Personalized Prediction of Glycemic Responses to Food in Women with Gestational Diabetes: Gut Microbiota Matters

Personalized Prediction of Glycemic Responses to Food in Women with Gestational Diabetes: Gut Microbiota Matters | 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 Article Personalized Prediction of Glycemic Responses to Food in Women with Gestational Diabetes: Gut Microbiota Matters Polina V. Popova, Artem O. Isakov, Anastasia N. Rusanova, Stanislav I. Sitkin, and 19 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4850670/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 07 Feb, 2025 Read the published version in npj Biofilms and Microbiomes → Version 1 posted 10 You are reading this latest preprint version Abstract We aimed to develop a prediction model for postprandial glycemic response (PPGR) in pregnant women with gestational diabetes mellitus (GDM) and to explore the influence of gut microbial data on prediction accuracy. We enrolled 105 pregnant women (70 GDM and 35 healthy). Participants underwent continuous glucose monitoring (CGM) for 7 days and provided detailed food diaries. Stool samples were collected at 28.8 ± 3.6 gestational weeks, followed by 16S rRNA gene sequence analysis. We developed machine learning algorithms for predicting PPGR, incorporating CGM measurements, meal content, lifestyle factors, biochemical parameters, anthropometrics, and gut microbiota data. The accuracy of the models with and without gut microbiota were compared. PPGR prediction models were created based on 2,706 meals with measured PPGRs. The integration of microbiome data in models increased the explained variance in peak glycemic levels (GLUmax) from 34–42% and the explained variance in the incremental area under the glycemic curve 120 minutes after meal start (iAUC120) from 50–52%. The final model performed better than the model based solely on carbohydrate count in terms of correlation between predicted and measured PPGRs (r = 0.72 vs r = 0.51 for iAUC120 and r = 0.66 vs r = 0.35 for GLUmax). After summing the SHAP values of associated features, the microbiome emerged as the fourth most impactful parameter for GLUmax and iAUC120 prediction, following meal composition, CGM measurements, and meal context. Microbiome features rank among the top 5 most impactful parameters in predicting PPGR in women with GDM. Biological sciences/Microbiology Health sciences/Health care glycemic response personalized nutrition microbiome postprandial glycemia Intestinibacter bartlettii Butyricicoccus faecihominis “Lachnoclostridium edouardi” Ruminococcus champanellensis Lachnospira eligens continuous glucose monitoring. Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Gestational diabetes mellitus (GDM) represents a prevalent condition, impacting a substantial portion, approximately up to 9-26%, of pregnancies [1]. GDM can lead to pregnancy complications, including but not limited to increased cesarean section rate, birth trauma, nerve palsy, neonatal hypoglycemia in the short-term period [2], and increased risks of obesity, type 2 diabetes, and cardiovascular diseases in both mothers and their offspring throughout life [3]. It is crucial to uphold normal glycemic levels during pregnancy to mitigate adverse pregnancy outcomes and disrupt the cyclical transmission of predisposition to metabolic diseases across generations [3,4]. The most common treatment for GDM is diet and lifestyle modification, reportedly effective without adding medications for achieving glucose control in 70 - 85% of women [5]. However, according to real life observations, many women with GDM do not achieve target glucose levels and the rate of pregnancy complications remains high in women with GDM [6]. Nutritional studies and guidelines concerning GDM concentrate on average characteristics across populations [5]. The historically prevailing approach to predicting postprandial glucose responses (PPGRs) to food involves relying on the carbohydrate content of the meal [5], despite evidence suggesting its inadequacy as a predictor [7]. Alternative methods include the glycemic index, which assesses the postprandial glucose response to a specific food, and the derived glycemic load [5]. Attributing a singular postprandial glucose response (PPGR) to each food implies that the response is entirely inherent to the food itself. Yet, recent studies exploring interindividual variations in PPGRs have revealed significant variability in how different individuals respond to identical foods [7, 8]. The distinctive PPGRs of an individual were shown to be influenced by their biological traits (such as gut microbiome composition and genetic variation) and lifestyle factors [7, 8, 9]. Several studies integrated gut microbial features into the models predicting PPGRs in healthy individuals [7, 8] or patients with type 1 diabetes [9]. However, only a few evaluated the impact of microbiome on the prediction of PPGRs, through variance components analyses [8] or the SHapley Additive exPlanation (SHAP) method [9], and only one study exclusively concentrated on investigating the connections between bacterial species and the host glycemic regulation [10]. Furthermore, none of these studies directly compared the accuracy metrics of the models before and after addition of microbial data into the list of the input variables. They compared basic models including carbohydrates and premeal glucose level with a full model including multiple individual parameters apart from microbial features [7, 8, 9]. Moreover, although there is mounting evidence regarding the regulatory functions of the microbiome in normal and impaired glycemic responses among non-pregnant individuals [7 - 10], limited knowledge exists concerning the microbiome's impact on PPGRs among pregnant women, both with and without GDM. We have previously developed PPGR prediction models based on multiple individual parameters without microbiome for pregnant women with and without GDM [11]. The performance of our model in predicting the incremental area under the glucose curve in the 2 hours after the meal (iAUC120) (R =0.7) was comparable to the model created by Zeevy et al. for healthy individuals based on individual parameters and microbiome (R =0.7) [7]. However, our past model was inferior to the accuracy of the microbiome-based model subsequently developed by Berry et al. for iAUC120 prediction in healthy individuals (R =0.77) [8], leaving space for the improvement of our model, potentially with inclusion of microbiome data. Another distinguishing feature of the Berry model, in comparison to both our previous model and the Zeevi model, was the incorporation of genetic factors. This likely enhanced the predictive accuracy, as genetics constituted the second most crucial parameter group after serum glycemic markers, as indicated by the proportion of variance explained (R 2 ) [8]. The aim of this study was to develop a prediction model for PPGR in pregnant women with GDM and to explore the impact of microbial data on the model’s performance. An accurate PPGR prediction model holds promise in optimizing personalized diet recommendations to improve glucose control and pregnancy outcomes in women with GDM. Simultaneously, the identification of a distinct gut microbial signature affecting PPGR, a secondary aim of this project, could serve as a basis for the development of potential therapeutic interventions. Research Design and Methods Study Design We recruited pregnant women who participated in the randomized controlled trial “Genetic and Epigenetic Mechanisms of Developing Gestational Diabetes Mellitus and its Effects on the Fetus” (GEM-GDM), consented to be connected to a continuous glucose monitoring (CGM) system (CGMS) for at least 7 days, tracked information on food consumption in a designated mobile app, and provided stool samples. The parent GEM-GDM study aimed to compare different glycemic targets for women with GDM. It was registered at ClinicalTrials.gov (Identifier: NCT03610178) and its design is described elsewhere [ 12 ]. Briefly, gravidas with GDM were randomly assigned into two groups according to their glycemic goals: the first group had strict glycemic goals (< 5.1 mmol/L for fasting blood glucose (BG) and < 7.0 mmol/L for 1-hour postprandial BG), and the second group had less strict glycemic goals (< 5.3 mmol/L and < 7.8 mmol/L, respectively). For this study, the women from both groups were combined to create the GDM group. GDM was diagnosed using a single-step 75-g OGTT according to the recommendation of the International Association of Diabetes and Pregnancy Study Groups (IADPSG) [ 13 ]. Apart from women with GDM, we also included healthy pregnant women with normal values of plasma glucose during oral glucose tolerance test (OGTT) (controls). At study initiation, a physician acquired informed consent, recorded medical history and took anthropometric measurements (weight, height, waist circumference, blood pressure and heart-rate). Pregestational body mass index (BMI) was calculated by dividing self-reported pregestational weight (in kilograms) by the square of height (in meters). Blood tests, including fasting plasma glucose, lipid profile and HbA1c, were performed in the Almazov National Medical Research Centre laboratory. Participants filled out questionnaires concerning their lifestyle before and during pregnancy and were then connected to CGMS for 7–14 days during which they tracked information on meal consumption in the proprietary mobile app DiaCompanion, as described elsewhere [ 14 ]. Each consumed food item was recorded by selecting it from a database created by the authors on the basis of reference books of the Russian Academy of Medical Sciences and the US Department of Agriculture (USDA) Food Composition Databases (Release 28) with the expansion of additional items by certified dietitians. The distinctive feature of this food database, in addition to a wide selection of foods (more than 5500 items), is the presence of glycemic index (GI). Each food item in the database was assigned a dietary GI [ 15 ]. The CGM and meal-related data were processed using a previously described algorithm [ 15 ]. The study was approved by the local ethics committee of the Almazov National Medical Research Centre, Russia (protocol. №119). Inclusion and exclusion criteria Participation in this study was optional for participants of the GEM-GDM trial. The GDM group included pregnant women with GDM and a gestational age of ≥ 24 weeks at the start of CGM. The control group comprised pregnant women with normal glucose tolerance, confirmed by OGTT between 24 and 32 weeks of gestation. In addition to the inclusion criteria used in the GEM-GDM study, for this particular study, consent to be connected to a CGMS for 7 days and a capability to work with a mobile phone app for the recording of dietary intake in real time were required. Exclusion criteria included an active inflammatory or neoplastic disease, any known medical condition affecting glucose metabolism (with the exception of GDM), current insulin use, antibiotic usage 2 months prior to participation in the study, failure to provide a stool sample and submission of inaccurate food diaries through the app. Taking into account that accurate logging is crucial for analysis of PPGRs to food, a set of rules was formulated by the authors to filter negligently filled-in and misreported diaries: (1) more than 50% of the logged meals comprised of a single dish or a single dish with a single beverage; (2) the average amount of logged calories per day was less than 1000 kcal; (3) more than 50% of the logged weights of food items were rounded to the hundreds (excluding beverages); (4) the amount of logged snacks was less than 10% of all meal records [ 11 ]. Participants with misreported diaries were excluded. Lifestyle questionnaire The questionnaire comprised several sections covering various aspects: frequency of consuming staple items per week (such as fruits, pastries, skimmed dairy products, legumes, meat, sausage products, dried fruits, fish, whole-grain bread, sauces, vegetables, alcohol, sweet drinks, and coffee), levels of physical activity (daily walking duration categorized as 60 min/day; daily frequency of stair climbing categorized as 16 flights/day; frequency of engaging in sports activities categorized as 3 days/week), and smoking habits before and during pregnancy. Each section of the questionnaire was structured in a semi-quantitative manner. This questionnaire has been previously documented [ 15 ]. For the description of the parameters from lifestyle questionnaire included in the final dataset, please refer to Supplementary Table S1 . Blood samples were collected by a certified nurse after 8–12 hours of fasting. The blood panel included measurement of glycosylated hemoglobin (HbA1c%), plasma glucose, total and HDL cholesterol, triglycerides, insulin, leptin. and fructosamine levels in the central lab of the Almazov National Medical Research Centre. Plasma glucose concentration was determined by the glucose oxidase method in fresh plasma samples. HbA1c was measured in fresh whole blood samples using high performance liquid chromatography (HPLC) (D10 HbA1c). Blood for genotyping of pregnant women and serum for other biochemical analysis were stored at -80°C