pathway / total number of metabolites in the pathway), the bubble size represents the number of hit metabolites, the color depth reflects significance (-log10q), and the background represents the characteristics of all metabolites detected and passed quality control in this study.

[0027] Figure 6Cross-omics coupling of maternal metabolites and fetal molecules: (A) O2PLS joint loading plot revealing covariation relationships between maternal DMs and HUVEC DEGs on the same potential component (red, DMs; green, DEGs). (B) O2PLS joint loading plot revealing covariation relationships between maternal DMs and umbilical cord blood DEPs on the same potential component (red, DMs; green, DEPs). (C) Gene-metabolite triangulation heatmap showing significantly paired covariation relationships; color-coded correlation coefficients r (-1 to 1), cell labels indicating r values. (D) Protein-metabolite triangulation heatmap showing significantly paired covariation relationships; color-coded correlation coefficients r (-1 to 1), cell labels indicating r values.

[0028] Figure 7 To assess the performance of dual-path screening for key candidate metabolites, biomarkers were screened from 33 candidate differential metabolites using two complementary pathways. In MetaboAnalyst / Cytoscape–CytoHubba, the intersection of the top 10% metabolites ranked by each of the five algorithms (MCC, degree centrality, bottleneck centrality, proximity centrality, and betweenness centrality) identified D-mannose, betaine, taurine, and pyruvate (A), which showed high ROC performance (B) and clear differences in expression distribution between groups (C). Based on Pearson correlation with groups (p<0.05), L-glutamine, taurine, pyrrole-2-carboxylic acid, and piperidine (D) were identified, also achieving high AUC values ​​(E) and significant differences between groups (F); the two pathways converged on taurine. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] 1. Sample and Data Collection This invention establishes an integrated multi-omics framework encompassing metabolomics from maternal peripheral blood, transcriptomics from HUVECs, and proteomics from offspring umbilical cord blood. GDM diagnosis is based on WHO / IADPSG criteria. The control group consisted of age-matched, healthy pregnant women with normal OGTT results at the same time. Eligible participants were limited to singleton pregnancies with complete clinical and high-quality biological samples. Furthermore, women with a history of diabetes, impaired glucose regulation, severe comorbidities, or using medications affecting glucose and lipid metabolism were excluded.

[0031] Twenty-seven mothers with gastroparesis (GDM) and 27 control mothers were enrolled for peripheral blood metabolomics analysis; three GDM mothers and three control mothers (isolated from their respective umbilical cords) underwent HUVEC transcriptomics analysis; and five GDM newborns and five control newborns underwent umbilical cord blood proteomics analysis. All samples were collected and processed according to standardized operating procedures to ensure consistency.

[0032] 2. Untargeted metabolomics analysis based on UHPLC-HRMS Untargeted metabolomics analysis of maternal peripheral blood plasma samples was performed using ultra-high performance liquid chromatography-high resolution mass spectrometry (UHPLC-HRMS). Samples were pretreated at 4°C before analysis, employing bipolar (ESI⁺ / ESI⁻) scanning mode. Raw mass spectrometry data were processed using XCMS software for peak detection, alignment, isotope removal, and adduct removal, generating characteristic peak tables in positive and negative ion modes. Missing values ​​were filled using k-nearest neighbor method, total ion current was normalized, log₂ transformed, and Pareto scaled. When significant instrument signal drift was detected, the ComBat algorithm was used for correction. The quality control standard required that the median relative standard deviation of metabolic characteristics in the pooled quality control samples not exceed 30%.

[0033] Principal component analysis was used to assess overall data quality, and orthogonal partial least squares discriminant analysis was employed to analyze differences between groups. Screening criteria for differentially expressed metabolites included: a variable importance projection value greater than 1, an absolute value of log2-fold change not less than 1, and a p-value less than 0.05 after multiple test correction. Metabolite identification was based on precise quality, retention time, and MS / MS spectra matched with the HMDB database. KEGG pathway enrichment analysis was performed on differentially expressed metabolites under both positive and negative ion modes.

[0034] 3. Transcriptomics analysis Umbilical cords were collected immediately postpartum, and primary human umbilical vein endothelial cells (HUVECs) were isolated and cultured. Total RNA was extracted from the cells using the TRIzol method, and RNA integrity was assessed by agarose gel electrophoresis. Library construction included enrichment of poly(A)+ mRNA using capture magnetic beads, followed by conversion to cDNA via first- and second-strand synthesis. The library was purified and size-selected using HieffNGS® DNA Selection Beads, amplified by PCR, and sequenced on the Illumina NovaSeq XPlus platform.

