Machine Learning-Driven Early Prediction of Spontaneous Preterm Birth Subtypes from Second-Trimester Plasma Metabolomic

Machine Learning-Driven Early Prediction of Spontaneous Preterm Birth Subtypes from Second-Trimester Plasma Metabolomic | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Machine Learning-Driven Early Prediction of Spontaneous Preterm Birth Subtypes from Second-Trimester Plasma Metabolomic Shen Yongmei, Wu Dan, Wang Hefei, Liu Tianxiang, Cao Jiasong, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8487634/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 14 You are reading this latest preprint version Abstract Background Preterm birth is a major cause of neonatal morbidity and mortality, with spontaneous preterm birth (sPTB) comprising preterm premature rupture of membranes (pPROM) and spontaneous preterm labor (sPL). Reliable early prediction remains challenging, particularly in distinguishing sPTB subtypes. Metabolomics offers a promising approach for identifying predictive biomarkers. Methods A single-center case-control study was conducted using archived maternal plasma samples. Participants included 70 pregnant women (30 term controls, 20 pPROM, 20 sPL) at 14–20 weeks’ gestation. Non-targeted metabolomic profiling was performed via liquid chromatography-mass spectrometry (LC-MS). Metabolite screening was carried out using LASSO regression, and pathway enrichment analysis was conducted. Machine learning models (logistic regression) were developed and validated. Statistical analyses included PLS-DA, ROC curves, Pearson correlation, and risk stratification. Results Nine metabolites associated with inflammatory activation, oxidative stress, and placental dysfunction were identified. LASSO models achieved high predictive accuracy (AUCs: 0.984 for controls, 0.964 for pPROM, 0.995 for sPL). Creatinine and LysoPC(P-16:0) was positively correlated with gestational age at blood sampling (R = 0.27/0.23), while phosphatidylcholine was negatively correlated with maternal age (R=-0.31). Gestational age at delivery was negatively correlated with BMI (R = − 0.51). High-risk stratification showed a decreasing probability of preterm birth with increasing gestation, while low-risk stratification remained stable. Conclusions Second-trimester plasma metabolomics combined with machine learning could effectively predict sPTB and distinguish its subtypes. These findings support the potential for early risk stratification and personalized intervention, though multicenter validation is needed for clinical translation. Premature birth Plasma metabolomics Premature premature rupture of membranes Spontaneous preterm birth Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Preterm birth, defined as birth before 37 weeks of gestation, is a leading cause of neonatal morbidity and mortality worldwide( 1 ). The current preterm birth rate is 10.6%, and it can be categorized into two main types: iatrogenic preterm birth, which is induced by medical interventions, and spontaneous preterm birth(sPTB), which includes preterm premature rupture of membranes (pPROM) and preterm labor with intact membranes (sPL)( 2 ). Despite efforts to predict preterm birth using various methods such as ultrasound measurements, combined indicators, lipid biomarkers, vaginal flora levels, and psychological factors like anxiety and happiness, there is still no reliable predictive model available. Further research is needed, particularly to distinguish between sPL and pPROM. A multitude of physiological changes and metabolic adaptations occur weekly during pregnancy. Metabolomics is a method that can comprehensively analyze metabolites in organisms, offering a new perspective for early disease diagnosis. In 2020, Mads Melbye et al. developed a metabolic clock that accurately predicts gestational age, aligning closely with early pregnancy ultrasound, the clinical gold standard( 3 ). Blood samples are the preferred type for predicting preterm birth due to the limitations of placenta and amniotic fluid sampling as well as the instability of urine. Although studies have investigated alterations in serum metabolites in preterm and normal pregnant women, it is important to note that these studies typically focus on gestational age after 24 weeks or between 29 + 0 weeks and 36 + 5 weeks( 4 , 5 ). The survival rate of premature infants at the earliest gestational age has increased significantly in developed countries, extending their viability limit to 22 to 23 weeks of gestation. However, their risk of death and morbidity remains higher( 6 ). Therefore, predicting premature birth at an earlier gestational age is crucial for enabling early intervention and treatment for these infants. Additionally, distinguishing between sPL and pPROM is important for implementing customized interventions to improve outcomes for both conditions. We employed liquid chromatography-mass spectrometry to examine the non-targeted metabolomics of maternal blood samples during the second trimester (14–20 weeks). This research revealed variations in metabolites among various types of preterm birth and used a logistic regression model to predict these types during the second trimester, providing methods and markers for earlier prediction of preterm birth. Material and Methods 1. Clinical data From October 2022 to October 2024, peripheral blood samples were collected from 3,000 pregnant women at 14-20 weeks of gestation and stored at −80°C. The inclusion criteria were singleton pregnancies conceived naturally, absence of underlying diseases, and no family history of genetic disorders. Preliminary exclusion criteria included fetal structural or chromosomal abnormalities, and pregnancy complications such as gestational diabetes and preeclampsia. Participants were categorized into three groups: (1) a normal control group comprising women who delivered at full term without complications; (2) a pPROM group involving women with spontaneous membrane rupture between 28 and 37 weeks of gestation; and (3) an sPL group consisting of women with intact fetal membranes spontaneous preterm labor between 28 and 37 weeks of gestation. The received samples were randomly divided into two distinct sets for metabolomic testing. The sets comprised 15 controls, 10 pPROM cases, and 10 sPL cases each. Each set was designated as either a test set or a validation set following established stratification protocols. All research procedures followed the ethical guidelines of the Declaration of Helsinki and were approved by the Research Ethics Committee of Tianjin Central Hospital of Obstetrics and Gynecology (Approval No. 2022KY068). Written consent was obtained from all participants. The selection process for the samples is outlined in Figure 1. 2. Sample Analysis and Data Processing For each plasma sample, 100 μL was mixed with 400 μL of 80% methanol aqueous solution. After placing the mixture in an ice bath for 5 minutes, it was centrifuged at 15,000 g for 20 minutes at 4°C. The supernatant was collected and diluted with MS-grade water to adjust the methanol content to 53%, followed by another centrifugation at 15,000 g for 20 minutes at 4°C. The resulting supernatant was then subjected to liquid chromatography-mass spectrometry (LC-MS) analysis using a Dionex Ultimate 3000 ultra-high-performance liquid chromatography system(7) . Chromatographic separation was performed using a Thermo Syncronis C18 column (2.1 mm × 100 mm, 1.7 μm). The mobile phase A consisted of water containing 0.1% formic acid (v/v) and 2 mM ammonium formate, while mobile phase B consisted of acetonitrile. The gradient elution conditions were as follows: 0–1 min, 95% A; 1–5 min, 95%–40% A; 5–8 min, 40%–0% A; 8–11 min, 0% A; 11–14 min, 0%–40% A; and 15–18 min, 95% A. Quality control (QC) samples were prepared by pooling equal volumes from each plasma sample. Samples were analyzed in random order, with one QC sample inserted every six experimental samples to monitor system stability and ensure the reliability of the experimental data. Mass spectrometry was performed using an electrospray ionization (ESI) source, operating in both positive and negative ion modes. The ESI voltage was set to 2.8 kV, with sheath gas flow at 35 arb, auxiliary gas flow at 10 arb, and a capillary temperature of 320°C. The full MS scan resolution was set to 70,000, with a scan range of 70–1050 m/z. The data-dependent secondary scan (full MS/dd-MS) had a resolution of 17,500, with stepped normalized collision energy (NCE) values of 20, 40, and 60 V. Raw data files (.raw) were imported into TraceFinder 3.2.0 software for library searching. Each metabolite was filtered based on retention time, mass-to-charge ratio, and other parameters, with a retention time deviation of 0.2 min and a mass accuracy threshold of 5 ppm used for peak alignment across different samples. Peak areas were quantified, and spectral matching was performed against the mzCloud database (https://www.mzcloud.org/) and a locally constructed database. The quantitative results were then normalized, providing identification and relative quantification of metabolites in plasma samples. Data processing was conducted using the Linux operating system, with the R (4.4.1) and Python (3.13.2) software environments. 3. Bioinformatics Analysis Metabolites were annotated using KEGG 3.13.2, HMDB, and LIPIDMaps databases. Data preprocessing was conducted using metaX software, followed by Partial Least Squares Discriminant Analysis (PLS-DA)(8). Differential metabolites were identified based on Variable Importance in the Projection (VIP) > 1, P < 0.05, and fold change (FC) ≥ 2 or ≤ 0.5. Permutation tests (100 iterations) were used to assess model overfitting by comparing R²Y and Q²Y values with real models. Volcano and KEGG enrichment bubble plots were created using ggplot2. Metabolites were selected with thresholds of VIP > 1, FC > 1.5, and P < 0.05. Heatmaps were generated with Pheatmap, and data were normalized with z-scores. Unsupervised clustering of the 426 most variable metabolites was conducted using K-means and Euclidean distance. To identify most variable metabolites, we chose the metabolites in 426 Least absolute shrinkage and selection operator (LASSO) regression was employed for metabolites selection. The samples with selected nine metabolites were randomly split into training and test sets based on the results of two metabolic panel tests. The logistic regression model was built with glm() with binomial logit option and validated on the test set. Box plots were created using geom_boxplot. Pearson correlations were analyzed with cor (), with significance ( P < 0.05) determined using cor.mtest (). Correlation plots were generated with corrplot. 