until the analysis. Serum fasting insulin levels were measured using the electrochemiluminescence immunoassay (Roche Diagnostics, GmbH, Germany). The homeostatic model assessment (HOMA) index was calculated using the following formula: fasting serum insulin (m IU/L) × fasting plasma glucose (mmol/L)/(22.5) as an insulin resistance indicator. Total cholesterol, HDL-C, LDL-C, VLDL-C, and triglyceride levels were measured utilizing enzymatic colorimetric methods with diagnostic reagent system designed for the Cobas Integra Autoanalyzer. Serum leptin levels were measured using an enzyme-linked immunosorbent assay (ELISA) as recommended by the manufacturer (Diagnostics Biochem Canada Inc., Canada). Continuous glucose monitoring (CGM) was conducted using the iPro2TM system from Medtronic, MN, USA. This system utilizes EnliteTM sensors placed subcutaneously to measure interstitial glucose levels. To align CGM readings with blood glucose levels, participants also utilized finger-prick measurements with the Accu Chek Performa from Roche, Germany. Participants were instructed to perform four daily blood glucose measurements. To enhance accuracy, participants were specifically asked to measure blood glucose levels before meals, following recommendations outlined in the iPro2 manual. Calibration of CGM measurements was performed using the CareLink online software from Medtronic, following the guidelines provided in the iPro2 manual. Food diary tracking was facilitated through our proprietary mobile app DiaCompanion. Each consumed food item was recorded by selecting it from a database created by the authors on the basis of reference books of the Russian Academy of Medical Sciences and the US Department of Agriculture (USDA) Food Composition Databases (Release 28) with the expansion of additional items by certified dietitians. The distinctive feature of this food database, in addition to a wide selection of foods (more than 5,500 items), is the presence of glycemic index (GI). Each food item in the database was assigned a dietary GI [ 15 ]. Participants were instructed to meticulously log their daily activities using this platform. They were required to document precise details, including the components and weights of each meal, sleep and wake-up times. Participants were informed of the importance of accurate logging, particularly emphasizing the correct timing of meal logging and accurate recording of food components. Research physicians conducted weekly reviews of each participant's loggings. Any uncertainties in the logs were addressed directly with the participants. Meal preprocessing Before a meal and corresponding PPGR were added to the dataset for model training, the following filters were implemented (mainly to exclude recordings with incorrect timing): (1) a meal followed by a subsequent meal less than 60 minutes after its start; (2) a meal on the peak of a CGM-curve: an increase in glucose levels by more than 1 mmol/L during an hour preceding the index meal; (3) a meal on the falling edge of a CGM peak; and (4) a meal with inadequately low PPGR (iAUC120 40 g) [ 11 ]. Glucose level at baseline was considered as the lowest glucose value within ± 15 min from self-reporting of the meal in the app. DNA and genotyping of blood samples Genomic DNA was extracted from blood samples using the FlexiGene DNA Kit from Qiagen, (Hilden, Germany). Genotyping of the following variants: HKDC1 (rs10762264), MTNR1B (rs10830963 and rs1387153), GCK (rs1799884), KCNJ11 (rs5219), IGF2BP2 (rs4402960), TCF7L2 (rs7903146), CDKAL1 (rs7754840), FTO (rs9939609), and IRS1 (rs1801278), was conducted through real-time PCR utilizing custom kits from Applied Biosystems, based in the USA. The procedures recommended by the manufacturer were followed meticulously. Each primer tube contained a concentrated mixture of SNP Genotyping Assay Mix, comprising polymorphism-specific direct and reverse primers, along with two TaqMan MGB probes: one tagged with VIC dye for allele 1 identification and the other tagged with FAM dye for allele 2 identification. Following replication of 10% of the samples, the discordance rate was determined to be less than 0.1%. Microbiome: DNA extraction DNA was extracted from all collected samples using the PowerSoil DNA Isolation Kit (MO BIO, Carlsbad, CA, USA) according to the manufacturer's instructions and following a 2 min bead beating step (BioSpec, Bartlesville, OK, USA). Next, the variable V4 region was PCR-amplified using the 515F and 806R barcoded primers following the Earth Microbiome Project protocol [ 16 ]. Each PCR reaction contained 25µl with ~ 40 ng/µl of DNA, 2 µl 515F (forward, 10µM) primer, 2 µl 806R (reverse, 10µM) primer, and 25 µl PrimeSTAR Max PCR Readymix (Takara, Mountain view, CA, USA). PCR conditions were as follows: 30 cycles of denaturation at 98°C for 10 sec, annealing at 55°C for 5 sec, and extension at 72°C for 20 sec, followed by a final elongation at 72°C for 1 min. Amplicons were purified using AMPure magnetic beads (Beckman Coulter, Indianapolis, IN, USA) and quantified using the Picogreen dsDNA quantitation kit (Thermofisher, Waltham, MA, USA). Equimolar amounts of DNA from individual samples were pooled and sequenced using the Illumina MiSeq platform at the Genomic Center at the Bar-Ilan University, Azrieli Faculty of Medicine. Appropriate negative and positive controls were included at all stages of analysis. Bioinformatics and microbiome analysis The quality of raw reads was assessed with FastQC v. 0.11.9 [ 17 ] and MultiQC v. 1.14 [ 18 ]. Reads were trimmed and filtered with Trimmomatic v. 0.39 [ 19 ] (SE -phred 33 HEADCROP 31 ILLUMINACLIP:2:30:10 SLIDINGWINDOW:4:15 MINLEN:150). The remaining reads were processed with the DADA2 pipeline v. 3.6.2. [ 20 ], including additional trimming, denoising, and error correction. The derived sequences - amplicon sequence variants (ASVs) were clustered using MMseqs2 v. 13.45111 [ 21 ] (identity 99%, coverage 80%). The resultant representative sequences were treated as operative taxonomic units (OTUs). We clustered ASVs to OTUs to reduce the number of sequencing errors inherent in ASVs and avoid false diversity. The OTUs were returned to DADA2 for taxonomy assignment with SILVA SSU database v.138.1 [ 22 ]. Sequences classified as eukaryotes were removed. Only samples containing more than 10,000 reads were used for downstream analysis. The bioinformatics analysis was conducted using R packages. Permutational multivariate analysis of variance (PERMANOVA) was performed with vegan v2.6.4 [ 23 ]. PCoA (Principal coordinates analysis) and alpha-diversity were performed with phyloseq v1.42.0 [ 24 ] and ggplot2 v. 3.3.6 ( https://github.com/tidyverse/ggplot2 ). Linear discriminant analysis Effect Size (LefSe) was conducted using microbiomeMarker v1.4.0 [ 25 ] with default parameters. Models for the prediction of postprandial glucose response We used two measures of PPGR characteristics: iAUC120 [ 7 ] and the peak glucose level within 120 minutes after the meal start (GLUmax, mmol/L). The latter indicator was chosen because the recommended timing of blood glucose self-monitoring for pregnant women is established in the time interval when glycemic levels are highest. The peak glycemic level in pregnant women with diabetes mellitus is reached 45–75 minutes after a meal, which is the reason for the recommendation to measure glycemia one hour after a meal [ 26 ]. However, the peak blood glucose (BG) level is less sensitive to inaccurate logging of meal start time than 1-hr postprandial BG. Outliers in the target variable were removed using the Tukey’s Interquartile Range method [ 27 ] resulting in the final dataset of 2,633 meals with PPGRs for GLUmax prediction model and 2,628 meals for iAUC120 prediction. We used the gradient boosting algorithm LightGBM [ 28 ] to predict both indices and improved its performance with Optuna hyperparameter optimization [ 29 ]. Within Optuna, we adopted the Tree-structured Parzen Estimator (TPESampler) for sampling within the hyperparameter space, and the Asynchronous Successive Halving Algorithm (SuccessiveHalvingPruner) was implemented to eliminate underperforming trials efficiently. The optimal configuration of hyperparameters that emerged from our analysis included a “num_boost_round” of 4700, a “learning_rate” of 0.0015, “max_depth” of 11, “num_leaves” of 30, “min_sum_hessian_in_leaf” at 0.12, “bagging_fraction” of 0.55, “bagging_freq” of 10, “feature_fraction” of 0.4, “lambda_l1” at 0.006 and “lambda_l2” at 0.007. The data was divided into training and test sets with a ratio of 70:30, ensuring that records from the same patient were only included in one set to prevent data leakage and potential bias in performance metrics. To enhance the robustness of the training process, we employed 3-fold cross-validation. The final metrics, including mean absolute error (MAE), mean squared error (MSE), root mean squared error (RMSE), Pearson’s correlation coefficient (R), and the coefficient of determination (R 2 ), were calculated on the held-out test sample in the Python scikit-learn library. Other statistical analyses . To describe the patients in our data frame, we employed the bootstrap hypothesis test [ 30 ]. The main idea behind this method is to repeatedly draw random subsamples of the data with replacement in order to estimate the distribution of the test statistic and make decisions about the significance of differences. The bootstrap test does not require any assumptions about the underlying distribution of the original data, making it more robust compared to parametric tests such as the t-test. Feature selection and evaluation of input parameters We selected 164 features as model inputs, including features characterizing meal content, anthropometric measures, gynecological data, blood tests results, CGM-derived features, lifestyle questionnaire data, and genetic and microbiome features (Supplementary Table S1 ). In order to avoid model overfitting, we used several approaches to decrease the number of input variables. From the original lifestyle questionnaire characterizing the consumption of certain product groups and physical activity, described elsewhere [ 11 , 31 ], we selected the parameters with significant Spearman correlations with iAUC120 and/or GLUmax (Supplementary Table S2). Microbial features were selected based on the results of LefSe analysis. For this purpose, all participants were divided into two groups based on the average levels of PPGR indices: group 1 – below the median and group 2 – equal or above the median for the group. The relative abundances (RA) of bacterial taxa differentially enriched in these groups were used as input variables. Among genetic factors, we selected rs10830963 and rs1387153 variants in MTNR1B previously shown by our group to be associated with the results of OGTT in pregnant women [ 31 ]. SHAP methods were utilized for enhancing model interpretability [ 32 ]. SHAP values were computed in two ways. First, calculations were made individually for each feature to denote the average alteration in the model’s output when conditioning on that specific feature. Second, the additive nature of SHAP values was employed to assess the impact of various feature groups on the model. Availability of Data The datasets generated and analyzed in the current study are available in a github repository [ https://github.com/artemisak/MicrobesAndGlucouseAnalysis?tab=readme-ov-file ]. Results In total, 152 participants were recruited to the study. After exclusion of 3 women who did not provide CGM data, 2 women with antibiotics intake during the study period, 34 women with inaccurate food diaries, 2 women with less than 6 meals left after filtering, and 6 microbiota samples with low read count (<10,000 reads), 105 participants (77 women with GDM and 28 healthy pregnant women) were included in the final analysis (Fig. 1). The characteristics of the participants are in Table 1. Women with GDM did not differ from the control group in terms of age and gestational age upon initiation of continuous glucose monitoring. Patients with GDM had higher body mass index (BMI) before pregnancy. As expected, healthy pregnant women had lower plasma glucose levels during OGTT and hemoglobin A1C (HbA1C) upon inclusion into the study. Patients with GDM consumed lower amounts of carbohydrates (28.4 ± 10.9 vs 36.6 ± 10.8 g) and higher amounts of proteins (17.0 ± 5.2 vs 13.8 ± 2.9 g) per meal compared to healthy women (Table 1). Presumably due to this fact, iAUC120 and GLUmax levels did not significantly differ between the groups and even tended to be lower in women with GDM compared to their healthy counterparts who were not dieting (0.52 ± 0.29 vs 0.63 ± 0.28 and 6.2 ± 0.6 vs 6.4 ± 0.6 mmol/L, respectively) (Table 1). For comprehensive details on lifestyle assessments and baseline blood tests, please refer to supplementary Table S3. Microbial features in women with higher and lower PPGRs As there was no difference in the levels of GLUmax and iAUC120 between women with and without GDM during CGM, we combined their data for selection of microbial features associated with higher and lower PPGRs. The medians for iAUC120 and GLUmax in the cohort were 0.527 and 6.254 mmol/L, respectively. Participants with median PPGR indices (iAUC120 or