[0035] The bioinformatics analysis workflow is as follows: Raw sequencing data were filtered using FASTP software; rRNA mapping reads were removed using Bowtie2, and clean reads were aligned to the human reference genome GRCh38 using HISAT2; transcript assembly and quantification were performed using StringTie to obtain reads per million transcripts; differentially expressed genes were screened using the DESeq2 software package with a false discovery rate of less than 0.05 and |log2FC| ≥ 1 as thresholds. GO functional annotation and KEGG pathway enrichment analysis were performed on the differentially expressed genes, and a protein-protein interaction network was constructed using the STRING database and visualized using Cytoscape software.

[0036] 4. Proteomics analysis Umbilical cord blood was collected postpartum, plasma was separated and aliquoted, and stored at -80°C. For proteomics analysis, plasma proteins were enriched using magnetic nanoparticle beads, followed by bead denaturation, Lys-C / trypsin digestion, and C18 desalting before loading. Peptides were analyzed using a nanoElute 2 coupled to a time-to-free (TOF) Pro 2. Raw files were processed in Spectronaut 19 software, using dynamic iRT calibration and a target-decoy competition search based on mutational decoys. The search was performed against a human reference proteome, with FDR controlled at 1% for precursor, peptide, and protein levels. Local normalization was applied, and proteomics quantification was performed using MaxLFQ.

[0037] 5. Multi-omics integrated analysis To analyze the cross-layer coupling problem along the maternal-placental-fetal axis, DMs (maternal plasma) from maternal metabolomics were used as metabolic inputs, the top 100 DEGs (differentially expressed genes) from HUVEC transcriptomics were used as transcriptomic inputs, and the top 80 DEPs (differentially expressed proteins) from umbilical cord blood proteomics were used as proteomic inputs. Only participants with paired data at all three levels were included. Each dataset underwent log2 transformation and Z-score normalization; ComBat correction was applied when batch or time drift was detected. Cross-omics modeling used bidirectional O2PLS in mixOmics. The number of joint / orthogonal components was selected through repeated 5–7 fold cross-validation. At the paired level, the full matrix of gene-metabolite and protein-metabolite Pearson correlations was calculated, and pairs with orientations consistent with and significantly correlated with joint loads were retained to construct a bipartite network, further supplemented with high-confidence gene-gene and protein-protein edges from STRING to generate a multilayer network. In Cytoscape, hub nodes are defined using CytoHubba, and functional enrichment analysis is performed through hypergeometric tests, with significant cross-layer edges mapped to shared paths to identify coupled modules.

[0038] 6. Selection of candidate biomarkers Starting with differentially identified metabolites from non-targeted metabolomics, two complementary prioritization pathways were implemented to obtain robust candidates. In the network importance pathway, metabolite networks were constructed in MetaboAnalyst and Cytoscape using KEGG reaction adjacency and intra-cohort correlation; node centrality was scored using CytoHubba, and the network priority set was defined by the intersection of the top 10% of metabolites ranked by each algorithm.

[0039] In the statistical correlation approach, group states are binarized, Pearson correlation between each metabolite and the group is calculated, and significant metabolites are ranked according to |r|.

[0040] Consistency filtering retained metabolites that were aligned with the O2PLS combined payload, showed significant gene-metabolite or protein-metabolite correlations, and mapped to the KEGG pathway enriched in this study. Metabolites identified in both pathways were included in diagnostic evaluation. ROC analysis estimated AUC through 1000 bootstrap iterations; the optimal cutoff value was defined using the Youden index, and sensitivity and specificity were reported. AUCs were compared using the DeLong test, and internal stability was assessed using five-fold cross-validation. Sensitivity analysis used logistic regression to adjust for age, gestational age, and BMI. Final candidates were annotated according to MSI guidelines; ROC was performed in R, and network analysis was performed in Cytoscape.

[0041] 7. Statistical Analysis Statistical analysis and data visualization were performed using R and GraphPad Prism. Data are expressed as mean ± standard deviation. Normality was assessed using the Shapiro–Wilk test, and one-way or two-way ANOVA was used for comparisons between groups. FDR correction was performed using the BH method. Unless otherwise stated, two-sided p < 0.05 or q < 0.05 was considered statistically significant. Multivariate analysis, correlation tests, and ROC-based assessments followed the analytical criteria defined in the respective omics sections.