4 Statistical analyses We conducted statistical analysis on the clinical and demographic characteristics of all plasma samples. After normality and homogeneity of variance tests, we employed the Kruskal-Wallis H test on Age, BMI, Delivery gestational age, and Birthweight. For the Blood sampling gestational age, we conducted a one-way ANOVA with Tukey's post hoc test. Additionally, Fisher's exact test was used for the gender of newborns. The baseline data obtained were presented in Table 1. Results 1. Clinical Characteristics The sample collection process for the Normal, pPROM, and sPL groups is detailed in Figure 1, with baseline data summarized in Table 1. No statistically significant differences were observed in baseline characteristics across the three groups. The mean maternal age was approximately 30 years. Significant differences in BMI and birth weight were noted between the pPROM and sPL groups compared to the term group. Plasma samples were predominantly collected between 12 and 13 weeks of gestation, with a consistent newborn sex ratio across groups. Gestational age at delivery and newborn weight in the term and preterm groups were consistent with the expected clinical characteristics of these conditions. Table 1. Clinical and demographic information on pregnancies included in the study Control n=30 pPROM n=20 sPL n=20 p value Age，years a 28.61（27.05，30.16） 30.47（28.79，32.16） 31.65（28.99，34.31） 0.18 BMI,kg/m 2 a 20.93（20.29，21.57） 22.49（20.98，24.00） * 23.73（22.65，24.81） # ＜0.001 Blood sampling gestational age,weeks b 12.58±0.50 12.83±0.61 12.41±0.39 0.035 Delivery gestational age,weeks a 39.83（39.46，40.19） 34.56（33.69，35.43） * 34.61（33.44，35.77） # ＜0.001 Male infant (n,%) c 15 (50) 11 (55) 13(65) 0.621 Birthweight, Kg a 3.37（3.24，3.50） 2354.74（2110.81，2598.66） * 2368.50（2082.40，2654.60） # ＜0.001 a : Data presented as ‘mean(95% confidence interval for difference)’ and analyzed by kruskal-Wallis H test b : Data presented as ‘mean±standard’ deviation and analyzed by one-way ANOVA with Tukey’s posthoc test. c : Data presented as ‘n (%)’ and analyzed by Fisher’s exact test * : Normal vs. pPROM, p ＜0.05 # : Normal vs. sPL, p ＜0.05 2. Global Distribution of Maternal Plasma Metabolites in pPROM and sPL To assess the overall distribution of metabolites and uncover group-specific differences, Partial Least Squares Discriminant Analysis (PLS-DA) was conducted on maternal plasma samples from the pPROM and sPL groups, in comparison with the normal group. The analysis revealed significant differences in metabolite profiles between the preterm groups and the normal ( Figure 2A ). Permutation testing was further applied to evaluate the explanatory and predictive power of the PLS-DA models. The results showed no evidence of overfitting, with R² values between 0.7 and 0.8, indicating strong model performance suitable for further screening and analysis ( Figure 2B ). A volcano plot of the 745 identified metabolites revealed 77 metabolites with significant differences in the pPROM group compared to normals, with 60 upregulated and 17 downregulated ( Figure 2C ). In the sPL group, 179 metabolites showed significant differences, 13 of which were upregulated, while the remainder were downregulated. A heatmap based on the differentially expressed metabolites was generated, visually representing the abundance variations across groups, providing an intuitive understanding of metabolite distribution differences ( Figure 2D ). KEGG pathway analysis of the identified differential metabolites showed significant enrichment in the arachidonic acid metabolism pathway within the pPROM group, particularly involving 6-Keto-prostaglandin F1α and prostaglandin E2. In the sPL group, significant enrichment was observed in the biosynthesis of secondary metabolites pathway. Key metabolites included tyramine, xanthine, 9-oxononanoic acid, biliverdin, caffeine, carvone, cis-aconitate, citrulline, coumarin, eucalyptol, ferulic acid, genistein, geranyl diphosphate, histamine, L-tryptophan, nicotinic acid, pantothenic acid, pilocarpine, pipecolic acid, porphobilinogen, and pulegone. Additionally, the monoterpenoid biosynthesis pathway was significantly enriched, involving metabolites such as carvone, eucalyptol, geranyl diphosphate, and pulegone ( Figure 2E, F ). 3. Development of Predictive Models for Different Types of Preterm Birth To identify metabolomic biomarkers for pPROM and sPL, metabolites from the pPROM and sPL groups were divided into training and validation sets based on the results of two metabolic panel tests. In the training set, 9 most important metabolites were identified using Least absolute shrinkage and selection operator (LASSO) regression, and logistic regression was used to build predictive models for the pPROM, sPL, and normal groups. These metabolites were: All-Trans-13,14-Dihydroretinol, Alpha-ketoisocaproic acid, Creatin/Creatinine, Kynurenic acid, LysoPC(P-16:0), LysoPE(0:0/20:4), Phosphocholine, Skatole. A boxplot revealed significant differences in metabolite concentrations between the control, pPROM, and sPL groups. Notably, the sPL group exhibited distinct metabolite profiles compared to both controls and pPROM ( Figure 3A ). The ROC curve illustrating the predictive performance based on these 9 metabolites is presented in Figure 4B. Receiver Operating Characteristic (ROC) curves assessed diagnostic accuracy, with the control group showing excellent performance (AUC = 0.984, 95% CI: 0.953–1). The pPROM group demonstrated moderate diagnostic ability (AUC = 0.964, 95% CI: 0.909–1), while the sPL group exhibited exceptional accuracy (AUC = 0.995, 95% CI: 0.982–1), indicating the high diagnostic potential of these metabolites, particularly for sPL. Kaplan-Meier survival curves were used to evaluate risk stratification, revealing significant differences in risk probabilities for both pPROM (p = 0.0076) and sPL (p = 0.0081) ( Figure 3B ). 4. Maternal Metabolic Determinants of Gestational Age Progression and Preterm Delivery Risk Subsequent analysis revealed correlations between nine metabolites and maternal age, BMI, blood sampling gestational age, and delivery gestational age. Creatin or creatinine showed significant correlations with other eight metabolites. Only phosphatoline exhibited a negative correlation with maternal age (R=-0.31), whereas Creatinine and LysoPC (P-16-0) were positively correlated with blood sampling gestational age (R=0.27, 0.23). Alpha ketoisocaproic acid, Creatin/Creatinine, Kynurenic acid, and Phosphatoline were correlated with BMI, while All-Trans-13,14-Dihydroretinol, Alpha-ketoisocaproic acid, Kynurenic acid, LysoPE (0:0/20:4), and Phosphocholine were correlated with delivery gestational age. Furthermore, a correlation was observed between BMI and delivery gestational age(R=-0.51) ( Figure 4A ). Subsequent analysis on metabolites and the risk of delivery gestational age indicated that high-risk groups had a higher likelihood of preterm birth, which decreased as gestational age progressed, while low-risk groups consistently maintained a lower risk ( Figure 4 B, C ). Discussion This study identified a panel of 9 plasma metabolites-All-Trans-13,14-Dihydroretinol, α-ketoisocaproic acid, Creatin/Creatinine, Kynurenic acid, LysoPC (P-16:0), LysoPE (0:0/20:4), Phosphocholine, and Skatole-that effectively distinguish between spontaneous preterm labor (sPL) and preterm premature rupture of membranes (pPROM) as early as 12–13 weeks of gestation. PLS-DA models confirmed distinct metabolite profiles between preterm and term groups, without overfitting. Pathway analysis revealed significant enrichment of arachidonic acid metabolism in pPROM, notably involving prostaglandin E2 and 6-keto-prostaglandin F1α, while the sPL group showed enrichment in secondary metabolite biosynthesis. A logistic regression model based on these metabolites demonstrated high predictive accuracy, with AUCs of 0.995 for sPL, 0.964 for pPROM, and 0.984 for controls. Several metabolites correlated with BMI and gestational age at delivery, and risk stratification analysis confirmed that higher metabolic risk scores were associated with earlier delivery. Preterm birth is the leading cause of neonatal morbidity and mortality worldwide, but there is a lack of definitive biomarkers to accurately predict the risk of preterm birth, especially in distinguishing between sPL and pPROM. Metabolomics provides a new approach for early disease detection. Previous studies have examined changes in serum metabolites in premature and normal pregnant women after 24 weeks of gestation or between 29 and 36 + 5 weeks, but have not effectively distinguished between sPL and pPROM ( 4 , 5 ). Currently, there is limited research on using ultra-performance liquid chromatography-mass spectrometry (UPLC-MS) to identify alterations in blood metabolomics during preterm labor, particularly between 14 and 10 weeks of gestation. Blood samples are the preferred type of sample for predicting preterm birth, as sampling placenta and amniotic fluid is limited and urine is unstable. A study comparing metabolites in early preterm and term placentas found higher levels of dysregulated metabolites in early preterm placentas. However, there was no significant difference in the placental metabolome between spontaneous preterm and term births, suggesting that placentas of different gestational ages may produce different metabolites ( 9 ). Significant changes in metabolites such as amino acids, fatty acids, lipids, hormones, and bile acids have been reported in the peripheral blood of premature patients after 24 weeks, with fatty acids being the most significant ( 4 ). Additionally, a study found that lipid levels in the peripheral blood of preterm births after 29 weeks were increased ( 5 ). The studies mentioned above suggest that lipid metabolism may play a significant role in the occurrence and progression of preterm birth. However, it is unclear whether various types of spontaneous preterm birth are associated with lipid metabolism. Our results showed that the differential metabolites of different types of premature birth were not enriched in pathways related to lipid metabolism. The main metabolic pathways involved in sPL and pPROM were biosynthesis of secondary metabolites pathway and arachidonic acid metabolism pathway respectively. Specifically, the pPROM group shows alterations in the arachidonic acid metabolism biosynthetic pathway, with metabolites such as 6-keto-prostaglandin F1a and prostaglandin