GLUmax, respectively) below these numbers were considered to have lower PPGRs, and those with median PPGR indices equal to or above the cohort median comprised the subgroup with higher PPGRs. Linear discriminant analysis revealed 18 bacterial taxa exhibiting significantly higher scores in the subgroup of women with higher iAUC120 and 21 bacterial taxa with higher scores in the subgroup with lower iAUC120, P < 0.05 for all (Fig. 2). Bacterial taxa displaying notably higher scores in women with higher iAUC120 included Dorea (Lachnospiraceae), Fusicatenibacter (Lachnospiraceae), Ruminococcus torques group (Oscillospiraceae), Prevotella 9 (Bacteroidia, Prevotellaceae), Coprococcus comes (Lachnospiraceae), Roseburia (Lachnospiraceae), “Lachnoclostridium edouardi” (Lachnospiraceae), Marvinbryantia (Lachnospiraceae), Anaerobutyricum hallii (basonym: Eubacterium hallii ) (Lachnospiraceae), Colidextribacter (Bacillota). Taxa with a higher score in the subgroup with lower iAUC120 included Oscillospiraceae UCG-002 , Muribaculaceae (Bacteroidota, Bacteroidia), Ruminococcus champanellensis (Oscillospiraceae), Christensenellaceae R-7 group (Clostridia), Parabacteroides distasonis (Bacteroidia, Tannerellaceae), Blautia (Lachnospiraceae), Sellimonas (Lachnospiraceae), Eisenbergiella tayi (Lachnospiraceae) and Bilophila wadsworthia (Deltaproteobacteria, Desulfovibrionaceae) (Fig. 2). All bacterial taxa distinguished by LefSe were included to input variables for creation of PPGR prediction models. When comparing women with higher and lower GLUmax, 7 taxa were enriched in the subgroup with higher GLUmax, including Clostridia UCG 014 and “Lachnoclostridium” (Lachnospiraceae), and 8 taxa were enriched in the subgroup with lower GLUmax, including Methanosphaera (Methanobacteria), Lachnospira eligens (basonym: Eubacterium eligens ) (Lachnospiraceae), Butyricicoccus faecihominis (Oscillospiraceae), Intestinibacter bartlettii (Clostridia, Peptostreptococcaceae), Sellimonas (Lachnospiraceae), E. tayi (Lachnospiraceae), Christensenellaceae R-7 group (Clostridia) (Fig. 3). Predicting individual postprandial responses We assessed the overall extent to which different combinations of input variables predict personal postprandial responses: iAUC120 and GLUmax. A total of 750 days of concurrent CGM usage and meal logging resulted in 3,514 meals to be analyzed with their PPGRs. Meal filtering (see RESEARCH DESIGN AND METHODS, Meal preprocessing) reduced the dataset to 2,706 meals. After removal of outliers in the target variable, the final dataset comprised 2,633 meals with PPGRs for GLUmax prediction model and 2,628 meals for iAUC120 prediction. Prediction models for both indices were developed utilizing gradient boosting algorithms, with the following combinations of input variables: 1) only carbohydrate content of the meal (carbs); 2) clinically available parameters (anthropometric, biochemical, lifestyle questionnaire, meal content and meal context, CGM data); 3) model 2 parameters + microbial features (the full model). For the full list of features please see the Supplementary Table 1. Validation of the model was performed using a three-fold cross-validation scheme (see RESEARCH DESIGN AND METHODS). In the context of predicting GLUmax, the first model that relied solely on the amount of carbohydrates in a meal demonstrated the lowest correlation with PPGRs (R = 0.35) and accounted for only 5% of the variation in glycemic response (Fig. 4A). The second model based on clinically available parameters achieved a significantly higher correlation (R = 0.62) and explained 34 % of variance (Fig. 4B). Adding microbiome features (Fig. 4C) further increased the predictive ability with an R of 0.66 and a coefficient of determination of 42%. Likewise, in the prediction of iAUC120, a model based solely on the carbohydrate content of meals demonstrated a relatively weak correlation (R = 0.51) and explained only 26% of the variation in glycemic response (Fig. 5A). The addition of parameter groups, as described above, resulted in an increase in correlation between CGM-measured and predicted values (R = 0.71, R 2 = 0.50). Addition of microbial features to this model slightly increased the accuracy of prediction (R = 0.72, R 2 = 0.52) (Fig. 5B-5C). Because the performance of a model can also be affected by non-linear relationships between measured and predicted values, we also assessed MAE, MSE and RMSE for the models with higher performance (models 2-3, Table 2). As shown in Table 2, adding microbial features decreased MAE, MSE and RMSE for GLUmax prediction, but did not influence these parameters characterizing prediction of iAUC120. Exploring factors influencing the prediction of postprandial glycemic responses Following the examination of different models predicting PPGRs, our subsequent focus was on understanding the individual factors influencing prediction accuracy, including microbial features and other parameters comprising the full model. For this purpose, we conducted feature attribution analysis employing SHAP [19]. The features that exerted the greatest influence on iAUC120 prediction, as indicated by the highest mean absolute SHAP value, encompassed the carbohydrate content of the meal, glycemic load of the meal, amount of starch in the meal, and CGM-derived parameters characterizing glucose levels preceding the meal (glucose level 10 minutes before meal and glucose rise from 240 minutes before the meal to meal start) (Fig. 6A). The most influential parameters for the prediction of GLUmax were the glucose levels at the onset of the meal (GLU0), the carbohydrate content of the meal, glycemic load of the meal, RA of I. bartlettii , and the amount of protein consumed up to 6 hours before the meal (Fig. 6B). Among the 20 most influential parameters for the prediction of iAUC120 or GLUmax, the algorithm selected the RA of the following bacterial taxa: I. bartlettii, “L. edouardi”, B. faecihominis (for iAUC120), and I. bartlettii, L. eligens (basonym: Eubacterium eligens ), and R. champanellensis (for GLUmax) (Fig. 6 A,B). Notably, I. bartlettii ranked fourth among influential parameters for the prediction of GLUmax and was selected by the algorithm among the top parameters both for iAUC120 and for GLUmax prediction. In order to assess the cumulative influence of microbial composition and other feature groups on the model, we summed the SHAP values of associated features (Fig. 7). These examinations revealed that the meal composition had the most significant effect on prediction of iAUC120, followed by CGM-derived data, meal context, and microbial composition (Fig. 8). Оn the contrary, for the prediction of GLUmax the main predictor group was the CGM-derived data, followed by meal composition, meal context, and microbial data also taking the fourth place (Fig. 7). Discussion Recently, a high interpersonal difference in PPGRs was revealed, and the gut microbiota has been shown to be a factor underlying this variability. Furthermore, the gut microbiota has been used to enhance the accuracy of PPGR prediction in healthy volunteers and individuals with type 1 diabetes [ 7 – 9 ]. However, to our knowledge, no study has explored the impact of the gut microbiome on PPGRs in pregnant women with or without GDM. Pregnancy is characterized by substantial alterations in all types of metabolism as well as by dynamic changes in gut microbial composition [ 33 ]. This fact, in line with the importance of achieving target postprandial glucose levels for improved pregnancy outcomes, underscores the importance of comprehensive evaluation of factors underlying PPGRs in pregnant women for the construction of more accurate PPGR prediction models for personalized dietary advice in this specific population. Our study has shown that microbiome features are among the top 10 most impactful individual parameters on the PPGR prediction in pregnant women with GDM. The cumulative influence of microbial composition was the fourth among the ten most impactful feature groups following CGM-derived data, meal composition and meal context for the prediction of GLUmax and iAUC120. Of note, adding microbiome features increased the predictive ability of the model with the increment of the coefficient of determination from 34–42%. These findings underscore the potential role of the microbiome in the regulation of glycemic control even though the addition of microbiome features had lower impact on the accuracy metrics of iAUC120 prediction. The latter could be a consequence of relatively high accuracy of iAUC120 prediction even before microbiome addition. The postprandial glycemic predictions in our study (with R = 0.72 for the model predicting iAUC120) closely resembled those documented by Zeevi et al. (with an R value of 0.70) in healthy subjects [ 7 ]. The performance of our models was also superior to that of the models developed by Shilo et al. for patients with type 1 diabetes (R = 0.72 vs 0.59 for iAUC120 prediction and R = 0.66 vs 0.61 for GLUmax prediction) [ 9 ]. Furthermore, Shilo et al. included PPGRs from the same patient in training and validation datasets, while in our study, we separated participants between datasets so PPGRs of a participant from a training dataset could not be analyzed in the test or validation datasets. If Shilo et al. followed the same protocol, the difference in model performance might be even more pronounced. However, type 1 diabetes patients have much greater glucose variability and excursions, thus complicating the task of accurate PPGR prediction in this group of patients. In the biggest study of PPGRs in healthy individuals, to date, Berry et al. obtained the highest accuracy of iAUC120 prediction with R = 0.75 in the validation cohort [ 8 ]. They likely reached the maximum accuracy which could be anticipated judging by the correlation between PPGRs to repeated standard meals (intraindividual variability) of 0.7–0.77 reported by Zeevi et al. [ 7 ]. A potential reason for the lower performance of our model is almost 10-fold smaller sample size and a lower number of included genetic variants compared to the study by Berry et al. [ 8 ]. Compared to our previous study on PPGRs prediction without implementing microbiome data [ 11 ], our current model exhibited only a slight increase of R (0.72 vs 0.7). However, for the previous model, we had a larger dataset (3,240 records of meals and corresponding PPGRs) making direct comparisons inappropriate. Concerning the gut microbiome, our study identified bacterial taxa differentially enriched in pregnant women with higher and lower mean PPGR. According to LefSe analyses, most taxa belong to Lachnospiraceae and Oscillospiraceae, and some families of Bacteroidia. These families are represented by the most functionally active bacteria involved in dietary fiber degradation and short-chain fatty acid (SCFA) biosynthesis [ 34 ]. SCFAs, especially butyrate, are generally considered beneficial metabolites that reduce the risk of GDM. However, excess SCFAs can activate gluconeogenesis, leading to hyperglycemia and insulin resistance [ 35 ]. Even though, in general, Lachnospiraceae and Oscillospiraceae are considered useful symbionts that interact beneficially with the host, among them, some taxa carry a “dual” function. For example, Anaerobutyricum hallii and some Blautia species, are considered pathobionts that can cause harm to the host [ 35 ]. Further, among the taxa with higher abundance in women with higher iAUC120 or GLUmax, several are of interest for discussion as taxa potentially contributing to GDM pathogenesis. Prevotella 9 is now characterized as the new genus Segatella with the type species Segatella copri (basonym: Prevotella copri ). Although S. copri is considered to be associated with health [ 36 ], a significant positive association between increases in Prevotella 9 and higher GDM risk was identified [ 37 ], and an increased abundance of Prevotella was reported in GDM patients [ 38 ]. Coprococcus comes is a butyrate producer [ 34 ], usually considered beneficial. However, in the FINRISK-2002 cohort, the strongest association with higher statin-associated new-onset type 2 diabetes risk was observed for C. comes [ 39 ], which aligns with our results. A possible explanation for this may be the ability of C. comes to produce the highest butyrate levels [ 34 ], which can lead to its excess. A. hallii (basonym: Eubacterium hallii ) is also associated with health [ 36 ]. However, GDM patients who failed to control glycemic levels were characterized by increased A. hallii [ 40 ], which corresponds to the results of our study. In women with lower iAUC120 or GLUmax, some taxa also had higher abundance, conversely suggesting a protective effect against higher PPGR. Oscillospiraceae UCG-002 , previously Ruminococcaceae UCG-002 , was more abundant in the normal glucose tolerance group than in GDM. Previous research found it was reduced in early pregnancy in women with subsequent GDM and was negatively correlated with fasting blood glucose levels [ 41 ]. Oscillospiraceae UCG-002 was also negatively associated with the homeostasis model assessment of insulin resistance (HOMA-IR) index and served as a marker of intestinal phytoestrogen enterolactone production [ 42 ]. Christensenellaceae R-7 group is a beneficial genus: elevated abundance