[0042] 1. HUVECs in GDM pregnancies exhibit transcriptional reprogramming in terms of inflammation and metabolism. To investigate transcriptional alterations in HUVECs associated with GDM and explore their potential links to maternal metabolic changes along the fetal-maternal axis, this invention performed RNA-Seq analysis and used DESeq2 to identify DEGs (see [link to DESeq2]). Figure 1 ). Figure 1 The volcano plot of A reveals a clear pattern of gene dysregulation. To depict key regulatory nodes in this transcriptomic landscape, node centrality was evaluated in Cytoscape using five algorithms, and the intersection of the top 10% of genes from each of the five algorithms identified 19 pivotal genes (see [link to Cytoscape]). Figure 1 B). Unsupervised clustering of hub genes clearly distinguished GDM from control samples (see B). Figure 1 C). For subsequent cross-omics integration, the top 100 DEGs were selected for O2PLS modeling and transcript-metabolite correlation analysis.

[0043] KEGG pathway enrichment analysis of the top 100 DEGs further elucidated the functional relevance of these transcriptional changes, with significantly enriched pathways including those associated with inflammatory and metabolic pathways (see [link to KEGG analysis]). Figure 1 D). These enriched terms collectively highlight two core functional modules: inflammatory / stress signaling and the reprogramming of energy, lipid, and amino acid metabolism.

[0044] 2. Proteomic alterations in umbilical cord blood in GDM highlight lipid metabolism, coagulation / inflammation. To characterize changes in fetal circulating proteomics and further explore the fetal-maternal axis in GDM, umbilical cord blood proteomics were quantitatively analyzed using DIA. Figure 2 As shown in Figure A, differential expression analysis was performed using a linear model, revealing 96 DEPs. To identify key functional effectors on the fetal side, we evaluated node centrality in Cytoscape using five algorithms; the intersection of the top 10% of proteins from each algorithm identified 17 pivotal proteins (see Figure A). Figure 2B). Unsupervised clustering of pivotal proteins clearly distinguished GDM from control samples (see B). Figure 2 (C) In order to integrate with metabolomics across omics, the top 80 DEPs selected based on comprehensive ranking were used for subsequent protein-metabolite correlation analysis.

[0045] To explore the biological significance of these proteins, a KEGG pathway enrichment analysis was performed on 80 DEPs, revealing their broad involvement in lipid and cholesterol metabolism pathways, complement and coagulation processes, endothelial adhesion and ECM interactions, and energy metabolism (see [link to analysis]). Figure 2 (D). These enriched pathways collectively characterize four main functional modules: lipid homeostasis and lipoprotein metabolism, coagulation / inflammatory response, endothelial adhesion and ECM remodeling, and energy metabolism. This modular organization provides a functional framework for the joint analysis of maternal metabolic changes and fetal circulating protein alterations in GDM.

[0046] 3. Clear separation of maternal metabolic profiles between GDM and healthy control groups To assess systemic metabolic differences between women with GDM and women with normal glucose tolerance during pregnancy, untargeted metabolomics analysis of maternal plasma was performed using UHPLC-HRMS. Figure 3 A and Figure 3 As shown in Figure B, PCA revealed tight clustering of QC samples in both ESI+ and ESI- modes, confirming analytical stability and reproducibility. Simultaneously, PCA revealed a clear separation between GDM and the control group, indicating significant differences in the overall metabolite profile. OPLS-DA further enhanced inter-group differentiation, observing significant clustering in both ESI+ and ESI- modes (see Figure B). Figure 3 C and Figure 3 D). Figure 3 E and Figure 3 The F-test confirmed the model's robustness, with no evidence of overfitting. Furthermore, the OPLS-DA model demonstrated high explanatory and predictive power.

[0047] Figure 3 G and Figure 3 Volcano plot visualization in H revealed extensive metabolic alterations, showing a large number of significantly upregulated and downregulated features in both ESI+ and ESI- modes. Based on the criteria of VIP>1, |log2~FC| ≥ 1, and P<0.05, a total of 262 differentially expressed metabolites were identified in ESI+ mode, which constituted the candidate metabolite set for all subsequent cross-omics and biomarker analyses.