E2 being involved. Prostaglandin E2 is known to be used for inducing labor in cases of premature rupture of membranes before delivery ( 10 ). Therefore, abnormal levels of prostaglandin E2 or its metabolites may contribute to the development of pPROM. The nine metabolites identified in this study -All-Trans-13,14-Dihydroretinol, α-ketoisocaproic acid (KIC), creatinine (via SUA/SCr ratio), kynurenic acid, LysoPC, LysoPE, phosphocholine, and skatole—collectively underscore the multifactorial metabolic dysregulation underlying preterm birth (PTB). All-Trans-13,14-Dihydroretinol, a vitamin A derivative, exhibits elevated mid-trimester levels in PTB cases, suggesting its role in placental maldevelopment due to disrupted angiogenesis and retinoid signaling ( 11 , 12 ). Similarly, KIC, a leucine catabolite, demonstrates reduced levels in maternal blood, implicating energy metabolism impairment via inhibition of insulin-stimulated glucose transport in myotubes ( 13 , 14 ). The SUA/SCr (serum uric acid/serum creatine) ratio, reflecting maternal renal function and fetal kidney stress, emerges as a predictive biomarker for PTB, with elevated creatinine levels related to acute neonatal kidney injury. Creatine or creatinine exhibited significant correlations with eight other metabolites in this study, underscoring the relevance of these compounds in preterm birth( 15 – 17 ). Diminished kynurenine levels in preterm placental and fetal compartments correlate with neurodevelopmental risks, likely mediated by kynurenic acid imbalance, thus emphasizing immune-metabolic crosstalk within the kynurenine pathway( 18 , 19 ). Lysophospholipids (LysoPC/LysoPE), elevated in adverse pregnancies, drive inflammation and trophoblast dysfunction-LysoPE (16:0) promotes aberrant trophoblast invasion, while LysoPC amplifies prostaglandin-mediated inflammatory cascades ( 20 – 22 ). Phosphocholine depletion in PTB neonates, associated with membrane instability and apoptosis, underscores its role in placental lipid signaling ( 23 , 24 ). Finally, skatole, though lacking direct PTB evidence, may exacerbate systemic inflammation via TLR activation and gut microbiota dysbiosis, aligning with emerging links between maternal microbiome shifts and preterm labor ( 25 , 26 ). Phosphatoline was the only metabolite showing a negative correlation with maternal age in our study. Creatinine and LysoPC (P-16-0) were positively correlated with gestational age at blood sampling. Maternal age and gestational age at blood collection had minimal impact on metabolite prediction. Five of the nine metabolites analyzed were associated with BMI and gestational age, suggesting a potential link between BMI and premature birth, warranting further investigation. Certain metabolites were related to gestational age, with high-risk groups showing an increased risk of preterm birth, indicating the potential use of metabolites for predicting delivery time. Premature birth is characterized by disruptions in metabolic networks, inflammation, oxidative stress, and placental dysfunction. Understanding the alterations and interactions of these markers can aid in early prediction and intervention in premature birth. Although this study employed ultra-performance liquid chromatography-mass spectrometry (UPLC-MS) for high-sensitivity metabolite detection, collected plasma samples during the second trimester (14–20 weeks), identified potential biomarkers in the peripheral blood of pregnant women with preterm birth, described metabolite changes across different subtypes of preterm birth, and developed a multiclass prediction model capable of distinguishing between spontaneous preterm labor (sPL) and preterm premature rupture of membranes (pPROM), certain limitations remain. The findings are derived from a single-center cohort with a limited sample size, which restricts the generalizability of the results. Although nine promising metabolites associated with preterm birth were identified, external validation in larger, multicenter populations is essential to confirm their predictive value and support future clinical translation. Furthermore, longitudinal profiling of metabolite dynamics throughout gestation is required to refine the predictive model and enhance its clinical applicability. Conclusions This study identified key metabolic pathways, including arachidonic acid metabolism in pPROM and secondary metabolite biosynthesis in sPL, revealing distinct pathophysiological mechanisms underlying these subtypes. The identified metabolites are involved in inflammatory processes, oxidative stress, and placental dysfunction, reflecting the multifactorial pathophysiology of preterm birth. The study also discovered a panel of nine second-trimester plasma metabolites that effectively predict spontaneous preterm birth (sPTB) and differentiate between preterm premature rupture of membranes (pPROM) and spontaneous preterm labor (sPL) as early as 14–20 weeks of gestation. Using machine learning approaches, a predictive model with high discriminative accuracy was developed. In contrast to previous studies focused on late gestation, this work provides novel metabolic insights at an earlier stage of pregnancy. Future research should further elucidate the mechanistic interactions among maternal-fetal immune responses, gut microbiota-derived signals such as skatole, and metabolic dysregulation, while promoting clinical translation through standardized assay development and longitudinal profiling to improve predictive reliability and support personalized intervention strategies. Declarations Ethics approval and consent to participate The studies involving human participants were reviewed and approved by Human Research Ethics Committee of Tianjin Central Hospital of Obstetrics and Gynecology(NO.2022KY068). All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. All participants signed a written informed consent form prior to their participation in this study. Consent for publication Consent for publication were obtained from that person. Availability of data and materials The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Competing interests All authors declare that they have no conflict of interest. Funding This work was supported by Noncommunicable Chronic Diseases-National Science and Technology Major Project (2023ZD0517600), Tianjin Health Research Project (TJWJ2025MS031), Tianjin Key Medical Discipline Construction Project (TJYXZDXK-3-029C). Authors' contributions All authors had full access to all the data in the study and accept responsibility to submit for publication. These authors meet the four criteria for authorship as specified in the ICMJE recommendations: Yongmei Shen contributed to the conceptualisation, funding acquisition, investigation, methodology, project administration, formal analysis and writing-original draft and writing- review and editing. Dan Wu contributed to the investigation, methodology, project administration, data curation, formal analysis. Hefei Wang contributed to the methodology, project administration, data curation writing-original draft. Tianxiang Liu contributed to the methodology,data curation. Yaqi Li, Maolin Nie, Rongxin Wei contributed to data curation, sample collection and processing. Jiaosong Cao and Qimei Lin contributed to the writing-review and editing. Junli He contributed to sample collection. Yongjun Piao contributed to methodology, formal analysis and writing-original draft. Ying Chang contributed to the conceptualisation, funding acquisition, investigation,writing- review. Acknowledgements We wish to acknowledge and thank all study participants. References Ohuma EO, Moller AB, Bradley E, Chakwera S, Hussain-Alkhateeb L, Lewin A, et al. National, regional, and global estimates of preterm birth in 2020, with trends from 2010: a systematic analysis. Lancet. 2023;402(10409):1261-71. Prediction and Prevention of Spontaneous Preterm Birth: ACOG Practice Bulletin, Number 234. Obstet Gynecol. 2021;138(2):e65-e90. Liang L, Rasmussen MH, Piening B, Shen X, Chen S, Röst H, et al. Metabolic Dynamics and Prediction of Gestational Age and Time to Delivery in Pregnant Women. Cell. 2020;181(7):1680-92.e15. Lizewska B, Teul J, Kuc P, Lemancewicz A, Charkiewicz K, Goscik J, et al. Maternal Plasma Metabolomic Profiles in Spontaneous Preterm Birth: Preliminary Results. Mediators Inflamm. 2018;2018:9362820. Virgiliou C, Gika HG, Witting M, Bletsou AA, Athanasiadis A, Zafrakas M, et al. Amniotic Fluid and Maternal Serum Metabolic Signatures in the Second Trimester Associated with Preterm Delivery. J Proteome Res. 2017;16(2):898-910. Barfield WD. Public Health Implications of Very Preterm Birth. Clin Perinatol. 2018;45(3):565-77. Want EJ, Masson P, Michopoulos F, Wilson ID, Theodoridis G, Plumb RS, et al. Global metabolic profiling of animal and human tissues via UPLC-MS. Nat Protoc. 2013;8(1):17-32. Wen B, Mei Z, Zeng C, Liu S. metaX: a flexible and comprehensive software for processing metabolomics data. BMC Bioinformatics. 2017;18(1):183. Cifkova E, Karahoda R, Stranik J, Abad C, Kacerovsky M, Lisa M, Staud F. Metabolomic analysis of the human placenta reveals perturbations in amino acids, purine metabolites, and small organic acids in spontaneous preterm birth. Excli j. 2024;23:264-82. Zhang Y, Wang J, Yu Y, Xie C, Xiao M, Ren L. Misoprostol versus prostaglandin E2 gel for labor induction in premature rupture of membranes after 34 weeks of pregnancy. Int J Gynaecol Obstet. 2015;130(3):214-8. Naik VD, Lee J, Wu G, Washburn S, Ramadoss J. Effects of nutrition and gestational alcohol consumption on fetal growth and development. Nutr Rev. 2022;80(6):1568-79. Ma G, Chen Y, Liu X, Gao Y, Deavila JM, Zhu MJ, Du M. Vitamin a supplementation during pregnancy in shaping child growth outcomes: A meta-analysis. Crit Rev Food Sci Nutr. 2023;63(33):12240-55. Vlaardingerbroek H, Roelants JA, Rook D, Dorst K, Schierbeek H, Vermes A, et al. Adaptive regulation of amino acid metabolism on early parenteral lipid and high-dose amino acid administration in VLBW infants - a randomized, controlled trial. Clin Nutr. 2014;33(6):982-90. Mann G, Adegoke OAJ. Effects of ketoisocaproic acid and inflammation on glucose transport in muscle cells. Physiol Rep. 2021;9(1):e14673. Zhang E, Su S, Gao S, Zhang Y, Wang J, Liu J, et al. Elevated serum uric acid to creatinine ratio is associated with adverse pregnancy outcomes: a prospective birth cohort study. Int J Med Sci. 2024;21(9):1612-21. Wu Y, Wang H, Pei J, Jiang X, Tang J. Acute kidney injury in premature and low birth weight neonates: a systematic review and meta-analysis. Pediatr Nephrol. 