was associated with reduced visceral adipose tissue and a healthier metabolic profile [ 43 , 44 ]. Parabacteroides distasonis may protect against inflammation and obesity; however, increased abundance of P. distasonis was previously reported in GDM [ 38 ]. Sellimonas is an acetate producer, associated with a reduced type 2 diabetes risk [ 45 ] and has been linked to low polycystic ovary syndrome (PCOS) risk [ 46 ]. Eisenbergiella tayi produces butyrate, lactate, acetate, and succinate and is thought to be potentially beneficial. However, E. tayi was associated with the disease state [ 36 ]. Women who developed GDM showed a significantly higher abundance of Eisenbergiella in early pregnancy, and Eisenbergiella was also positively correlated with fasting blood glucose levels [ 41 ], which contradicts our results. Bilophila wadsworthia is associated with the metabolism of fatty acid esters of hydroxy fatty acids, which improves glucose homeostasis, stimulates insulin sensitivity, and has anti-inflammatory effects [ 47 ]. After including bacterial taxa distinguished by LefSe as input variables for creation of PPGR prediction models, microbiome features were categorized as either advantageous or disadvantageous. As the RA of these taxa increased, the algorithm projected a decrease or increase in postprandial glucose response, respectively. Among bacterial features, the greatest contribution to iAUC120 prediction was made by the RAs of I. bartlettii , B. faecihominis , and “L. edouardi” . The most impactful bacterial features for the prediction of GLUmax were I. bartlettii, R. champanellensis , and L. eligens . A higher abundance of these bacteria was associated with lower PPGRs. Notably, I. bartlettii was selected by the algorithm both for the prediction of iAUC120 and GLUmax among the top 20 parameters. I. bartlettii can produce indoleacetic and phenylacetic acids, acetate, isovalerate, and isobutyrate. Due to the latter's production, Intestinibacter might be beneficial to host lipid and glucose metabolism and intestinal barrier integrity, which may explain the inverse association of Intestinibacter with diabetes [ 48 ]. L. eligens produces butyrate, acetate, and lactate, and promotes the production of the anti-inflammatory cytokine IL-10. L. eligens was reduced in early pregnancy in women with subsequent GDM [ 41 ]. The abundance of L. eligens was significantly higher in the healthy controls than in the obese individuals [ 49 ]. Additionally, L. eligens was positively associated with adherence to a Mediterranean diet [ 50 ]. B. faecihominis is a butyrate producer and was included in the stool-derived microbial ecosystem therapeutics to combat Clostridioides difficile infection as a beneficial bacterium [ 51 ]. R. champanellensis is a cellulose-degrading bacterium [ 52 ]. The strongest association with lower statin-associated new-onset type 2 diaberes risk was observed for R. champanellensis [ 39 ]. “L. edouardi” was associated with an increased risk of GDM [ 37 ] and heightened type 2 diabetes risk [ 45 ]. The limitation of our study is a relatively small sample size. Further studies in other cohorts and populations of pregnant women are needed to confirm our findings concerning certain bacterial taxa associated with PPGRs. There is further room for improvement, such as conducting more comprehensive assessments of contextual factors than those employed in the current study. For example, including data on physical activity preceding meals and integrating extensive 'omics' data could improve the predictive capacity of these algorithms. It is essential to delve deeper into understanding the functional roles of bacterial taxa that were the most influential for PPGR prediction in our study. The insights gained from this data could pave the way for the future advancement of probiotic or autoprobiotic therapies aimed at enhancing glycemic regulation. Probiotics with metabolic effects that target functionally active bacteria, predominantly belonging to Clostridia and Bacterodia, which play a key role in maintaining the balance of the intestinal microbiota, seem promising [ 53 ]. Conclusion Our study highlights the emerging role of the gut microbiota in the interpersonal variability of PPGRs. While previous research extensively utilized microbiota for PPGR prediction in healthy individuals and those with type 1 diabetes, our study fills an important gap by examining its impact on PPGRs in pregnant women, particularly those with GDM. Our findings indicate that microbiome features rank among the top parameters influencing PPGR prediction in pregnant women with GDM. Specifically, certain bacterial taxa were identified as significantly associated with variations in PPGRs. Incorporating microbiome data enhanced the accuracy of our predictions, highlighting the potential of microbiota-based interventions for optimizing glycemic control. Declarations Funding This work was financially supported by the Ministry of Science and Higher Education of the Russian Federation (Agreement No. 075-15-2022-301) and the Ministry of Innovation, Science & Technology, Israel. The funders played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript. Author Contribution P.P., E.P., E.G., O.K. and E.S. were involved in the conception, design, and conduct of the study and the analysis and interpretation of the results. A.I., A.R., and S.S. designed and conducted the analyses, interpreted the results, and wrote the manuscript. A.A., E.V., A.T., I.N., Ek. S., A.E., El.S. and T. P. provided data and interpreted the results, M.K. and E.V. performed laboratory analysis and interpreted the results., S.Z., E.R., C.E., and S.T. performed fecal specimen processing, sample sequencing and interpreted the results. L.V. performed blood specimen processing, genotyping assays and interpreted the results, P.P., A.I., A.R., and S.S. wrote the first draft of the manuscript, and all authors edited, reviewed, and approved the final version of the manuscript. P.P. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Data Availability The datasets generated and analyzed in the current study are available in a github repository [https://github.com/artemisak/MicrobesAndGlucouseAnalysis?tab=readme-ov-file]. References Sacks DA, et al. Frequency of gestational diabetes mellitus at collaborating centers based on IADPSG consensus panel-recommended criteria. Diabetes Care 2012;35:526-8.https://doi.org/10.2337/dc11-1641 HAPO Study Cooperative Research Group; Metzger BE, et al. Hyperglycemia and adverse pregnancy outcomes. N Engl J Med 2008;358:1991-2002.https://doi.org/10.1056/NEJMoa0707943 Hajj NE, Schneider E, Lehnen H, Haaf T. Epigenetics and life-long consequences of an adverse nutritional and diabetic intrauterine environment. Reproduction 2014;148(6):R111-20.https://doi.org/10.1530/REP-14-0334 Popova P, Castorino K, Grineva EN, Kerr D. Gestational diabetes mellitus diagnosis and treatment goals: measurement and measures. Minerva Endocrinol 2016;41(4):421-32. PMID: 26824326. American Diabetes Association Professional Practice Committee. 15. Management of Diabetes in Pregnancy: Standards of Care in Diabetes-2024. Diabetes Care 2024;47(Suppl 1):S282-S294.https://doi.org/10.2337/dc24-S015 Koning SH, et al. Neonatal and obstetric outcomes in diet- and insulin-treated women with gestational diabetes mellitus: a retrospective study. BMC Endocr Disord 2016;16(1):52.https://doi.org/10.1186/s12902-016-0136-4 Zeevi D, et al. Personalized Nutrition by Prediction of Glycemic Responses. Cell 2015;163(5):1079-1094.https://doi.org/10.1016/j.cell.2015.11.001 Berry SE, et al. Human postprandial responses to food and potential for precision nutrition. Nat Med 2020;26(6):964-973.https://doi.org/10.1038/s41591-020-0934-0 Shilo S, et al. Prediction of Personal Glycemic Responses to Food for Individuals With Type 1 Diabetes Through Integration of Clinical and Microbial Data. Diabetes Care 2022;45(3):502-511.https://doi.org/10.2337/dc21-1048 Shilo S, et al. The gut microbiome of adults with type 1 diabetes and its association with the host glycemic control. Diabetes Care 2022;45(3):555-563.https://doi.org/10.2337/dc21-1656 Pustozerov EA, et al. Machine learning approach for postprandial blood glucose prediction in gestational diabetes mellitus. IEEE Access 2020;8:219308-219321.https://doi.org/10.1109/ACCESS.2020.3042483 Popova P, et al. A randomised, controlled study of different glycaemic targets during gestational diabetes treatment: Effect on the level of adipokines in cord blood and ANGPTL4 expression in human umbilical vein endothelial cells. Int J Endocrinology 2018;2018:6481658.https://doi.org/10.1155/2018/6481658 International Association of Diabetes and Pregnancy Study Groups Consensus Panel; Metzger BE, et al. International association of diabetes and pregnancy study groups recommendations on the diagnosis and classification of hyperglycemia in pregnancy. Diabetes Care 2010;33(3):676-682.https://doi.org/10.2337/dc09-1848 Pustozerov E, et al. Development and evaluation of a mobile personalized blood glucose prediction system for patients with gestational diabetes mellitus. JMIR Mhealth Uhealth 2018;6(1):e6.https://doi.org/10.2196/mhealth.9236 Pustozerov E, et al. The role of glycemic index and glycemic load in the development of real-time postprandial glycemic response prediction models for patients with gestational diabetes. Nutrients 2020;12(2):302.https://doi.org/10.3390/nu12020302 Caporaso JG, et al. Ultra-high-throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. ISME J 2012;6(8):1621-4.https://doi.org/10.1038/ismej.2012.8 Andrews, S. (2010). FastQC: A Quality Control Tool for High Throughput Sequence Data [Online]. Available online at: http://www.bioinformatics.babraham.ac.uk/projects/fastqc/ Ewels, P. et al. MultiQC: summarize analysis results for multiple tools and samples in a single report, Bioinformatics, Volume 32, Issue 19, October 2016, Pages 3047-3048, https://doi.org/10.1093/bioinformatics/btw354 Bolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics. 2014 Aug 1;30(15):2114-20. doi: 10.1093/bioinformatics/btu170. Callahan BJ, et al. (2016). "DADA2: High-resolution sample inference from Illumina amplicon data." Nature Methods, 13, 581-583. doi: 10.1038/nmeth.3869.https://doi.org/10.1038/nmeth.3869 Mirdita M, Steinegger M, Breitwieser F, Söding J, Levy Karin E. Fast and sensitive taxonomic assignment to metagenomic contigs. Bioinformatics. 2021 Sep 29;37(18):3029-3031. doi: 10.1093/bioinformatics/btab184. PMID: 33734313; PMCID: PMC8479651.https://doi.org/10.1093/bioinformatics/btab184 Quast, C. et al. The SILVA ribosomal RNA gene database project: Improved data processing and web-based tools. Nucleic Acids Res. 41, D590-D596 (2012).https://doi.org/10.1093/nar/gks1219 Oksanen, F.J., et al. (2017) Vegan: Community Ecology Package. R package Version 2.4-3. https://CRAN.R-project.org/package=vegan McMurdie, P. J., & Holmes, S. (2013). phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data. PloS one, 8(4), e61217. https://doi.org/10.1371/journal.pone.0061217 Yang Cao et al. MicrobiomeMarker: an R/Bioconductor package for microbiome marker identification and visualization, Bioinformatics, Volume 38, Issue 16, August 2022, Pages 4027-4029, https://doi.org/10.1093/bioinformatics/btac438 Bühling KJ, et al. Optimal timing for postprandial glucose measurement in pregnant women with diabetes and a non-diabetic pregnant population evaluated by the Continuous Glucose Monitoring System (CGMS). J Perinat Med 2005;33:125-31.https://doi.org/10.1515/JPM.2005.024 Tukey, J. W. (1977). Exploratory Data Analysis. Addison-Wesley Ke, G., et al. (2017). Lightgbm: A highly efficient gradient boostin decision tree. Advances in Neural Information Processing Systems, 30, 3146-3154. Akiba T, Sano S, Yanase T, Ohta T, Koyama M. 2019. Optuna: a next-generation hyperparameter optimization framework. KDD '19: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data; p. 2623-2631. doi: 10.1145/3292500.3330701. Efron, B. (1979). Bootstrap methods: Another look at the jackknife. The Annals of Statistics, 7(1), 1-26https://doi.org/10.1214/aos/1176344552 Popova PV, et al. Effect of gene-lifestyle interaction on gestational diabetes risk. Oncotarget 2017;8:112024-112035.https://doi.org/10.18632/oncotarget.22999 Lundberg SM, Erion G, Chen H, et al. From local explanations to global understanding with explainable AI for trees. Nat Mach Intell 2020;2:56-67.https://doi.org/10.1038/s42256-019-0138-9 Nuriel-Ohayon M, Neuman H, Koren O. Microbial Changes during Pregnancy, Birth, and Infancy. Front Microbiol 2016;7:1031.https://doi.org/10.3389/fmicb.2016.01031 Abdugheni R, et al. Metabolite profiling of human-originated Lachnospiraceae at the strain level. iMeta 2022;1(4):e58.https://doi.org/10.1002/imt2.58 Hu R, et al. Gut Microbiota and Critical Metabolites: Potential Target in Preventing Gestational Diabetes Mellitus? Microorganisms 2023;11(7):1725.https://doi.org/10.3390/microorganisms11071725 Giliberti R, Cavaliere S, Mauriello IE, Ercolini D, Pasolli E. Host phenotype classification from human microbiome data is mainly driven by the presence of microbial taxa. PLoS Comput Biol. 2022;18(4):e1010066.https://doi.org/10.1371/journal.pcbi.1010066 Wu X, et al. Investigating causal associations among gut microbiota, gut