[0048] 4. Identification of core candidate metabolites of GDM from maternal serum To identify the metabolites that contributed most strongly to intergroup separation, hierarchical clustering was performed on the differentially expressed metabolite set, and the importance of variables in the OPLS-DA model was assessed. Figure 4 As shown in Figure A, the heatmap of the top 20 metabolites reveals a clear abundance pattern, largely separating samples by group and demonstrating strong inter-group separation ability. The VIP score was then used to quantify the relative contribution of each metabolite to the OPLS-DA model. In ESI+ mode, choline, taurine, creatine, pyruvate, and glycerophosphate choline ranked among the metabolites with the highest VIP scores, see [Figure A]. Figure 4 B. In the ESI-mode, L-glutamine, betaine, D-mannose, L-arginine, and vanillylmandelic acid were identified as the most significant contributors, see [link to ESI-mode]. Figure 4 C. The strong consistency between the clustering patterns in the heatmap and the VIP rankings highlights a coherent set of metabolites that not only exhibit significant inter-group differences but also possess strong discriminative power. Therefore, these are nominated as key candidates for subsequent pathway enrichment analysis and biomarker evaluation.

[0049] 5. Abnormalities in amino acid, lipid, and nucleotide pathways in maternal serum of patients with GDM To elucidate the systemic metabolic disorders associated with GDM, KEGG pathway over-enrichment analysis was performed on identified differentially metabolites. For example... Figure 5 In ESI+ mode, pathways including tyrosine metabolism, secondary metabolite biosynthesis, protein digestion and absorption, and ABC transporters were significantly enriched. Figure 5 In ES-mode, the data further highlighted profound alterations in pyrimidine metabolism, linoleic acid metabolism, arginine biosynthesis, nucleotide metabolism, fatty acid biosynthesis, and arachidonic acid metabolism. Overall, these findings indicate a widespread and coordinated dysregulation of the interconnected networks of amino acid, lipid, and nucleotide metabolism in the serum of pregnant women with GDM. This multi-pathway reprogramming is consistent with the widespread metabolic disturbances specific to GDM and provides a functional framework for subsequent multi-omics integration and prioritization of mechanistic biomarker candidates.

[0050] 6. Cross-omics correlation analysis between maternal metabolites and fetal transcripts / proteins To investigate the relationship between maternal circulating metabolism and fetal transcripts / proteins, a bidirectional O2PLS model was constructed, integrating 262 DMs from maternal serum analysis with the top 100 DEGs from HUVEC transcript analysis or the top 80 DEPs from umbilical cord blood serum analysis. The combined loading revealed coordinated covariation of functionally relevant molecular features along shared potential components (see [link to O2PLS model]). Figure 6Specifically, metabolites including betaine, choline, and taurine clustered with key genes involved in lipid and cholesterol metabolism, while D-mannose and pyruvate occupied adjacent loading spaces with genes involved in carbon metabolism and redox homeostasis. Paired Pearson correlation analysis further identified significant gene-metabolite associations (see [link to analysis]). Figure 6 (C), which includes several strong relationships. Overall, these results depict a tightly coupled system linking maternal metabolism to fetal transcriptional programs, centered on the intertwining of lipid / cholesterol homeostasis and amino acid / carbon metabolism with redox processes, thus providing biologically interpretable clues for subsequent integration with umbilical cord blood proteomics and biomarker prioritization.

[0051] To further analyze the relationship between maternal circulating metabolism and fetal circulating proteins, an integrated protein-metabolite O2PLS model was established using the first 80 DEPs and 262 maternal DMs from umbilical cord blood serum (see [link to model]). Figure 6 B). The combined loads converged into two main functional clusters: a lipid / cholesterol metabolism-lipoprotein transport module, in which betaine, choline, glycerophosphatecholine, and taurine covariate with proteins involved in lipid homeostasis; and an unsaturated fatty acid / redox module, in which metabolites associated with linoleic acid / arachidonic acid metabolism cluster near proteins involved in coagulation-complement cascades and endothelial adhesion. This organization suggests that maternal lipid and redox alterations are associated with coordinated variations in fetal lipoprotein transport, coagulation / inflammatory responses, and endothelial function. Paired Pearson correlation analysis confirmed several significant protein-metabolite pairs (see [link to Pearson correlation analysis]). Figure 6 (D) This integrated analysis reveals a multilayered coupling system between maternal metabolites and fetal circulating proteins, organized around lipid homeostasis and lipoprotein transport, coagulation-complement activation, endothelial adhesion / ECM remodeling, and unsaturated fatty acid / redox processes, thus providing consistent cross-layered evidence for subsequent network integration and biomarker prioritization.