2022;37(2):275-87. Go H, Momoi N, Kashiwabara N, Haneda K, Chishiki M, Imamura T, et al. Neonatal and maternal serum creatinine levels during the early postnatal period in preterm and term infants. PLoS One. 2018;13(5):e0196721. van Zundert SK, Broekhuizen M, Smit AJ, van Rossem L, Mirzaian M, Willemsen SP, et al. The Role of the Kynurenine Pathway in the (Patho) physiology of Maternal Pregnancy and Fetal Outcomes: A Systematic Review. Int J Tryptophan Res. 2022;15:11786469221135545. Morrens M. The role of the kynurenine pathway in cognitive function in brain Disorders: Insights and challenges. Brain Behav Immun. 2025;127:110-1. Tsutsumi T, Yasuda K, Neya M, Okada H, Tokumura A. Lysophosphatidic acid production from lysophosphatidylcholine by lysophospholipase D activity of autotaxin in plasma of women with normal and adverse pregnancies. Prostaglandins Other Lipid Mediat. 2022;163:106670. Zhang Z, He P, Chen D, Tan Y, Chen A, Bian Z, Chen T. Active metabolomics identify potential functional metabolites for preeclampsia prevention. Clin Chim Acta. 2024;560:119717. Zhong W, Li Y, Zhong H, Cheng Y, Chen Q, Zhao X, et al. Exploring the mechanism of anti-chronic heart failure effect of qiweiqiangxin І granules based on metabolomics. Front Pharmacol. 2023;14:1111007. Bernhard W, Poets CF, Franz AR. Choline and choline-related nutrients in regular and preterm infant growth. Eur J Nutr. 2019;58(3):931-45. Dobrynina EA, Zykova VA, Surovtsev NV. In-plane and out-of-plane gigahertz sound velocities of saturated and unsaturated phospholipid bilayers from cryogenic to room temperatures. Chem Phys Lipids. 2023;256:105335. Georges HM, Norwitz ER, Abrahams VM. Predictors of Inflammation-Mediated Preterm Birth. Physiology (Bethesda). 2025;40(1):0. Lu X, Shi Z, Jiang L, Zhang S. Maternal gut microbiota in the health of mothers and offspring: from the perspective of immunology. Front Immunol. 2024;15:1362784. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 07 May, 2026 Reviews received at journal 20 Feb, 2026 Reviews received at journal 10 Feb, 2026 Reviewers agreed at journal 10 Feb, 2026 Reviewers agreed at journal 09 Feb, 2026 Reviews received at journal 06 Feb, 2026 Reviews received at journal 04 Feb, 2026 Reviewers agreed at journal 04 Feb, 2026 Reviewers agreed at journal 28 Jan, 2026 Reviewers invited by journal 28 Jan, 2026 Editor invited by journal 01 Jan, 2026 Editor assigned by journal 01 Jan, 2026 Submission checks completed at journal 01 Jan, 2026 First submitted to journal 31 Dec, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8487634","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":581976018,"identity":"4c0fe88e-3885-431b-a4b7-9eec59d0da55","order_by":0,"name":"Shen Yongmei","email":"","orcid":"","institution":"Tianjin Central Hospital of Gynecology Obstetrics","correspondingAuthor":false,"prefix":"","firstName":"Shen","middleName":"","lastName":"Yongmei","suffix":""},{"id":581976019,"identity":"1bacc47a-5340-4f17-84b8-c0e2cee048fe","order_by":1,"name":"Wu Dan","email":"","orcid":"","institution":"Tianjin Central Hospital of Gynecology Obstetrics","correspondingAuthor":false,"prefix":"","firstName":"Wu","middleName":"","lastName":"Dan","suffix":""},{"id":581976020,"identity":"844f317e-ce24-4887-af56-b8f536d79652","order_by":2,"name":"Wang Hefei","email":"","orcid":"","institution":"Tianjin Medical University","correspondingAuthor":false,"prefix":"","firstName":"Wang","middleName":"","lastName":"Hefei","suffix":""},{"id":581976021,"identity":"b6eaf317-531e-4da9-bb51-7f66f90014ff","order_by":3,"name":"Liu Tianxiang","email":"","orcid":"","institution":"Nankai University","correspondingAuthor":false,"prefix":"","firstName":"Liu","middleName":"","lastName":"Tianxiang","suffix":""},{"id":581976022,"identity":"b8f90231-5287-4153-957b-c0f8647ffcc6","order_by":4,"name":"Cao Jiasong","email":"","orcid":"","institution":"Tianjin Central Hospital of Gynecology Obstetrics","correspondingAuthor":false,"prefix":"","firstName":"Cao","middleName":"","lastName":"Jiasong","suffix":""},{"id":581976023,"identity":"8105401e-aec9-42ba-89a7-de2e4e4852d9","order_by":5,"name":"Lin Qimei","email":"","orcid":"","institution":"Tianjin Central Hospital of Gynecology Obstetrics","correspondingAuthor":false,"prefix":"","firstName":"Lin","middleName":"","lastName":"Qimei","suffix":""},{"id":581976027,"identity":"afaab166-aaa8-415f-8507-d93f7b8e42c0","order_by":6,"name":"Li Yaqi","email":"","orcid":"","institution":"Tianjin Medical University","correspondingAuthor":false,"prefix":"","firstName":"Li","middleName":"","lastName":"Yaqi","suffix":""},{"id":581976029,"identity":"54b656b6-bdbc-48e1-917c-b23f356138bf","order_by":7,"name":"Nie Maolin","email":"","orcid":"","institution":"Tianjin Medical University","correspondingAuthor":false,"prefix":"","firstName":"Nie","middleName":"","lastName":"Maolin","suffix":""},{"id":581976032,"identity":"ab23ef7d-fd60-4cae-b3f4-70b73cfc44ee","order_by":8,"name":"Wei Rongxin","email":"","orcid":"","institution":"Tianjin Medical University","correspondingAuthor":false,"prefix":"","firstName":"Wei","middleName":"","lastName":"Rongxin","suffix":""},{"id":581976033,"identity":"af87b8bd-e42c-4f39-9839-ab419aad7517","order_by":9,"name":"He Junli","email":"","orcid":"","institution":"Tianjin Medical University","correspondingAuthor":false,"prefix":"","firstName":"He","middleName":"","lastName":"Junli","suffix":""},{"id":581976034,"identity":"58d97b1c-7ca6-483c-a145-41f4146dc64f","order_by":10,"name":"Piao Yongjun","email":"","orcid":"","institution":"Nankai University","correspondingAuthor":false,"prefix":"","firstName":"Piao","middleName":"","lastName":"Yongjun","suffix":""},{"id":581976036,"identity":"7d904982-0723-4289-b321-3c6f2ff3bacb","order_by":11,"name":"Chang Ying","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA+0lEQVRIiWNgGAWjYDACCTh5IP1HQgUDDwlaGA88kPhwhngtQMB88IHkzDYi3MU/u/nZw69tFnnybocTjHnn1cmYsx9g/PAxB48ld46ZG8u2SRQbnjmWkMy77TCPZU8Cs+TMbbi1GEgkmElLtkkkbpxxJuEw77YDPAYHEtiYefFqSf8G0TL//cdm3jl1PAbnHxDSkmMm+RGoZT7DgWTGmQ3MPAY3CNgicSOnTJrhnETiBoYDaQwfjh0GannYjNcv/DPSt0n+KKtLnN8A1JJQU2dvcD754IePeLSAADMvG9CFB+B8xgb86kFKfvxhYJAnrG4UjIJRMApGKgAAw/9W4xigjygAAAAASUVORK5CYII=","orcid":"","institution":"Tianjin Central Hospital of Gynecology Obstetrics","correspondingAuthor":true,"prefix":"","firstName":"Chang","middleName":"","lastName":"Ying","suffix":""}],"badges":[],"createdAt":"2025-12-31 08:23:39","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8487634/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8487634/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":101517089,"identity":"715f8b13-dab7-4129-9d13-023bfc8a0d56","added_by":"auto","created_at":"2026-01-30 16:22:56","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":148590,"visible":true,"origin":"","legend":"\u003cp\u003eSample collection process\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8487634/v1/a6a0b8b25370b301d4ceade6.png"},{"id":101517091,"identity":"8b1790f7-5ef7-4bbf-8c85-c5032d870303","added_by":"auto","created_at":"2026-01-30 16:22:57","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":562090,"visible":true,"origin":"","legend":"\u003cp\u003ePlasma metabolite characteristics of pPROM and sPL. A. PLS-DA diagram; B. Cross-validation diagram; C. Volcano diagram; D. Cluster diagram; E,F. KEGG analysis of differential metabolites inpPROM and sPL.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8487634/v1/39e351fe7f8be86918e85a25.png"},{"id":101517090,"identity":"252e57fc-74e4-4c7c-9669-765e871dce05","added_by":"auto","created_at":"2026-01-30 16:22:56","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":132940,"visible":true,"origin":"","legend":"\u003cp\u003epPROM and sPL prediction model A. Box plot of the nine selected metabolites in normal, pPROM, and sPL; B. ROC curve of the prediction model;\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8487634/v1/85184ac9bddf287f014d6167.png"},{"id":101517092,"identity":"85b2fa9c-cead-4ec9-81ba-77b19351fa04","added_by":"auto","created_at":"2026-01-30 16:22:57","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":841235,"visible":true,"origin":"","legend":"\u003cp\u003eMetabolomic Associations with Maternal Phenotypes and Gestational Outcomes \u003cstrong\u003eA\u003c/strong\u003e. Correlation analysis between metabolites, age, BMI, blood collection gestational age and delivery gestational age; \u003cstrong\u003eB\u003c/strong\u003e and \u003cstrong\u003eC\u003c/strong\u003e Kaplan-Meier analysis of pPROM and sPL.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8487634/v1/300f204c997b8b3a6f8c8e03.png"},{"id":101752699,"identity":"99428150-3d62-4a2d-a7b1-43601aa43afd","added_by":"auto","created_at":"2026-02-03 10:29:01","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2032925,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8487634/v1/853047db-2994-4504-8cfd-38b7c9d9705e.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Machine Learning-Driven Early Prediction of Spontaneous Preterm Birth Subtypes from Second-Trimester Plasma Metabolomic","fulltext":[{"header":"Introduction","content":"\u003cp\u003ePreterm birth, defined as birth before 37 weeks of gestation, is a leading cause of neonatal morbidity and mortality worldwide(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). The current preterm birth rate is 10.6%, and it can be categorized into two main types: iatrogenic preterm birth, which is induced by medical interventions, and spontaneous preterm birth(sPTB), which includes preterm premature rupture of membranes (pPROM) and preterm labor with intact membranes (sPL)(\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). Despite efforts to predict preterm birth using various methods such as ultrasound measurements, combined indicators, lipid biomarkers, vaginal flora levels, and psychological factors like anxiety and happiness, there is still no reliable predictive model available. Further research is needed, particularly to distinguish between sPL and pPROM.