microbiota-derived metabolites, and gestational diabetes mellitus: a bidirectional Mendelian randomization study. Aging (Albany NY) 2023;15(16):8345-8366.https://doi.org/10.18632/aging.204973 Ponzo V, et al. Diet-Gut Microbiota Interactions and Gestational Diabetes Mellitus (GDM). Nutrients 2019;11(2):330.https://doi.org/10.3390/nu11020330 Koponen K, et al. Role of Gut Microbiota in Statin-Associated New-Onset Diabetes-A Cross-Sectional and Prospective Analysis of the FINRISK 2002 Cohort. Arterioscler Thromb Vasc Biol 2024;44(2):477-487.https://doi.org/10.1161/ATVBAHA.123.319458 Ye G, et al. The Gut Microbiota in Women Suffering from Gestational Diabetes Mellitus with the Failure of Glycemic Control by Lifestyle Modification. J Diabetes Res 2019;2019:6081248.https://doi.org/10.1155/2019/6081248 Ma S, et al. Alterations in Gut Microbiota of Gestational Diabetes Patients During the First Trimester of Pregnancy. Front Cell Infect Microbiol 2020;10:58.https://doi.org/10.3389/fcimb.2020.00058 Atzeni A, et al. Taxonomic and Functional Fecal Microbiota Signatures Associated With Insulin Resistance in Non-Diabetic Subjects With Overweight/Obesity Within the Frame of the PREDIMED-Plus Study. Front Endocrinol (Lausanne) 2022;13:804455.https://doi.org/10.3389/fendo.2022.804455 Tavella T, et al. Elevated gut microbiome abundance of Christensenellaceae, Porphyromonadaceae and Rikenellaceae is associated with reduced visceral adipose tissue and healthier metabolic profile in Italian elderly. Gut Microbes 2021;13(1):1-19.https://doi.org/10.1080/19490976.2021.1880221 Alcazar M, et al. Gut microbiota is associated with metabolic health in children with obesity. Clin Nutr 2022;41(8):1680-1688.https://doi.org/10.1016/j.clnu.2022.06.007 Song S, Zhang Q, Zhang L, Zhou X, Yu J. A two-sample bidirectional Mendelian randomization analysis investigates associations between gut microbiota and type 2 diabetes mellitus. Front Endocrinol (Lausanne) 2024;15:1313651.https://doi.org/10.3389/fendo.2024.1313651PMid:38495787 PMCid:PMC10940336 Liang Y, et al. Gut microbiome and reproductive endocrine diseases: a Mendelian randomization study. Front Endocrinol (Lausanne) 2023;14:1164186.https://doi.org/10.3389/fendo.2023.1164186 Folz J, et al. Human metabolome variation along the upper intestinal tract. Nat Metab 2023;5(5):777-788.https://doi.org/10.1038/s42255-023-00777-z Luo K, et al. Metabolic and inflammatory perturbation of diabetes associated gut dysbiosis in people living with and without HIV infection. Genome Med 2024;16(1):59.https://doi.org/10.1186/s13073-024-01336-1 Hu X, et al. Integrative metagenomic analysis reveals distinct gut microbial signatures related to obesity. BMC Microbiol 2024;24(1):119.https://doi.org/10.1186/s12866-024-03278-5 Ghosh TS, et al. Mediterranean diet intervention alters the gut microbiome in older people reducing frailty and improving health status: the NU-AGE 1-year dietary intervention across five European countries. Gut. 2020;69(7):1218-1228.https://doi.org/10.1136/gutjnl-2019-319654 Carlucci C, et al. Effects of defined gut microbial ecosystem components on virulence determinants of Clostridioides difficile. Sci Rep 2019;9(1):885.https://doi.org/10.1038/s41598-018-37547-x Moraïs S, et al. Cryptic diversity of cellulose-degrading gut bacteria in industrialized humans. Science 2024;383(6688):eadj9223.https://doi.org/10.1126/science.adj9223 Pokrotnieks J, Sitkin S. He who controls Clostridia and Bacteroidia controls the gut microbiome: The concept of targeted probiotics to restore the balance of keystone taxa in irritable bowel syndrome. Neurogastroenterol Motil. 2024:e14805.https://doi.org/10.1111/nmo.14805 Tables Table 1 – Characteristics of the participants GDM N=77 (mean±SD) Healthy pregnant women N=28 (mean±SD) p-value Age, years 32.2 ± 4.3 31.4 ± 4.7 0.392 Gestational age (weeks) 30.1± 4.0 30.0± 2.0 0.947 BMI before pregnancy (kg/m 2 ) 24.7 ± 5.2 22.1 ± 3.6 0.017 Fasting PG (mmol/L) 5.1 ± 0.5 4.4 ± 0.4 <0.001 1-h postload glucose (mmol/L) 9.5 ± 1.5 6.7 ± 1.6 <0.001 2-h postload glucose (mmol/L) 8.3 ± 1.7 6.0 ±1.2 <0.001 HbA1C, % 5.0 ± 0.4 4.8 ± 0.3 0.020 Real-time meal logging Days logged per participant 7.2 ± 0.6 7.0 ± 0.8 0.78 Meals logged per participant 26.5 ± 3.7 25.3 ± 4.7 0.72 Energy intake per meal (kcal) 334.8 ± 99.8 346.3 ± 95.0 0.597 Carbohydrate intake per meal (g) 28.4 ± 10.9 36.6 ± 10.8 0.001 Glycaemic load per meal (g) 14.8 ± 7.5 20.4 ± 6.8 0.001 Fat intake per meal (g) 16.4 ± 5.9 15.3 ± 5.9 0.413 Protein intake per meal (g) 17.0 ± 5.2 13.8 ± 2.9 0.003 CGM – derived indices Mean GLUmax, mmol/L 6.2 ± 0.6 6.4 ± 0.6 0.083 Mean iAUC120, mmol/L 0.52 ± 0.29 0.63 ± 0.28 0.092 Notes: Comparisons were performed using the bootstrap hypothesis test. Table 2 – Accuracy of the models for predicting PPGRs and peak postprandial glycaemic levels based on clinical data with and without the addition of bacterial features GLUmax PPGR (iAUC120) Without microbiome With microbiome Without microbiome With microbiome MAE 0.49 0.46 0.30 0.30 MSE 0.38 0.33 0.15 0.15 RMSE 0.62 0.57 0.39 0.39 Pearson R 0.62 0.66 0.71 0.72 R 2 0.34 0.42 0.50 0.52 Notes: CGM – continuous glucose monitoring, GLUmax – peak postprandial glycaemic level, PPGR – postprandial glycemic response, iAUC120 – incremental area under glucose curve during 120 minutes after meal, MAE – mean absolute error, MSE – mean squared error, RMSE – root mean squared error, R – coefficient of correlation for predicted and observed values, R 2 – coefficient of determination. Additional Declarations No competing interests reported. Supplementary Files Supplementarymaterialsnpg.docx Cite Share Download PDF Status: Published Journal Publication published 07 Feb, 2025 Read the published version in npj Biofilms and Microbiomes → Version 1 posted Editorial decision: Revision requested 04 Nov, 2024 Reviews received at journal 03 Nov, 2024 Reviewers agreed at journal 28 Oct, 2024 Reviews received at journal 09 Oct, 2024 Reviewers agreed at journal 30 Sep, 2024 Reviewers agreed at journal 26 Sep, 2024 Reviewers invited by journal 26 Aug, 2024 Editor assigned by journal 21 Aug, 2024 Submission checks completed at journal 16 Aug, 2024 First submitted to journal 02 Aug, 2024 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-4850670","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":347659755,"identity":"909601ca-1b94-4e75-961e-e94d5ab4022a","order_by":0,"name":"Polina V. Popova","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAz0lEQVRIiWNgGAWjYFAD9gYgYWBBihaeAyAtEqRokUgAk4QVmrP3HnzMU8OQZ3Dz+dUNPwokGPjbuxPwarHsOZdszHOModjgdk7ZzR6gwyTOnN2AV4vBjRwzyRlsDIkzZ+ek3eABajGQyCVGyz+glpln0m7+IVaLxMc2hsR+CfZjt4myBeQXg499DMX8PDlst2UMJHgI+gUUYg8SvjHksbEff3bzzR8bOf72XgIOY+ABUf8TgHFpAGLx4FWOpIUBqIX9AUHVo2AUjIJRMDIBAF7yRQJXy5e9AAAAAElFTkSuQmCC","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":true,"prefix":"","firstName":"Polina","middleName":"V.","lastName":"Popova","suffix":""},{"id":347659759,"identity":"1adaaf39-a16c-4dbb-bec7-6320e34e447b","order_by":1,"name":"Artem O. Isakov","email":"","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":false,"prefix":"","firstName":"Artem","middleName":"O.","lastName":"Isakov","suffix":""},{"id":347659764,"identity":"d732ecb2-8a5e-4cf5-b7a6-94fa65ab5f68","order_by":2,"name":"Anastasia N. Rusanova","email":"","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":false,"prefix":"","firstName":"Anastasia","middleName":"N.","lastName":"Rusanova","suffix":""},{"id":347659767,"identity":"33f0a920-a17b-4fa9-9c7c-071971c6b924","order_by":3,"name":"Stanislav I. Sitkin","email":"","orcid":"","institution":"Almazov National Medical Research Center","correspondingAuthor":false,"prefix":"","firstName":"Stanislav","middleName":"I.","lastName":"Sitkin","suffix":""},{"id":347659768,"identity":"0640d88a-26e2-4a27-9b51-f8fead3d2764","order_by":4,"name":"Anna D. Anopova","email":"","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":false,"prefix":"","firstName":"Anna","middleName":"D.","lastName":"Anopova","suffix":""},{"id":347659769,"identity":"50f96211-a456-48df-b706-d59df05c8c56","order_by":5,"name":"Elena A. Vasukova","email":"","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":false,"prefix":"","firstName":"Elena","middleName":"A.","lastName":"Vasukova","suffix":""},{"id":347659770,"identity":"d4fb9f01-b0f6-49ce-b8a2-16670b1f41bf","order_by":6,"name":"Alexandra S. Tkachuk","email":"","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":false,"prefix":"","firstName":"Alexandra","middleName":"S.","lastName":"Tkachuk","suffix":""},{"id":347659771,"identity":"57d6632a-8eb7-4a57-b654-1c1c10bae30e","order_by":7,"name":"Irina S. Nemikina","email":"","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":false,"prefix":"","firstName":"Irina","middleName":"S.","lastName":"Nemikina","suffix":""},{"id":347659772,"identity":"17441390-9b6b-45e6-9a10-0ff95fe97589","order_by":8,"name":"Elizaveta A. Stepanova","email":"","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":false,"prefix":"","firstName":"Elizaveta","middleName":"A.","lastName":"Stepanova","suffix":""},{"id":347659773,"identity":"dd4ce4e2-6607-4246-aac9-18932910bb97","order_by":9,"name":"Angelina I. Eriskovskaya","email":"","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":false,"prefix":"","firstName":"Angelina","middleName":"I.","lastName":"Eriskovskaya","suffix":""},{"id":347659774,"identity":"e1d67539-8478-4641-b4e3-c58f4c15e23a","order_by":10,"name":"Ekaterina A. Stepanova","email":"","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":false,"prefix":"","firstName":"Ekaterina","middleName":"A.","lastName":"Stepanova","suffix":""},{"id":347659775,"identity":"71674ef4-df1c-4b37-9bc5-0b5a65e859be","order_by":11,"name":"Evgenii A. Pustozerov","email":"","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":false,"prefix":"","firstName":"Evgenii","middleName":"A.","lastName":"Pustozerov","suffix":""},{"id":347659776,"identity":"ab057224-0f5e-4cda-ae4e-7e9815b5b1f7","order_by":12,"name":"Maria A. Kokina","email":"","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":false,"prefix":"","firstName":"Maria","middleName":"A.","lastName":"Kokina","suffix":""},{"id":347659777,"identity":"82847685-2b8c-41f4-96ed-81cb9acfa9b1","order_by":13,"name":"Elena Y. Vasilieva","email":"","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":false,"prefix":"","firstName":"Elena","middleName":"Y.","lastName":"Vasilieva","suffix":""},{"id":347659778,"identity":"66db84f8-c31f-4def-b587-d6eadf1e57b0","order_by":14,"name":"Lyudmila B. Vasilyeva","email":"","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":false,"prefix":"","firstName":"Lyudmila","middleName":"B.","lastName":"Vasilyeva","suffix":""},{"id":347659779,"identity":"783da1a3-56ba-4af9-ba73-28923439a65e","order_by":15,"name":"Soha Zgairy","email":"","orcid":"","institution":"Bar-Ilan University","correspondingAuthor":false,"prefix":"","firstName":"Soha","middleName":"","lastName":"Zgairy","suffix":""},{"id":347659780,"identity":"2a152366-de0c-40ae-8e07-897eae631f3a","order_by":16,"name":"Elad Rubin","email":"","orcid":"","institution":"Bar-Ilan University","correspondingAuthor":false,"prefix":"","firstName":"Elad","middleName":"","lastName":"Rubin","suffix":""},{"id":347659781,"identity":"0521e194-3df7-448b-9ac1-4a7ff05561da","order_by":17,"name":"Carmel Even","email":"","orcid":"","institution":"Bar-Ilan University","correspondingAuthor":false,"prefix":"","firstName":"Carmel","middleName":"","lastName":"Even","suffix":""},{"id":347659782,"identity":"5f05b273-9935-41fa-bc57-5e27f1477eb9","order_by":18,"name":"Sondra Turjeman","email":"","orcid":"","institution":"Bar-Ilan University","correspondingAuthor":false,"prefix":"","firstName":"Sondra","middleName":"","lastName":"Turjeman","suffix":""},{"id":347659783,"identity":"3c5350a6-5150-44d5-b3f1-cdce7eb09881","order_by":19,"name":"Tatiana M. Pervunina","email":"","orcid":"","institution":"Almazov National Medical Research Center","correspondingAuthor":false,"prefix":"","firstName":"Tatiana","middleName":"M.","lastName":"Pervunina","suffix":""},{"id":347659784,"identity":"68668fb8-d6f0-445d-ab98-839e0d25e310","order_by":20,"name":"Elena N. Grineva","email":"","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":false,"prefix":"","firstName":"Elena","middleName":"N.","lastName":"Grineva","suffix":""},{"id":347659785,"identity":"95000f90-d7e9-448c-942d-eefad1ea5a6a","order_by":21,"name":"Omry Koren","email":"","orcid":"","institution":"Bar-Ilan University","correspondingAuthor":false,"prefix":"","firstName":"Omry","middleName":"","lastName":"Koren","suffix":""},{"id":347659786,"identity":"72d26861-0c6b-43cc-876a-bf2780523657","order_by":22,"name":"Evgeny V. Shlyakhto","email":"","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":false,"prefix":"","firstName":"Evgeny","middleName":"V.","lastName":"Shlyakhto","suffix":""}],"badges":[],"createdAt":"2024-08-02 22:53:16","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4850670/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4850670/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41522-025-00650-9","type":"published","date":"2025-02-07T15:58:04+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":66689383,"identity":"306c48e5-ed3a-4b48-a8bb-772bbb794812","added_by":"auto","created_at":"2024-10-15 13:49:39","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":77894,"visible":true,"origin":"","legend":"\u003cp\u003eCohort selection\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-4850670/v1/73b28d085e40ee18066fe53c.png"},{"id":66690868,"identity":"f3cfeedf-a0e7-48fd-9f6e-1e7d23e4bf6b","added_by":"auto","created_at":"2024-10-15 13:57:39","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":254808,"visible":true,"origin":"","legend":"\u003cp\u003eResults of LefSe analysis comparing relative abundance (RA) of microbial features of participants with PPGR (iAUC120) below and above median\u003c/p\u003e\n\u003cp\u003eRed indicates higher RA in patients with iAUC120 equal to or above median and turquoise indicates higher RA in patients with iAUC120 below median, ranked by the effect size.