[0052] 7. Identification of maternal metabolite biomarkers based on complementary evidence of network and correlation. Based on cross-omics association networks, 33 candidate metabolites (DMs) were first screened for subsequent biomarker priority ranking. Then, two complementary strategies were employed to rank these candidates. First, a network-based approach used five centrality algorithms, and based on the intersection of the top 10% rankings from each algorithm, D-mannose, betaine, taurine, and pyruvate were identified as the highest-ranked pivot metabolites (see [link to relevant documentation]). Figure 7 A). These metabolites demonstrated excellent diagnostic properties in this cohort and showed significant abundance variations between GDM and control groups (see A). Figure 7 B and Figure 7C). Secondly, an independent statistical correlation method, based on the Pearson correlation between metabolite abundance and group status, highlighted L-glutamine, taurine, pyrrole-2-carboxylic acid, and piperidine, which also achieved high AUC and significant intergroup separation (see [link to study]). Figure 7 (DF). The convergence of the two methods on taurine underscores their robustness as a candidate biomarker.

[0053] Notably, all prioritized metabolites closely corresponded to cross-omics regulatory networks. Key metabolite-gene / protein pairs aligned with the results of the O2PLS co-load. By integrating network centrality, discriminative power, and cross-layer consistency, betaine, taurine, pyruvate, and D-mannose were ultimately identified as a core group of maternal biomarkers for GDM to reflect fetal metabolic dysfunction. In this cohort, these metabolites not only effectively distinguished GDM from controls but also reflected functional alterations in the placental endothelial and fetal side proteomic profiles, supporting their potential for non-invasive metabolic readings of fetal-maternal status.

[0054] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A non-invasive biomarker for detecting fetal intrauterine conditions in patients with gestational diabetes, characterized in that, The biomarker is one or more of betaine, taurine, pyruvate, and D-mannose.

2. The application of a biomarker as described in claim 1 in the preparation of detection products, characterized in that, The testing product is used to assess the intrauterine metabolic health status of fetuses with GDM. The applications include: The concentration of the biomarker was determined based on peripheral blood samples from pregnant women; Based on the concentration of the biomarker, and in conjunction with a fetal status assessment model associated with the biomarker, the likelihood of GDM-related metabolic abnormalities in the pregnant woman and the fetus in the womb is assessed.

3. The application according to claim 2, characterized in that, The fetal status assessment model is associated with the biomarker, and the concentration of the biomarker is used as the input of the fetal status assessment model. The fetal status assessment model outputs risk assessment results for the pregnant woman to be tested having GDM and fetal metabolic status assessment results.

4. The application according to claim 2 or 3, characterized in that, The fetal status assessment model is a multi-omics integrated assessment model built on a fetal-anchored training set; the training set contains multi-omics data from paired samples of GDM pregnant women and healthy control pregnant women. The paired samples include: maternal peripheral blood samples, placental interface tissue or cell samples from the same pregnancy, and fetal cord blood samples; the multi-omics data include at least: metabolomic data from maternal peripheral blood samples, transcriptomic or proteomic data from placental interface tissue or cell samples, and proteomic or metabolomic data from fetal cord blood samples.

5. The application according to claim 4, characterized in that, The construction of the multi-omics integrated evaluation model includes: preprocessing and standardizing the multi-omics data of the training set; using a cross-omics data integration algorithm to mine the association network between maternal metabolomics data, placental interface molecular omics data, and fetal circulation molecular omics data; and based on the association network, screening out a set of maternal metabolic biomarkers and their feature weights that are consistent with the direction of molecular changes on the fetal side and have strong statistical association in the GDM state.

6. The application according to claim 5, characterized in that, The cross-omics data integration algorithm includes a bidirectional orthogonal partial least squares model; the mining of the association network includes calculating the correlation between maternal metabolites and fetal gene or protein expression levels, and identifying metabolite-molecule pairs with consistent orientation and correlation strength exceeding a preset threshold.

7. The application according to claim 6, characterized in that, The association network contains at least one of the following metabolite-molecule pairs: betaine with SCARB1 positive correlation pair, taurine with GPX8 negative correlation pair, pyruvate with GALK2 positive correlation pair, and D-mannose with AKR1B1 positive correlation pair.

8. The application according to claim 7, characterized in that, The assessment results output by the fetal status assessment model are presented in the form of GDM risk probability, risk level, and scores indicating the degree of abnormalities in fetal lipid metabolism, oxidative stress, and glucose metabolism.

9. The application according to claim 5, characterized in that, Placental interface tissue or cell samples include human umbilical vein endothelial cells, placental villus tissue, or trophoblast cells; placental interface molecular omics data are transcriptomic data from HUVECs. The fetal circulation molecular omics data are proteomic data from umbilical cord blood.

Images