\u003c/p\u003e \u003cp\u003eA multitude of physiological changes and metabolic adaptations occur weekly during pregnancy. Metabolomics is a method that can comprehensively analyze metabolites in organisms, offering a new perspective for early disease diagnosis. In 2020, Mads Melbye et al. developed a metabolic clock that accurately predicts gestational age, aligning closely with early pregnancy ultrasound, the clinical gold standard(\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e). Blood samples are the preferred type for predicting preterm birth due to the limitations of placenta and amniotic fluid sampling as well as the instability of urine. Although studies have investigated alterations in serum metabolites in preterm and normal pregnant women, it is important to note that these studies typically focus on gestational age after 24 weeks or between 29\u003csup\u003e+\u0026thinsp;0\u003c/sup\u003e weeks and 36\u003csup\u003e+\u0026thinsp;5\u003c/sup\u003e weeks(\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). The survival rate of premature infants at the earliest gestational age has increased significantly in developed countries, extending their viability limit to 22 to 23 weeks of gestation. However, their risk of death and morbidity remains higher(\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). Therefore, predicting premature birth at an earlier gestational age is crucial for enabling early intervention and treatment for these infants. Additionally, distinguishing between sPL and pPROM is important for implementing customized interventions to improve outcomes for both conditions.\u003c/p\u003e \u003cp\u003eWe employed liquid chromatography-mass spectrometry to examine the non-targeted metabolomics of maternal blood samples during the second trimester (14\u0026ndash;20 weeks). This research revealed variations in metabolites among various types of preterm birth and used a logistic regression model to predict these types during the second trimester, providing methods and markers for earlier prediction of preterm birth.\u003c/p\u003e"},{"header":"Material and Methods","content":"\u003cp\u003e1. \u003cstrong\u003eClinical data\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFrom October 2022 to October 2024, peripheral blood samples were collected from 3,000 pregnant women at 14-20 weeks of gestation and stored at \u0026minus;80\u0026deg;C. The inclusion criteria were singleton pregnancies conceived naturally, absence of underlying diseases, and no family history of genetic disorders. Preliminary exclusion criteria included fetal structural or chromosomal abnormalities, and pregnancy complications such as gestational diabetes and preeclampsia. Participants were categorized into three groups: (1) a normal control group comprising women who delivered at full term without complications; (2) a pPROM group involving women with spontaneous membrane rupture between 28 and 37 weeks of gestation; and (3) an sPL group consisting of women with intact fetal membranes spontaneous preterm labor between 28 and 37 weeks of gestation. The received samples were randomly divided into two distinct sets for metabolomic testing. The sets comprised 15 controls, 10 pPROM cases, and 10 sPL cases each. Each set was designated as either a test set or a validation set following established stratification protocols. All research procedures followed the ethical guidelines of the Declaration of Helsinki and were approved by the Research Ethics Committee of Tianjin Central Hospital of Obstetrics and Gynecology (Approval No. 2022KY068). Written consent was obtained from all participants. The selection process for the samples is outlined in Figure 1.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2. Sample Analysis and Data Processing\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor each plasma sample, 100 \u0026mu;L was mixed with 400 \u0026mu;L of 80% methanol aqueous solution. After placing the mixture in an ice bath for 5 minutes, it was centrifuged at 15,000 g for 20 minutes at 4\u0026deg;C. The supernatant was collected and diluted with MS-grade water to adjust the methanol content to 53%, followed by another centrifugation at 15,000 g for 20 minutes at 4\u0026deg;C. The resulting supernatant was then subjected to liquid chromatography-mass spectrometry (LC-MS) analysis using a Dionex Ultimate 3000 ultra-high-performance liquid chromatography system(7) . Chromatographic separation was performed using a Thermo Syncronis C18 column (2.1 mm \u0026times; 100 mm, 1.7 \u0026mu;m). The mobile phase A consisted of water containing 0.1% formic acid (v/v) and 2 mM ammonium formate, while mobile phase B consisted of acetonitrile. The gradient elution conditions were as follows: 0\u0026ndash;1 min, 95% A; 1\u0026ndash;5 min, 95%\u0026ndash;40% A; 5\u0026ndash;8 min, 40%\u0026ndash;0% A; 8\u0026ndash;11 min, 0% A; 11\u0026ndash;14 min, 0%\u0026ndash;40% A; and 15\u0026ndash;18 min, 95% A. Quality control (QC) samples were prepared by pooling equal volumes from each plasma sample. Samples were analyzed in random order, with one QC sample inserted every six experimental samples to monitor system stability and ensure the reliability of the experimental data. Mass spectrometry was performed using an electrospray ionization (ESI) source, operating in both positive and negative ion modes. The ESI voltage was set to 2.8 kV, with sheath gas flow at 35 arb, auxiliary gas flow at 10 arb, and a capillary temperature of 320\u0026deg;C. The full MS scan resolution was set to 70,000, with a scan range of 70\u0026ndash;1050 m/z. The data-dependent secondary scan (full MS/dd-MS) had a resolution of 17,500, with stepped normalized collision energy (NCE) values of 20, 40, and 60 V.\u003c/p\u003e\n\u003cp\u003eRaw data files (.raw) were imported into TraceFinder 3.2.0 software for library searching. Each metabolite was filtered based on retention time, mass-to-charge ratio, and other parameters, with a retention time deviation of 0.2 min and a mass accuracy threshold of 5 ppm used for peak alignment across different samples. Peak areas were quantified, and spectral matching was performed against the mzCloud database (https://www.mzcloud.org/) and a locally constructed database. The quantitative results were then normalized, providing identification and relative quantification of metabolites in plasma samples. Data processing was conducted using the Linux operating system, with the R (4.4.1) and Python (3.13.2) software environments.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3. Bioinformatics Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMetabolites were annotated using KEGG\u0026nbsp;3.13.2, HMDB, and LIPIDMaps databases. Data preprocessing was conducted using metaX software, followed by Partial Least Squares Discriminant Analysis (PLS-DA)(8). Differential metabolites were identified based on Variable Importance in the Projection (VIP) \u0026gt; 1, \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05, and fold change (FC) \u0026ge; 2 or \u0026le; 0.5. Permutation tests (100 iterations) were used to assess model overfitting by comparing R\u0026sup2;Y and Q\u0026sup2;Y values with real models. Volcano and KEGG enrichment bubble plots were created using ggplot2. Metabolites were selected with thresholds of VIP \u0026gt; 1, FC \u0026gt; 1.5, and \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05. Heatmaps were generated with Pheatmap, and data were normalized with z-scores.\u003c/p\u003e\n\u003cp\u003eUnsupervised clustering of the 426 most variable metabolites was conducted using K-means and Euclidean distance. To identify most variable metabolites, we chose the metabolites in 426 Least absolute shrinkage and selection operator (LASSO) regression was employed for metabolites selection. The samples with selected nine metabolites were randomly split into training and test sets based on the results of two metabolic panel tests. The logistic regression model was built with glm() with binomial logit option and validated on the test set. Box plots were created using geom_boxplot. Pearson correlations were analyzed with cor (), with significance (\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05) determined using cor.mtest (). Correlation plots were generated with corrplot.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4 Statistical analyses\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe conducted statistical analysis on the clinical and demographic characteristics of all plasma samples. After normality and homogeneity of variance tests, we employed the Kruskal-Wallis H test on Age, BMI, Delivery gestational age, and Birthweight. For the Blood sampling gestational age, we conducted a one-way ANOVA with Tukey\u0026apos;s post hoc test. Additionally, Fisher\u0026apos;s exact test was used for the gender of newborns. The baseline data obtained were presented in Table 1.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003e1. Clinical Characteristics\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe sample collection process for the Normal, pPROM, and sPL groups is detailed in Figure 1, with baseline data summarized in Table 1. No statistically significant differences were observed in baseline characteristics across the three groups. The mean maternal age was approximately 30 years. Significant differences in BMI and birth weight were noted between the pPROM and sPL groups compared to the term group. Plasma samples were predominantly collected between 12 and 13 weeks of gestation, with a consistent newborn sex ratio across groups. Gestational age at delivery and newborn weight in the term and preterm groups were consistent with the expected clinical characteristics of these conditions.\u003c/p\u003e\n\u003cp\u003eTable 1. Clinical and demographic information on pregnancies included in the study\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"634\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20.8531%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.8531%;\"\u003e\n \u003cp\u003eControl\u003c/p\u003e\n \u003cp\u003en=30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23.8547%;\"\u003e\n \u003cp\u003epPROM\u003c/p\u003e\n \u003cp\u003en=20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23.8547%;\"\u003e\n \u003cp\u003esPL\u003c/p\u003e\n \u003cp\u003en=20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10.5845%;\"\u003e\n \u003cp\u003e\u003cem\u003ep\u003c/em\u003e value\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20.8531%;\"\u003e\n \u003cp\u003eAge，years\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20.8531%;\"\u003e\n \u003cp\u003e28.61（27.05，30.16）\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23.8547%;\"\u003e\n \u003cp\u003e30.47（28.79，32.16）\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23.8547%;\"\u003e\n \u003cp\u003e31.65（28.99，34.31）\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10.5845%;\"\u003e\n \u003cp\u003e0.18\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20.8531%;\"\u003e\n \u003cp\u003eBMI,kg/m\u003csup\u003e2 a\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.8531%;\"\u003e\n \u003cp\u003e20.93（20.29，21.57）\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23.8547%;\"\u003e\n \u003cp\u003e22.49（20.98，24.00）\u003csup\u003e*\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23.8547%;\"\u003e\n \u003cp\u003e23.73（22.65，24.81）\u003cstrong\u003e\u003csup\u003e#\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10.5845%;\"\u003e\n \u003cp\u003e＜0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20.8531%;\"\u003e\n \u003cp\u003eBlood sampling gestational age,weeks\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.8531%;\"\u003e\n \u003cp\u003e12.58\u0026plusmn;0.