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-4850670/v1/e02061a624c049c118e291ba.png"},{"id":66690867,"identity":"c550a371-ad63-4078-a1d4-50e2e6ea9a3f","added_by":"auto","created_at":"2024-10-15 13:57:39","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":97033,"visible":true,"origin":"","legend":"\u003cp\u003eResults of LefSe analysis comparing relative abundance (RA) of microbial features of participants with GLUmax below and above median\u003c/p\u003e\n\u003cp\u003eRed indicates higher RA in patients with GLUmax equal to or above median and turquoise indicates higher RA in patients with GLUmax below median, ranked by the effect size.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-4850670/v1/34d38d62f5df58586dd8b03b.png"},{"id":66688929,"identity":"44f4f772-153d-4d11-a710-9e38d53b2faa","added_by":"auto","created_at":"2024-10-15 13:41:39","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":183909,"visible":true,"origin":"","legend":"\u003cp\u003eThe results of peak postprandial blood glucose prediction (GLUmax) with the test set. X scale - CGM-measured values, Y scale – predicted values.\u003c/p\u003e\n\u003cp\u003eNotes: A. baseline model - solely carbohydrate content of the meal (carbs); B. the model based on clinically available parameters (anthropometric, biochemical, lifestyle questionnaire, meal content, meal context, CGM data); C. full model - clinically available parameters + microbial features.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-4850670/v1/9c853f0b2921c987319c4d40.png"},{"id":66688926,"identity":"8328e44e-9587-4bea-bacd-6ca6e0ce989c","added_by":"auto","created_at":"2024-10-15 13:41:39","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":156975,"visible":true,"origin":"","legend":"\u003cp\u003eThe results of iAUC120 prediction with the test set. X scale - CGM-measured values, Y scale – predicted values.\u003c/p\u003e\n\u003cp\u003eNotes: A. baseline model - solely carbohydrate content of the meal (carbs); B. the model based on clinically available parameters (anthropometric, biochemical, lifestyle questionnaire, meal content, meal context, CGM data); C. full model - clinically available parameters + microbial features.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-4850670/v1/f8a7f62ba6c2a0fca58ee5a6.png"},{"id":66688927,"identity":"83469331-da35-48db-81cd-7d9c3d6fc7ce","added_by":"auto","created_at":"2024-10-15 13:41:39","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":156370,"visible":true,"origin":"","legend":"\u003cp\u003eSignificance level of the 20 most impactful variables of the model for predicting\u003c/p\u003e\n\u003cp\u003eiAUC120 (A) and GLUmax (B) based on full clinical data with the addition of bacterial features.\u003c/p\u003e\n\u003cp\u003eNotes: Higher values of the feature are indicated by colors close to red, lower values by colors close to blue. If a point of a certain color is located on the left side of the central axis, the feature has a downward effect on the target variable; if the point is located on the right side, the effect will be the opposite. For example, lower values of GLU0 (the long blue tail on the left of Fig. 6B) correspond to lower values of the target variable (GLUMax).\u003c/p\u003e\n\u003cp\u003eGLUb- glucose level before meal initiation. Numbers near «GLUb» represent the minutes prior to meal initiation in which the measurement was obtained. For example, «GLUb10» represents the glucose level 10 min prior to the meal; Kcal - the energy value of the meal; COC - combined oral contraceptive use any time before pregnancy (1 – yes, 0 – no); Sausages 1 - frequency of consuming sausage products before pregnancy. For more detailed description of the input features please refer to the supplementary table 1.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-4850670/v1/5f42c340f0484990bf74cfda.png"},{"id":66688922,"identity":"ab9f8101-943e-419c-b49d-4d96b332fc3c","added_by":"auto","created_at":"2024-10-15 13:41:39","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":49083,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP values (linear scale, absolute values) of the groups of features for the prediction of iAUC120 (A) and GLUmax (B).\u003c/p\u003e\n\u003cp\u003eThe groups of features are presented as follows: «meal composition» includes the nutritional content of the meal, «cgm_data» includes glucose values obtained from CGM devices; «meal_context» includes the nutritional content of meals consumed up to 12 hour prior the index meal; «Microbiome» includes RA of bacteria detected from stool samples; «genetics» includes rs10830963 and rs1387153 variants in MTNR1B gene. The full description of the parameters included in each feature group is listed in the supplementary table 1.\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-4850670/v1/1ec8ed7261a9ee74673c4a73.png"},{"id":75930500,"identity":"868918bc-cdff-4cd3-b214-271f3cfce9c6","added_by":"auto","created_at":"2025-02-10 16:12:30","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1861100,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4850670/v1/64658ef5-0f60-4c5d-a629-b65c19a1e833.pdf"},{"id":66689386,"identity":"178c4e59-e7ea-4562-9b3d-0a1471a6f1ab","added_by":"auto","created_at":"2024-10-15 13:49:39","extension":"docx","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":55098,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarymaterialsnpg.docx","url":"https://assets-eu.researchsquare.com/files/rs-4850670/v1/a776adcc3d4803610028c9b1.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Personalized Prediction of Glycemic Responses to Food in Women with Gestational Diabetes: Gut Microbiota Matters","fulltext":[{"header":"Introduction","content":"\u003cp\u003eGestational diabetes mellitus (GDM) represents a prevalent condition, impacting a substantial portion, approximately up to 9-26%, of pregnancies\u0026nbsp;[1].\u0026nbsp;GDM can lead to pregnancy complications, including but not limited to\u0026nbsp;increased cesarean section rate,\u0026nbsp;birth trauma, nerve palsy, neonatal hypoglycemia in the short-term period\u0026nbsp;[2], and increased risks of obesity, type 2 diabetes, and cardiovascular diseases in both mothers and their offspring throughout life\u0026nbsp;[3]. It is crucial to uphold normal glycemic levels during pregnancy to mitigate adverse pregnancy outcomes and disrupt the cyclical transmission of predisposition to metabolic diseases across generations\u0026nbsp;[3,4].\u003c/p\u003e\n\u003cp\u003eThe\u0026nbsp;most common\u0026nbsp;treatment for GDM is diet and lifestyle modification, reportedly effective without adding medications for achieving glucose control in 70 - 85% of women [5]. However, according to real life observations, many women with GDM do not achieve target glucose levels and the rate of pregnancy complications remains high in women with GDM [6].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eNutritional studies and guidelines concerning GDM concentrate on average characteristics across populations\u0026nbsp;[5]. The historically prevailing approach to predicting postprandial glucose responses (PPGRs) to food involves relying on the carbohydrate content of the meal\u0026nbsp;[5], despite evidence suggesting its inadequacy as a predictor\u0026nbsp;[7]. Alternative methods include the glycemic index, which assesses the postprandial glucose response to a specific food, and the derived glycemic load\u0026nbsp;[5]. Attributing a singular postprandial glucose response (PPGR) to each food implies that the response is entirely inherent to the food itself. Yet, recent studies exploring interindividual variations in PPGRs have revealed significant variability in how different individuals respond to identical foods\u0026nbsp;[7, 8]. The distinctive PPGRs of an individual were shown to be influenced by their biological traits (such as gut microbiome composition and genetic variation) and lifestyle factors\u0026nbsp;[7, 8, 9].\u003c/p\u003e\n\u003cp\u003eSeveral studies\u0026nbsp;integrated gut microbial features into the models predicting PPGRs in healthy individuals [7, 8] or patients with type 1 diabetes [9]. However, only\u0026nbsp;a\u0026nbsp;few evaluated the impact of\u0026nbsp;microbiome on the prediction of PPGRs, through variance components analyses\u0026nbsp;[8] or the SHapley Additive exPlanation (SHAP) method [9], and only one study exclusively concentrated on\u0026nbsp;investigating the connections between bacterial species and the host glycemic regulation\u0026nbsp;[10]. Furthermore, none of these studies directly compared the accuracy metrics of the models before and after addition\u0026nbsp;of microbial data into the list of the input variables. They compared basic models including carbohydrates and premeal glucose level\u0026nbsp;with a full\u0026nbsp;model including multiple individual parameters apart from microbial features [7, 8, 9].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eMoreover, although there is mounting evidence regarding the regulatory functions of the microbiome in normal and impaired glycemic responses among non-pregnant individuals\u0026nbsp;[7 - 10], limited knowledge exists concerning the microbiome\u0026apos;s impact on PPGRs among pregnant women, both with and without GDM.\u0026nbsp;We have previously developed\u0026nbsp;PPGR prediction models based on multiple individual parameters without microbiome for pregnant women with and without GDM\u0026nbsp;[11]. The performance of our model in predicting\u0026nbsp;the incremental area under the glucose curve in the 2 hours after the meal (iAUC120)\u0026nbsp;(R =0.7) was\u0026nbsp;comparable to the model created by Zeevy et al. for healthy individuals\u0026nbsp;based on\u0026nbsp;individual parameters and microbiome\u0026nbsp;(R =0.7) [7]. However, our past model was inferior to the accuracy of the microbiome-based model subsequently developed by Berry et al.\u0026nbsp;for\u0026nbsp;iAUC120 prediction in\u0026nbsp;healthy individuals\u0026nbsp;(R =0.77) [8], leaving space for the improvement of our model, potentially with inclusion of microbiome data.\u0026nbsp;Another distinguishing feature of the Berry model, in comparison to both our previous model and the Zeevi model, was the incorporation of genetic factors. This likely enhanced the predictive accuracy, as genetics constituted the second most crucial parameter group after serum glycemic markers, as indicated by the proportion of variance explained (R\u003csup\u003e2\u003c/sup\u003e) [8].\u003c/p\u003e\n\u003cp\u003eThe aim of this study was to develop a prediction model for PPGR in pregnant women with GDM and to explore the impact of microbial data on the model\u0026rsquo;s performance. An accurate PPGR prediction model holds promise in optimizing personalized diet recommendations to improve glucose control and pregnancy outcomes in women with GDM. Simultaneously, the identification of a distinct gut microbial signature affecting PPGR, a secondary aim of this project, could serve as a basis for the development of potential therapeutic interventions.\u003c/p\u003e"},{"header":"Research Design and Methods","content":"\n\u003ch3\u003eStudy Design\u003c/h3\u003e\n\u003cp\u003eWe recruited pregnant women who participated in the randomized controlled trial \u0026ldquo;Genetic and Epigenetic Mechanisms of Developing Gestational Diabetes Mellitus and its Effects on the Fetus\u0026rdquo; (GEM-GDM), consented to be connected to a continuous glucose monitoring (CGM) system (CGMS) for at least 7 days, tracked information on food consumption in a designated mobile app, and provided stool samples. The parent GEM-GDM study aimed to compare different glycemic targets for women with GDM. It was registered at ClinicalTrials.gov (Identifier: NCT03610178) and its design is described elsewhere [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Briefly, gravidas with GDM were randomly assigned into two groups according to their glycemic goals: the first group had strict glycemic goals (\u0026lt;\u0026thinsp;5.1 mmol/L for fasting blood glucose (BG) and \u0026lt;\u0026thinsp;7.0 mmol/L for 1-hour postprandial BG), and the second group had less strict glycemic goals (\u0026lt;\u0026thinsp;5.3 mmol/L and \u0026lt;\u0026thinsp;7.8 mmol/L, respectively). For this study, the women from both groups were combined to create the GDM group. GDM was diagnosed using a single-step 75-g OGTT according to the recommendation of the International Association of Diabetes and Pregnancy Study Groups (IADPSG) [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eApart from women with GDM, we also included healthy pregnant women with normal values of plasma glucose during oral glucose tolerance test (OGTT) (controls).