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23.8547%;\"\u003e\n \u003cp\u003e12.83\u0026plusmn;0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23.8547%;\"\u003e\n \u003cp\u003e12.41\u0026plusmn;0.39\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10.5845%;\"\u003e\n \u003cp\u003e0.035\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20.8531%;\"\u003e\n \u003cp\u003eDelivery gestational age,weeks\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.8531%;\"\u003e\n \u003cp\u003e39.83（39.46，40.19）\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23.8547%;\"\u003e\n \u003cp\u003e34.56（33.69，35.43）\u003csup\u003e*\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23.8547%;\"\u003e\n \u003cp\u003e34.61（33.44，35.77）\u003csup\u003e#\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10.5845%;\"\u003e\n \u003cp\u003e＜0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20.8531%;\"\u003e\n \u003cp\u003eMale infant (n,%)\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.8531%;\"\u003e\n \u003cp\u003e15 (50)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23.8547%;\"\u003e\n \u003cp\u003e11 (55)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23.8547%;\"\u003e\n \u003cp\u003e13(65)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10.5845%;\"\u003e\n \u003cp\u003e0.621\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20.8531%;\"\u003e\n \u003cp\u003eBirthweight, Kg\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.8531%;\"\u003e\n \u003cp\u003e3.37（3.24，3.50）\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23.8547%;\"\u003e\n \u003cp\u003e2354.74（2110.81，2598.66）\u003cstrong\u003e\u003csup\u003e*\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 23.8547%;\"\u003e\n \u003cp\u003e2368.50（2082.40，2654.60）\u003cstrong\u003e\u003csup\u003e#\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10.5845%;\"\u003e\n \u003cp\u003e＜0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003csup\u003ea\u003c/sup\u003e: Data presented as \u0026lsquo;mean(95% confidence interval for difference)\u0026rsquo;\u0026nbsp;and analyzed by kruskal-Wallis H test\u003c/p\u003e\n\u003cp\u003e\u003csup\u003eb\u003c/sup\u003e: Data presented as \u0026lsquo;mean\u0026plusmn;standard\u0026rsquo; deviation and analyzed by one-way ANOVA with Tukey\u0026rsquo;s posthoc test.\u003c/p\u003e\n\u003cp\u003e\u003csup\u003ec\u003c/sup\u003e: Data presented as \u0026lsquo;n (%)\u0026rsquo; and analyzed by Fisher\u0026rsquo;s exact test\u003c/p\u003e\n\u003cp\u003e\u003csup\u003e*\u003c/sup\u003e: Normal vs. pPROM, \u003cem\u003ep\u003c/em\u003e＜0.05\u003c/p\u003e\n\u003cp\u003e\u003csup\u003e#\u003c/sup\u003e: Normal vs. sPL, \u003cem\u003ep\u003c/em\u003e＜0.05\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2. Global Distribution of Maternal Plasma Metabolites in pPROM and sPL\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo assess the overall distribution of metabolites and uncover group-specific differences, Partial Least Squares Discriminant Analysis (PLS-DA) was conducted on maternal plasma samples from the pPROM and sPL groups, in comparison with the normal group. The analysis revealed significant differences in metabolite profiles between the preterm groups and the normal (\u003cstrong\u003eFigure 2A\u003c/strong\u003e). Permutation testing was further applied to evaluate the explanatory and predictive power of the PLS-DA models. The results showed no evidence of overfitting, with R\u0026sup2; values between 0.7 and 0.8, indicating strong model performance suitable for further screening and analysis (\u003cstrong\u003eFigure 2B\u003c/strong\u003e). A volcano plot of the 745 identified metabolites revealed 77 metabolites with significant differences in the pPROM group compared to normals, with 60 upregulated and 17 downregulated (\u003cstrong\u003eFigure 2C\u003c/strong\u003e). In the sPL group, 179 metabolites showed significant differences, 13 of which were upregulated, while the remainder were downregulated. A heatmap based on the differentially expressed metabolites was generated, visually representing the abundance variations across groups, providing an intuitive understanding of metabolite distribution differences (\u003cstrong\u003eFigure 2D\u003c/strong\u003e). KEGG pathway analysis of the identified differential metabolites showed significant enrichment in the arachidonic acid metabolism pathway within the pPROM group, particularly involving 6-Keto-prostaglandin F1\u0026alpha; and prostaglandin E2. In the sPL group, significant enrichment was observed in the biosynthesis of secondary metabolites pathway. Key metabolites included tyramine, xanthine, 9-oxononanoic acid, biliverdin, caffeine, carvone, cis-aconitate, citrulline, coumarin, eucalyptol, ferulic acid, genistein, geranyl diphosphate, histamine, L-tryptophan, nicotinic acid, pantothenic acid, pilocarpine, pipecolic acid, porphobilinogen, and pulegone. Additionally, the monoterpenoid biosynthesis pathway was significantly enriched, involving metabolites such as carvone, eucalyptol, geranyl diphosphate, and pulegone (\u003cstrong\u003eFigure 2E, F\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3. Development of Predictive Models for Different Types of Preterm Birth\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo identify metabolomic biomarkers for pPROM and sPL, metabolites from the pPROM and sPL groups were divided into training and validation sets based on the results of two metabolic panel tests. In the training set, 9 most important metabolites were identified using Least absolute shrinkage and selection operator (LASSO) regression, and logistic regression was used to build predictive models for the pPROM, sPL, and normal groups. These metabolites were: All-Trans-13,14-Dihydroretinol, Alpha-ketoisocaproic acid, Creatin/Creatinine, Kynurenic acid, LysoPC(P-16:0), LysoPE(0:0/20:4), Phosphocholine, Skatole. A boxplot revealed significant differences in metabolite concentrations between the control, pPROM, and sPL groups. Notably, the sPL group exhibited distinct metabolite profiles compared to both controls and pPROM (\u003cstrong\u003eFigure 3A\u003c/strong\u003e). The ROC curve illustrating the predictive performance based on these 9 metabolites is presented in Figure 4B. Receiver Operating Characteristic (ROC) curves assessed diagnostic accuracy, with the control group showing excellent performance (AUC = 0.984, 95% CI: 0.953\u0026ndash;1). The pPROM group demonstrated moderate diagnostic ability (AUC = 0.964, 95% CI: 0.909\u0026ndash;1), while the sPL group exhibited exceptional accuracy (AUC = 0.995, 95% CI: 0.982\u0026ndash;1), indicating the high diagnostic potential of these metabolites, particularly for sPL. Kaplan-Meier survival curves were used to evaluate risk stratification, revealing significant differences in risk probabilities for both pPROM (p = 0.0076) and sPL (p = 0.0081) (\u003cstrong\u003eFigure 3B\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4. Maternal Metabolic Determinants of Gestational Age Progression and Preterm Delivery Risk\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSubsequent analysis revealed correlations between nine metabolites and maternal age, BMI, blood sampling gestational age, and delivery gestational age. Creatin or creatinine showed significant correlations with other eight metabolites. Only phosphatoline exhibited a negative correlation with maternal age (R=-0.31), whereas Creatinine and LysoPC (P-16-0) were positively correlated with blood sampling gestational age (R=0.27, 0.23). Alpha ketoisocaproic acid, Creatin/Creatinine, Kynurenic acid, and Phosphatoline were correlated with BMI, while All-Trans-13,14-Dihydroretinol, Alpha-ketoisocaproic acid, Kynurenic acid, LysoPE (0:0/20:4), and Phosphocholine were correlated with delivery gestational age. Furthermore, a correlation was observed between BMI and delivery gestational age(R=-0.51) (\u003cstrong\u003eFigure 4A\u003c/strong\u003e). Subsequent analysis on metabolites and the risk of delivery gestational age indicated that high-risk groups had a higher likelihood of preterm birth, which decreased as gestational age progressed, while low-risk groups consistently maintained a lower risk (\u003cstrong\u003eFigure 4 B, C\u003c/strong\u003e).\u0026nbsp;\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study identified a panel of 9 plasma metabolites-All-Trans-13,14-Dihydroretinol, α-ketoisocaproic acid, Creatin/Creatinine, Kynurenic acid, LysoPC (P-16:0), LysoPE (0:0/20:4), Phosphocholine, and Skatole-that effectively distinguish between spontaneous preterm labor (sPL) and preterm premature rupture of membranes (pPROM) as early as 12\u0026ndash;13 weeks of gestation. PLS-DA models confirmed distinct metabolite profiles between preterm and term groups, without overfitting. Pathway analysis revealed significant enrichment of arachidonic acid metabolism in pPROM, notably involving prostaglandin E2 and 6-keto-prostaglandin F1α, while the sPL group showed enrichment in secondary metabolite biosynthesis. A logistic regression model based on these metabolites demonstrated high predictive accuracy, with AUCs of 0.995 for sPL, 0.964 for pPROM, and 0.984 for controls. Several metabolites correlated with BMI and gestational age at delivery, and risk stratification analysis confirmed that higher metabolic risk scores were associated with earlier delivery.\u003c/p\u003e \u003cp\u003ePreterm birth is the leading cause of neonatal morbidity and mortality worldwide, but there is a lack of definitive biomarkers to accurately predict the risk of preterm birth, especially in distinguishing between sPL and pPROM. Metabolomics provides a new approach for early disease detection. Previous studies have examined changes in serum metabolites in premature and normal pregnant women after 24 weeks of gestation or between 29 and 36\u0026thinsp;+\u0026thinsp;5 weeks, but have not effectively distinguished between sPL and pPROM (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). Currently, there is limited research on using ultra-performance liquid chromatography-mass spectrometry (UPLC-MS) to identify alterations in blood metabolomics during preterm labor, particularly between 14 and 10 weeks of gestation.