\u003c/p\u003e \u003cp\u003eAt study initiation, a physician acquired informed consent, recorded medical history and took anthropometric measurements (weight, height, waist circumference, blood pressure and heart-rate). Pregestational body mass index (BMI) was calculated by dividing self-reported pregestational weight (in kilograms) by the square of height (in meters). Blood tests, including fasting plasma glucose, lipid profile and HbA1c, were performed in the Almazov National Medical Research Centre laboratory. Participants filled out questionnaires concerning their lifestyle before and during pregnancy and were then connected to CGMS for 7\u0026ndash;14 days during which they tracked information on meal consumption in the proprietary mobile app DiaCompanion, as described elsewhere [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Each consumed food item was recorded by selecting it from a database created by the authors on the basis of reference books of the Russian Academy of Medical Sciences and the US Department of Agriculture (USDA) Food Composition Databases (Release 28) with the expansion of additional items by certified dietitians. The distinctive feature of this food database, in addition to a wide selection of foods (more than 5500 items), is the presence of glycemic index (GI). Each food item in the database was assigned a dietary GI [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. The CGM and meal-related data were processed using a previously described algorithm [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. The study was approved by the local ethics committee of the Almazov National Medical Research Centre, Russia (protocol. №119).\u003c/p\u003e \u003cdiv id=\"Sec2\" class=\"Section2\"\u003e \u003ch2\u003eInclusion and exclusion criteria\u003c/h2\u003e \u003cp\u003eParticipation in this study was optional for participants of the GEM-GDM trial.\u003c/p\u003e \u003cp\u003eThe GDM group included pregnant women with GDM and a gestational age of \u0026ge;\u0026thinsp;24 weeks at the start of CGM. The control group comprised pregnant women with normal glucose tolerance, confirmed by OGTT between 24 and 32 weeks of gestation.\u003c/p\u003e \u003cp\u003eIn addition to the inclusion criteria used in the GEM-GDM study, for this particular study, consent to be connected to a CGMS for 7 days and a capability to work with a mobile phone app for the recording of dietary intake in real time were required.\u003c/p\u003e \u003cp\u003eExclusion criteria included an active inflammatory or neoplastic disease, any known medical condition affecting glucose metabolism (with the exception of GDM), current insulin use, antibiotic usage 2 months prior to participation in the study, failure to provide a stool sample and submission of inaccurate food diaries through the app. Taking into account that accurate logging is crucial for analysis of PPGRs to food, a set of rules was formulated by the authors to filter negligently filled-in and misreported diaries: (1) more than 50% of the logged meals comprised of a single dish or a single dish with a single beverage; (2) the average amount of logged calories per day was less than 1000 kcal; (3) more than 50% of the logged weights of food items were rounded to the hundreds (excluding beverages); (4) the amount of logged snacks was less than 10% of all meal records [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Participants with misreported diaries were excluded.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eLifestyle questionnaire\u003c/h2\u003e \u003cp\u003eThe questionnaire comprised several sections covering various aspects: frequency of consuming staple items per week (such as fruits, pastries, skimmed dairy products, legumes, meat, sausage products, dried fruits, fish, whole-grain bread, sauces, vegetables, alcohol, sweet drinks, and coffee), levels of physical activity (daily walking duration categorized as \u0026lt;\u0026thinsp;30 min/day, 30\u0026ndash;60 min/day, or \u0026gt;\u0026thinsp;60 min/day; daily frequency of stair climbing categorized as \u0026lt;\u0026thinsp;4 flights/day, 4\u0026ndash;16 flights/day, or \u0026gt;\u0026thinsp;16 flights/day; frequency of engaging in sports activities categorized as \u0026lt;\u0026thinsp;2 days/week, 2\u0026ndash;3 days/week, or \u0026gt;\u0026thinsp;3 days/week), and smoking habits before and during pregnancy. Each section of the questionnaire was structured in a semi-quantitative manner. This questionnaire has been previously documented [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. For the description of the parameters from lifestyle questionnaire included in the final dataset, please refer to Supplementary Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cb\u003eBlood samples\u003c/b\u003e were collected by a certified nurse after 8\u0026ndash;12 hours of fasting. The blood panel included measurement of glycosylated hemoglobin (HbA1c%), plasma glucose, total and HDL cholesterol, triglycerides, insulin, leptin. and fructosamine levels in the central lab of the Almazov National Medical Research Centre. Plasma glucose concentration was determined by the glucose oxidase method in fresh plasma samples. HbA1c was measured in fresh whole blood samples using high performance liquid chromatography (HPLC) (D10 HbA1c). Blood for genotyping of pregnant women and serum for other biochemical analysis were stored at -80\u0026deg;C until the analysis. Serum fasting insulin levels were measured using the electrochemiluminescence immunoassay (Roche Diagnostics, GmbH, Germany). The homeostatic model assessment (HOMA) index was calculated using the following formula: fasting serum insulin (m IU/L) \u0026times; fasting plasma glucose (mmol/L)/(22.5) as an insulin resistance indicator. Total cholesterol, HDL-C, LDL-C, VLDL-C, and triglyceride levels were measured utilizing enzymatic colorimetric methods with diagnostic reagent system designed for the Cobas Integra Autoanalyzer. Serum leptin levels were measured using an enzyme-linked immunosorbent assay (ELISA) as recommended by the manufacturer (Diagnostics Biochem Canada Inc., Canada).\u003c/p\u003e \u003cp\u003e\u003cb\u003eContinuous glucose monitoring (CGM)\u003c/b\u003e was conducted using the iPro2TM system from Medtronic, MN, USA. This system utilizes EnliteTM sensors placed subcutaneously to measure interstitial glucose levels. To align CGM readings with blood glucose levels, participants also utilized finger-prick measurements with the Accu Chek Performa from Roche, Germany. Participants were instructed to perform four daily blood glucose measurements. To enhance accuracy, participants were specifically asked to measure blood glucose levels before meals, following recommendations outlined in the iPro2 manual. Calibration of CGM measurements was performed using the CareLink online software from Medtronic, following the guidelines provided in the iPro2 manual.\u003c/p\u003e \u003cp\u003e \u003cb\u003eFood diary tracking\u003c/b\u003e was facilitated through our proprietary mobile app DiaCompanion. Each consumed food item was recorded by selecting it from a database created by the authors on the basis of reference books of the Russian Academy of Medical Sciences and the US Department of Agriculture (USDA) Food Composition Databases (Release 28) with the expansion of additional items by certified dietitians. The distinctive feature of this food database, in addition to a wide selection of foods (more than 5,500 items), is the presence of glycemic index (GI). Each food item in the database was assigned a dietary GI [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Participants were instructed to meticulously log their daily activities using this platform. They were required to document precise details, including the components and weights of each meal, sleep and wake-up times. Participants were informed of the importance of accurate logging, particularly emphasizing the correct timing of meal logging and accurate recording of food components. Research physicians conducted weekly reviews of each participant's loggings. Any uncertainties in the logs were addressed directly with the participants.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eMeal preprocessing\u003c/h2\u003e \u003cp\u003eBefore a meal and corresponding PPGR were added to the dataset for model training, the following filters were implemented (mainly to exclude recordings with incorrect timing): (1) a meal followed by a subsequent meal less than 60 minutes after its start; (2) a meal on the peak of a CGM-curve: an increase in glucose levels by more than 1 mmol/L during an hour preceding the index meal; (3) a meal on the falling edge of a CGM peak; and (4) a meal with inadequately low PPGR (iAUC120\u0026thinsp;\u0026lt;\u0026thinsp;=\u0026thinsp;0.3 mmol/L\u0026lowast;h) to a considerable amount of carbohydrates (\u0026gt;\u0026thinsp;40 g) [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Glucose level at baseline was considered as the lowest glucose value within \u0026plusmn;\u0026thinsp;15 min from self-reporting of the meal in the app.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eDNA and genotyping of blood samples\u003c/h2\u003e \u003cp\u003eGenomic DNA was extracted from blood samples using the FlexiGene DNA Kit from Qiagen, (Hilden, Germany). Genotyping of the following variants: HKDC1 (rs10762264), MTNR1B (rs10830963 and rs1387153), GCK (rs1799884), KCNJ11 (rs5219), IGF2BP2 (rs4402960), TCF7L2 (rs7903146), CDKAL1 (rs7754840), FTO (rs9939609), and IRS1 (rs1801278), was conducted through real-time PCR utilizing custom kits from Applied Biosystems, based in the USA. The procedures recommended by the manufacturer were followed meticulously. Each primer tube contained a concentrated mixture of SNP Genotyping Assay Mix, comprising polymorphism-specific direct and reverse primers, along with two TaqMan MGB probes: one tagged with VIC dye for allele 1 identification and the other tagged with FAM dye for allele 2 identification. Following replication of 10% of the samples, the discordance rate was determined to be less than 0.1%.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eMicrobiome: DNA extraction\u003c/h2\u003e \u003cp\u003eDNA was extracted from all collected samples using the PowerSoil DNA Isolation Kit (MO BIO, Carlsbad, CA, USA) according to the manufacturer's instructions and following a 2 min bead beating step (BioSpec, Bartlesville, OK, USA). Next, the variable V4 region was PCR-amplified using the 515F and 806R barcoded primers following the Earth Microbiome Project protocol [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. Each PCR reaction contained 25\u0026micro;l with ~\u0026thinsp;40 ng/\u0026micro;l of DNA, 2 \u0026micro;l 515F (forward, 10\u0026micro;M) primer, 2 \u0026micro;l 806R (reverse, 10\u0026micro;M) primer, and 25 \u0026micro;l PrimeSTAR Max PCR Readymix (Takara, Mountain view, CA, USA). PCR conditions were as follows: 30 cycles of denaturation at 98\u0026deg;C for 10 sec, annealing at 55\u0026deg;C for 5 sec, and extension at 72\u0026deg;C for 20 sec, followed by a final elongation at 72\u0026deg;C for 1 min. Amplicons were purified using AMPure magnetic beads (Beckman Coulter, Indianapolis, IN, USA) and quantified using the Picogreen dsDNA quantitation kit (Thermofisher, Waltham, MA, USA). Equimolar amounts of DNA from individual samples were pooled and sequenced using the Illumina MiSeq platform at the Genomic Center at the Bar-Ilan University, Azrieli Faculty of Medicine. Appropriate negative and positive controls were included at all stages of analysis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eBioinformatics and microbiome analysis\u003c/h2\u003e \u003cp\u003eThe quality of raw reads was assessed with FastQC v. 0.11.9 [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e] and MultiQC v. 1.14 [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Reads were trimmed and filtered with Trimmomatic v. 0.39 [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e] (SE -phred 33 HEADCROP 31 ILLUMINACLIP:2:30:10 SLIDINGWINDOW:4:15 MINLEN:150). The remaining reads were processed with the DADA2 pipeline v. 3.6.2. [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e], including additional trimming, denoising, and error correction. The derived sequences - amplicon sequence variants (ASVs) were clustered using MMseqs2 v. 13.45111 [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e] (identity 99%, coverage 80%). The resultant representative sequences were treated as operative taxonomic units (OTUs). We clustered ASVs to OTUs to reduce the number of sequencing errors inherent in ASVs and avoid false diversity. The OTUs were returned to DADA2 for taxonomy assignment with SILVA SSU database v.138.1 [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Sequences classified as eukaryotes were removed. Only samples containing more than 10,000 reads were used for downstream analysis. The bioinformatics analysis was conducted using R packages. Permutational multivariate analysis of variance (PERMANOVA) was performed with vegan v2.6.4 [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. PCoA (Principal coordinates analysis) and alpha-diversity were performed with phyloseq v1.42.0 [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e] and ggplot2 v. 3.3.6 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/tidyverse/ggplot2\u003c/span\u003e\u003cspan address=\"https://github.com/tidyverse/ggplot2\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Linear discriminant analysis Effect Size (LefSe) was conducted using microbiomeMarker v1.4.0 [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e] with default parameters.