\u003c/p\u003e \u003cp\u003eBlood samples are the preferred type of sample for predicting preterm birth, as sampling placenta and amniotic fluid is limited and urine is unstable. A study comparing metabolites in early preterm and term placentas found higher levels of dysregulated metabolites in early preterm placentas. However, there was no significant difference in the placental metabolome between spontaneous preterm and term births, suggesting that placentas of different gestational ages may produce different metabolites (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e). Significant changes in metabolites such as amino acids, fatty acids, lipids, hormones, and bile acids have been reported in the peripheral blood of premature patients after 24 weeks, with fatty acids being the most significant (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e). Additionally, a study found that lipid levels in the peripheral blood of preterm births after 29 weeks were increased (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). The studies mentioned above suggest that lipid metabolism may play a significant role in the occurrence and progression of preterm birth. However, it is unclear whether various types of spontaneous preterm birth are associated with lipid metabolism. Our results showed that the differential metabolites of different types of premature birth were not enriched in pathways related to lipid metabolism. The main metabolic pathways involved in sPL and pPROM were biosynthesis of secondary metabolites pathway and arachidonic acid metabolism pathway respectively. Specifically, the pPROM group shows alterations in the arachidonic acid metabolism biosynthetic pathway, with metabolites such as 6-keto-prostaglandin F1a and prostaglandin E2 being involved. Prostaglandin E2 is known to be used for inducing labor in cases of premature rupture of membranes before delivery (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e). Therefore, abnormal levels of prostaglandin E2 or its metabolites may contribute to the development of pPROM.\u003c/p\u003e \u003cp\u003eThe nine metabolites identified in this study -All-Trans-13,14-Dihydroretinol, α-ketoisocaproic acid (KIC), creatinine (via SUA/SCr ratio), kynurenic acid, LysoPC, LysoPE, phosphocholine, and skatole\u0026mdash;collectively underscore the multifactorial metabolic dysregulation underlying preterm birth (PTB). All-Trans-13,14-Dihydroretinol, a vitamin A derivative, exhibits elevated mid-trimester levels in PTB cases, suggesting its role in placental maldevelopment due to disrupted angiogenesis and retinoid signaling (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). Similarly, KIC, a leucine catabolite, demonstrates reduced levels in maternal blood, implicating energy metabolism impairment via inhibition of insulin-stimulated glucose transport in myotubes (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). The SUA/SCr (serum uric acid/serum creatine) ratio, reflecting maternal renal function and fetal kidney stress, emerges as a predictive biomarker for PTB, with elevated creatinine levels related to acute neonatal kidney injury. Creatine or creatinine exhibited significant correlations with eight other metabolites in this study, underscoring the relevance of these compounds in preterm birth(\u003cspan additionalcitationids=\"CR16\" citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e). Diminished kynurenine levels in preterm placental and fetal compartments correlate with neurodevelopmental risks, likely mediated by kynurenic acid imbalance, thus emphasizing immune-metabolic crosstalk within the kynurenine pathway(\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e). Lysophospholipids (LysoPC/LysoPE), elevated in adverse pregnancies, drive inflammation and trophoblast dysfunction-LysoPE (16:0) promotes aberrant trophoblast invasion, while LysoPC amplifies prostaglandin-mediated inflammatory cascades (\u003cspan additionalcitationids=\"CR21\" citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e). Phosphocholine depletion in PTB neonates, associated with membrane instability and apoptosis, underscores its role in placental lipid signaling (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e). Finally, skatole, though lacking direct PTB evidence, may exacerbate systemic inflammation via TLR activation and gut microbiota dysbiosis, aligning with emerging links between maternal microbiome shifts and preterm labor (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e). Phosphatoline was the only metabolite showing a negative correlation with maternal age in our study. Creatinine and LysoPC (P-16-0) were positively correlated with gestational age at blood sampling. Maternal age and gestational age at blood collection had minimal impact on metabolite prediction. Five of the nine metabolites analyzed were associated with BMI and gestational age, suggesting a potential link between BMI and premature birth, warranting further investigation. Certain metabolites were related to gestational age, with high-risk groups showing an increased risk of preterm birth, indicating the potential use of metabolites for predicting delivery time. Premature birth is characterized by disruptions in metabolic networks, inflammation, oxidative stress, and placental dysfunction. Understanding the alterations and interactions of these markers can aid in early prediction and intervention in premature birth.\u003c/p\u003e \u003cp\u003eAlthough this study employed ultra-performance liquid chromatography-mass spectrometry (UPLC-MS) for high-sensitivity metabolite detection, collected plasma samples during the second trimester (14\u0026ndash;20 weeks), identified potential biomarkers in the peripheral blood of pregnant women with preterm birth, described metabolite changes across different subtypes of preterm birth, and developed a multiclass prediction model capable of distinguishing between spontaneous preterm labor (sPL) and preterm premature rupture of membranes (pPROM), certain limitations remain. The findings are derived from a single-center cohort with a limited sample size, which restricts the generalizability of the results. Although nine promising metabolites associated with preterm birth were identified, external validation in larger, multicenter populations is essential to confirm their predictive value and support future clinical translation. Furthermore, longitudinal profiling of metabolite dynamics throughout gestation is required to refine the predictive model and enhance its clinical applicability.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eThis study identified key metabolic pathways, including arachidonic acid metabolism in pPROM and secondary metabolite biosynthesis in sPL, revealing distinct pathophysiological mechanisms underlying these subtypes. The identified metabolites are involved in inflammatory processes, oxidative stress, and placental dysfunction, reflecting the multifactorial pathophysiology of preterm birth. The study also discovered a panel of nine second-trimester plasma metabolites that effectively predict spontaneous preterm birth (sPTB) and differentiate between preterm premature rupture of membranes (pPROM) and spontaneous preterm labor (sPL) as early as 14\u0026ndash;20 weeks of gestation. Using machine learning approaches, a predictive model with high discriminative accuracy was developed. In contrast to previous studies focused on late gestation, this work provides novel metabolic insights at an earlier stage of pregnancy. Future research should further elucidate the mechanistic interactions among maternal-fetal immune responses, gut microbiota-derived signals such as skatole, and metabolic dysregulation, while promoting clinical translation through standardized assay development and longitudinal profiling to improve predictive reliability and support personalized intervention strategies.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe studies involving human participants were reviewed and approved by Human Research Ethics Committee of Tianjin Central Hospital of Obstetrics and Gynecology(NO.2022KY068). All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. All participants signed a written informed consent form prior to their participation in this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConsent for publication were obtained from that person.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors declare that they have no conflict of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by Noncommunicable Chronic Diseases-National Science and Technology Major Project (2023ZD0517600),\u0026nbsp;Tianjin Health Research\u0026nbsp;Project (TJWJ2025MS031), Tianjin Key Medical Discipline Construction Project (TJYXZDXK-3-029C).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026apos; contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors had full access to all the data in the study and accept responsibility to submit for publication. These authors meet the four criteria for authorship as specified in the ICMJE recommendations: Yongmei Shen contributed to the conceptualisation, funding acquisition, investigation, methodology, project administration, formal analysis and writing-original draft and writing- review and editing. Dan Wu contributed to the investigation, methodology, project administration, data curation, formal analysis. Hefei Wang contributed to the methodology, project administration, data curation writing-original draft. Tianxiang Liu contributed to the methodology,data curation. Yaqi Li, Maolin Nie, Rongxin Wei contributed to data curation, sample collection and processing. Jiaosong Cao and Qimei Lin contributed to the writing-review and editing. Junli He contributed to sample collection. Yongjun Piao contributed to methodology, formal analysis and writing-original draft. Ying Chang contributed to the conceptualisation, funding acquisition, investigation,writing- review.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe wish to acknowledge and thank all study participants.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eOhuma EO, Moller AB, Bradley E, Chakwera S, Hussain-Alkhateeb L, Lewin A, et al. National, regional, and global estimates of preterm birth in 2020, with trends from 2010: a systematic analysis. Lancet. 2023;402(10409):1261-71.\u003c/li\u003e\n\u003cli\u003ePrediction and Prevention of Spontaneous Preterm Birth: ACOG Practice Bulletin, Number 234. Obstet Gynecol. 2021;138(2):e65-e90.\u003c/li\u003e\n\u003cli\u003eLiang L, Rasmussen MH, Piening B, Shen X, Chen S, R\u0026ouml;st H, et al. Metabolic Dynamics and Prediction of Gestational Age and Time to Delivery in Pregnant Women. Cell. 2020;181(7):1680-92.e15.\u003c/li\u003e\n\u003cli\u003eLizewska B, Teul J, Kuc P, Lemancewicz A, Charkiewicz K, Goscik J, et al. Maternal Plasma Metabolomic Profiles in Spontaneous Preterm Birth: Preliminary Results. Mediators Inflamm. 2018;2018:9362820.\u003c/li\u003e\n\u003cli\u003eVirgiliou C, Gika HG, Witting M, Bletsou AA, Athanasiadis A, Zafrakas M, et al. Amniotic Fluid and Maternal Serum Metabolic Signatures in the Second Trimester Associated with Preterm Delivery. J Proteome Res. 2017;16(2):898-910.\u003c/li\u003e\n\u003cli\u003eBarfield WD. Public Health Implications of Very Preterm Birth. Clin Perinatol. 2018;45(3):565-77.\u003c/li\u003e\n\u003cli\u003eWant EJ, Masson P, Michopoulos F, Wilson ID, Theodoridis G, Plumb RS, et al. Global metabolic profiling of animal and human tissues via UPLC-MS. Nat Protoc. 2013;8(1):17-32.\u003c/li\u003e\n\u003cli\u003eWen B, Mei Z, Zeng C, Liu S. metaX: a flexible and comprehensive software for processing metabolomics data. BMC Bioinformatics. 2017;18(1):183.\u003c/li\u003e\n\u003cli\u003eCifkova E, Karahoda R, Stranik J, Abad C, Kacerovsky M, Lisa M, Staud F. Metabolomic analysis of the human placenta reveals perturbations in amino acids, purine metabolites, and small organic acids in spontaneous preterm birth. Excli j. 2024;23:264-82.\u003c/li\u003e\n\u003cli\u003eZhang Y, Wang J, Yu Y, Xie C, Xiao M, Ren L. Misoprostol versus prostaglandin E2 gel for labor induction in premature rupture of membranes after 34 weeks of pregnancy. Int J Gynaecol Obstet. 2015;130(3):214-8.\u003c/li\u003e\n\u003cli\u003eNaik VD, Lee J, Wu G, Washburn S, Ramadoss J. Effects of nutrition and gestational alcohol consumption on fetal growth and development. Nutr Rev. 2022;80(6):1568-79.\u003c/li\u003e\n\u003cli\u003eMa G, Chen Y, Liu X, Gao Y, Deavila JM, Zhu MJ, Du M. Vitamin a supplementation during pregnancy in shaping child growth outcomes: A meta-analysis. Crit Rev Food Sci Nutr. 2023;63(33):12240-55.\u003c/li\u003e\n\u003cli\u003eVlaardingerbroek H, Roelants JA, Rook D, Dorst K, Schierbeek H, Vermes A, et al. Adaptive regulation of amino acid metabolism on early parenteral lipid and high-dose amino acid administration in VLBW infants - a randomized, controlled trial. Clin Nutr. 2014;33(6):982-90.\u003c/li\u003e\n\u003cli\u003eMann G, Adegoke OAJ. Effects of ketoisocaproic acid and inflammation on glucose transport in muscle cells. Physiol Rep. 2021;9(1):e14673.\u003c/li\u003e\n\u003cli\u003eZhang E, Su S, Gao S, Zhang Y, Wang J, Liu J, et al. Elevated serum uric acid to creatinine ratio is associated with adverse pregnancy outcomes: a prospective birth cohort study. Int J Med Sci. 2024;21(9):1612-21.\u003c/li\u003e\n\u003cli\u003eWu Y, Wang H, Pei J, Jiang X, Tang J. Acute kidney injury in premature and low birth weight neonates: a systematic review and meta-analysis. Pediatr Nephrol. 2022;37(2):275-87.\u003c/li\u003e\n\u003cli\u003eGo H, Momoi N, Kashiwabara N, Haneda K, Chishiki M, Imamura T, et al. Neonatal and maternal serum creatinine levels during the early postnatal period in preterm and term infants. PLoS One. 2018;13(5):e0196721.\u003c/li\u003e\n\u003cli\u003evan Zundert SK, Broekhuizen M, Smit AJ, van Rossem L, Mirzaian M, Willemsen SP, et al. The Role of the Kynurenine Pathway in the (Patho) physiology of Maternal Pregnancy and Fetal Outcomes: A Systematic Review. Int J Tryptophan Res. 2022;15:11786469221135545.\u003c/li\u003e\n\u003cli\u003eMorrens M. The role of the kynurenine pathway in cognitive function in brain Disorders: Insights and challenges. Brain Behav Immun. 2025;127:110-1.\u003c/li\u003e\n\u003cli\u003eTsutsumi T, Yasuda K, Neya M, Okada H, Tokumura A. Lysophosphatidic acid production from lysophosphatidylcholine by lysophospholipase D activity of autotaxin in plasma of women with normal and adverse pregnancies. Prostaglandins Other Lipid Mediat. 2022;163:106670.\u003c/li\u003e\n\u003cli\u003eZhang Z, He P, Chen D, Tan Y, Chen A, Bian Z, Chen T. Active metabolomics identify potential functional metabolites for preeclampsia prevention. Clin Chim Acta. 2024;560:119717.\u003c/li\u003e\n\u003cli\u003eZhong W, Li Y, Zhong H, Cheng Y, Chen Q, Zhao X, et al. Exploring the mechanism of anti-chronic heart failure effect of qiweiqiangxin І granules based on metabolomics. Front Pharmacol. 2023;14:1111007.\u003c/li\u003e\n\u003cli\u003eBernhard W, Poets CF, Franz AR. Choline and choline-related nutrients in regular and preterm infant growth. Eur J Nutr. 2019;58(3):931-45.\u003c/li\u003e\n\u003cli\u003eDobrynina EA, Zykova VA, Surovtsev NV. In-plane and out-of-plane gigahertz sound velocities of saturated and unsaturated phospholipid bilayers from cryogenic to room temperatures. Chem Phys Lipids. 2023;256:105335.\u003c/li\u003e\n\u003cli\u003eGeorges HM, Norwitz ER, Abrahams VM. Predictors of Inflammation-Mediated Preterm Birth. Physiology (Bethesda). 2025;40(1):0.\u003c/li\u003e\n\u003cli\u003eLu X, Shi Z, Jiang L, Zhang S. Maternal gut microbiota in the health of mothers and offspring: from the perspective of immunology. Front Immunol. 2024;15:1362784.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-pregnancy-and-childbirth","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"prch","sideBox":"Learn more about [BMC Pregnancy and Childbirth](http://bmcpregnancychildbirth.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/prch/default.aspx","title":"BMC Pregnancy and Childbirth","twitterHandle":"@BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Premature birth, Plasma metabolomics, Premature premature rupture of membranes, Spontaneous preterm birth","lastPublishedDoi":"10.21203/rs.3.rs-8487634/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8487634/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003ePreterm birth is a major cause of neonatal morbidity and mortality, with spontaneous preterm birth (sPTB) comprising preterm premature rupture of membranes (pPROM) and spontaneous preterm labor (sPL). Reliable early prediction remains challenging, particularly in distinguishing sPTB subtypes. Metabolomics offers a promising approach for identifying predictive biomarkers.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eA single-center case-control study was conducted using archived maternal plasma samples. Participants included 70 pregnant women (30 term controls, 20 pPROM, 20 sPL) at 14\u0026ndash;20 weeks\u0026rsquo; gestation. Non-targeted metabolomic profiling was performed via liquid chromatography-mass spectrometry (LC-MS). Metabolite screening was carried out using LASSO regression, and pathway enrichment analysis was conducted. Machine learning models (logistic regression) were developed and validated. Statistical analyses included PLS-DA, ROC curves, Pearson correlation, and risk stratification.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eNine metabolites associated with inflammatory activation, oxidative stress, and placental dysfunction were identified. LASSO models achieved high predictive accuracy (AUCs: 0.984 for controls, 0.964 for pPROM, 0.995 for sPL). Creatinine and LysoPC(P-16:0) was positively correlated with gestational age at blood sampling (R\u0026thinsp;=\u0026thinsp;0.27/0.23), while phosphatidylcholine was negatively correlated with maternal age (R=-0.31). Gestational age at delivery was negatively correlated with BMI (R\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.51). High-risk stratification showed a decreasing probability of preterm birth with increasing gestation, while low-risk stratification remained stable.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eSecond-trimester plasma metabolomics combined with machine learning could effectively predict sPTB and distinguish its subtypes. These findings support the potential for early risk stratification and personalized intervention, though multicenter validation is needed for clinical translation.\u003c/p\u003e","manuscriptTitle":"Machine Learning-Driven Early Prediction of Spontaneous Preterm Birth Subtypes from Second-Trimester Plasma Metabolomic","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-30 16:22:52","doi":"10.21203/rs.3.rs-8487634/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-05-07T18:31:45+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-20T11:19:34+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-10T18:05:23+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"246028413206643774608564774502963560154","date":"2026-02-10T17:53:37+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"129995665867176620393322002247135209504","date":"2026-02-09T18:20:46+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-06T19:13:36+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-04T21:21:07+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"294199101678579722710097116862553899621","date":"2026-02-04T19:12:30+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"122374414149371929397261676200989805624","date":"2026-01-28T11:05:31+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-01-28T09:16:15+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-01-02T04:39:59+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-01-01T23:24:32+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-01-01T23:24:19+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Pregnancy and Childbirth","date":"2025-12-31T08:15:58+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-pregnancy-and-childbirth","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"prch","sideBox":"Learn more about [BMC Pregnancy and Childbirth](http://bmcpregnancychildbirth.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/prch/default.aspx","title":"BMC Pregnancy and Childbirth","twitterHandle":"@BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"14f82589-e908-4791-a2c6-ad39ea1e706b","owner":[],"postedDate":"January 30th, 2026","published":true,"recentEditorialEvents":[{"type":"decision","content":"Revision requested","date":"2026-05-07T18:31:45+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[],"tags":[],"updatedAt":"2026-05-07T18:38:55+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-30 16:22:52","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8487634","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8487634","identity":"rs-8487634","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}