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eModels for the prediction of postprandial glucose response\u003c/h2\u003e \u003cp\u003eWe used two measures of PPGR characteristics: iAUC120 [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e] and the peak glucose level within 120 minutes after the meal start (GLUmax, mmol/L). The latter indicator was chosen because the recommended timing of blood glucose self-monitoring for pregnant women is established in the time interval when glycemic levels are highest. The peak glycemic level in pregnant women with diabetes mellitus is reached 45\u0026ndash;75 minutes after a meal, which is the reason for the recommendation to measure glycemia one hour after a meal [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. However, the peak blood glucose (BG) level is less sensitive to inaccurate logging of meal start time than 1-hr postprandial BG.\u003c/p\u003e \u003cp\u003eOutliers in the target variable were removed using the Tukey\u0026rsquo;s Interquartile Range method [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e] resulting in the final dataset of 2,633 meals with PPGRs for GLUmax prediction model and 2,628 meals for iAUC120 prediction.\u003c/p\u003e \u003cp\u003eWe used the gradient boosting algorithm LightGBM [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e] to predict both indices and improved its performance with Optuna hyperparameter optimization [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eWithin Optuna, we adopted the Tree-structured Parzen Estimator (TPESampler) for sampling within the hyperparameter space, and the Asynchronous Successive Halving Algorithm (SuccessiveHalvingPruner) was implemented to eliminate underperforming trials efficiently. The optimal configuration of hyperparameters that emerged from our analysis included a \u0026ldquo;num_boost_round\u0026rdquo; of 4700, a \u0026ldquo;learning_rate\u0026rdquo; of 0.0015, \u0026ldquo;max_depth\u0026rdquo; of 11, \u0026ldquo;num_leaves\u0026rdquo; of 30, \u0026ldquo;min_sum_hessian_in_leaf\u0026rdquo; at 0.12, \u0026ldquo;bagging_fraction\u0026rdquo; of 0.55, \u0026ldquo;bagging_freq\u0026rdquo; of 10, \u0026ldquo;feature_fraction\u0026rdquo; of 0.4, \u0026ldquo;lambda_l1\u0026rdquo; at 0.006 and \u0026ldquo;lambda_l2\u0026rdquo; at 0.007. The data was divided into training and test sets with a ratio of 70:30, ensuring that records from the same patient were only included in one set to prevent data leakage and potential bias in performance metrics. To enhance the robustness of the training process, we employed 3-fold cross-validation.\u003c/p\u003e \u003cp\u003eThe final metrics, including mean absolute error (MAE), mean squared error (MSE), root mean squared error (RMSE), Pearson\u0026rsquo;s correlation coefficient (R), and the coefficient of determination (R\u003csup\u003e2\u003c/sup\u003e), were calculated on the held-out test sample in the Python scikit-learn library.\u003c/p\u003e \u003cp\u003e \u003cb\u003eOther statistical analyses\u003c/b\u003e. To describe the patients in our data frame, we employed the bootstrap hypothesis test [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. The main idea behind this method is to repeatedly draw random subsamples of the data with replacement in order to estimate the distribution of the test statistic and make decisions about the significance of differences. The bootstrap test does not require any assumptions about the underlying distribution of the original data, making it more robust compared to parametric tests such as the t-test.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eFeature selection and evaluation of input parameters\u003c/h2\u003e \u003cp\u003eWe selected 164 features as model inputs, including features characterizing meal content, anthropometric measures, gynecological data, blood tests results, CGM-derived features, lifestyle questionnaire data, and genetic and microbiome features (Supplementary Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn order to avoid model overfitting, we used several approaches to decrease the number of input variables. From the original lifestyle questionnaire characterizing the consumption of certain product groups and physical activity, described elsewhere [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e], we selected the parameters with significant Spearman correlations with iAUC120 and/or GLUmax (Supplementary Table S2).\u003c/p\u003e \u003cp\u003eMicrobial features were selected based on the results of LefSe analysis. For this purpose, all participants were divided into two groups based on the average levels of PPGR indices: group 1 \u0026ndash; below the median and group 2 \u0026ndash; equal or above the median for the group. The relative abundances (RA) of bacterial taxa differentially enriched in these groups were used as input variables.\u003c/p\u003e \u003cp\u003eAmong genetic factors, we selected rs10830963 and rs1387153 variants in \u003cem\u003eMTNR1B\u003c/em\u003e previously shown by our group to be associated with the results of OGTT in pregnant women [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eSHAP methods were utilized for enhancing model interpretability [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. SHAP values were computed in two ways. First, calculations were made individually for each feature to denote the average alteration in the model\u0026rsquo;s output when conditioning on that specific feature. Second, the additive nature of SHAP values was employed to assess the impact of various feature groups on the model.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eAvailability of Data\u003c/h2\u003e \u003cp\u003eThe datasets generated and analyzed in the current study are available in a github repository [\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/artemisak/MicrobesAndGlucouseAnalysis?tab=readme-ov-file\u003c/span\u003e\u003cspan address=\"https://github.com/artemisak/MicrobesAndGlucouseAnalysis?tab=readme-ov-file\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003eIn total, 152 participants were recruited to the study. After exclusion of 3 women who did not provide CGM data, 2 women with antibiotics intake during the study period, 34 women with inaccurate food diaries, 2 women with less than 6 meals left after filtering, and 6 microbiota samples with low read count (\u0026lt;10,000 reads), 105 participants (77 women with GDM and 28 healthy pregnant women) were included in the final analysis (Fig. 1).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe characteristics of the participants are in\u0026nbsp;Table 1. Women with GDM did not differ\u0026nbsp;from the control\u0026nbsp;group in terms of age and gestational age upon initiation of continuous glucose monitoring. Patients with GDM had higher body mass index (BMI) before pregnancy. As expected, healthy pregnant women had lower plasma glucose levels during OGTT and hemoglobin A1C (HbA1C) upon inclusion into the study.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePatients with GDM consumed lower amounts of carbohydrates (28.4 \u0026plusmn; 10.9 vs 36.6 \u0026plusmn; 10.8 g) and higher amounts of proteins (17.0 \u0026plusmn; 5.2 vs 13.8 \u0026plusmn; 2.9 g) per meal compared to healthy women (Table 1). Presumably due to this fact, iAUC120 and GLUmax levels did not significantly differ between the groups and even tended to be lower in women with GDM compared to their healthy counterparts who were not dieting (0.52 \u0026plusmn; 0.29 vs 0.63 \u0026plusmn; 0.28 and 6.2 \u0026plusmn; 0.6 vs 6.4 \u0026plusmn; 0.6 mmol/L, respectively) (Table 1).\u0026nbsp;For comprehensive details on lifestyle assessments and baseline blood tests, please refer to supplementary Table S3.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMicrobial features in women with higher and lower PPGRs\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAs there was no difference in the levels of GLUmax and iAUC120 between women with and without GDM during CGM, we combined their data for selection of microbial features associated with higher and lower PPGRs. The medians for iAUC120 and GLUmax in the cohort were 0.527 and 6.254 mmol/L, respectively. Participants with\u0026nbsp;median\u0026nbsp;PPGR indices (iAUC120 or GLUmax, respectively) below these numbers were considered to have lower PPGRs, and those with\u0026nbsp;median PPGR indices equal to or above\u0026nbsp;the cohort median comprised the subgroup with higher PPGRs.\u003c/p\u003e\n\u003cp\u003eLinear discriminant analysis\u0026nbsp;revealed 18 bacterial taxa exhibiting significantly higher scores in the subgroup of women with higher\u0026nbsp;iAUC120 and 21 bacterial taxa with higher \u0026nbsp;scores in the subgroup with lower iAUC120, P \u0026lt; 0.05 for all (Fig. 2).\u0026nbsp;Bacterial taxa displaying notably higher scores in women with higher\u0026nbsp;iAUC120 included\u0026nbsp;\u003cem\u003eDorea\u003c/em\u003e (Lachnospiraceae), \u003cem\u003eFusicatenibacter\u003c/em\u003e (Lachnospiraceae), \u003cem\u003eRuminococcus torques group\u003c/em\u003e (Oscillospiraceae), \u003cem\u003ePrevotella 9\u003c/em\u003e (Bacteroidia, Prevotellaceae), \u003cem\u003eCoprococcus comes\u003c/em\u003e (Lachnospiraceae), \u003cem\u003eRoseburia\u003c/em\u003e (Lachnospiraceae), \u003cem\u003e\u0026ldquo;Lachnoclostridium edouardi\u0026rdquo;\u003c/em\u003e (Lachnospiraceae), \u003cem\u003eMarvinbryantia\u003c/em\u003e (Lachnospiraceae), \u003cem\u003eAnaerobutyricum hallii\u003c/em\u003e (basonym: \u003cem\u003eEubacterium hallii\u003c/em\u003e) (Lachnospiraceae), \u003cem\u003eColidextribacter\u003c/em\u003e (Bacillota).\u0026nbsp;Taxa\u0026nbsp;with a higher score in the subgroup with lower iAUC120 included\u0026nbsp;\u003cem\u003eOscillospiraceae UCG-002\u003c/em\u003e, Muribaculaceae (Bacteroidota, Bacteroidia), \u003cem\u003eRuminococcus champanellensis\u003c/em\u003e (Oscillospiraceae), \u003cem\u003eChristensenellaceae R-7 group\u003c/em\u003e (Clostridia), \u003cem\u003eParabacteroides distasonis\u003c/em\u003e (Bacteroidia, Tannerellaceae), \u003cem\u003eBlautia\u003c/em\u003e (Lachnospiraceae), \u003cem\u003eSellimonas\u003c/em\u003e (Lachnospiraceae), \u003cem\u003eEisenbergiella tayi\u003c/em\u003e (Lachnospiraceae) and \u003cem\u003eBilophila wadsworthia\u003c/em\u003e (Deltaproteobacteria, Desulfovibrionaceae)\u003cem\u003e\u0026nbsp;\u003c/em\u003e(Fig. 2). All bacterial taxa distinguished by LefSe were included to input variables for creation of PPGR prediction models.\u003c/p\u003e\n\u003cp\u003eWhen comparing women with higher and lower GLUmax, 7 taxa were enriched in the subgroup with higher GLUmax, including Clostridia UCG 014 and \u003cem\u003e\u0026ldquo;Lachnoclostridium\u0026rdquo;\u003c/em\u003e (Lachnospiraceae), and 8 taxa were enriched in the subgroup with lower GLUmax, including \u003cem\u003eMethanosphaera\u003c/em\u003e (Methanobacteria), \u003cem\u003eLachnospira eligens\u003c/em\u003e (basonym: \u003cem\u003eEubacterium eligens\u003c/em\u003e) (Lachnospiraceae), \u003cem\u003eButyricicoccus faecihominis\u003c/em\u003e (Oscillospiraceae), \u003cem\u003eIntestinibacter bartlettii\u003c/em\u003e (Clostridia, Peptostreptococcaceae), \u003cem\u003eSellimonas\u003c/em\u003e (Lachnospiraceae), \u003cem\u003eE. tayi\u003c/em\u003e (Lachnospiraceae), \u003cem\u003eChristensenellaceae R-7 group\u003c/em\u003e (Clostridia) (Fig. 3).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePredicting individual postprandial responses\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe assessed the overall extent to which different combinations of input variables predict personal postprandial responses: iAUC120 and GLUmax.\u0026nbsp;A total of 750 days of concurrent CGM usage and meal logging resulted in 3,514 meals to be analyzed with their PPGRs. Meal filtering (see RESEARCH DESIGN AND METHODS, Meal preprocessing) reduced the dataset to 2,706 meals. After removal of outliers in the target variable, the final dataset comprised 2,633 meals with PPGRs for GLUmax prediction model and 2,628\u0026nbsp;meals for iAUC120\u0026nbsp;prediction. Prediction models for both indices were developed utilizing\u0026nbsp;gradient boosting algorithms, with the following combinations of input variables: 1) only carbohydrate content of the meal (carbs); 2) clinically available parameters (anthropometric, biochemical, lifestyle questionnaire,\u0026nbsp;meal content and meal context, CGM data);\u0026nbsp;3) model\u0026nbsp;2\u0026nbsp;parameters + microbial features (the full model). For the full list of features please see the Supplementary Table 1. Validation of the model was performed\u0026nbsp;using a three-fold\u0026nbsp;cross-validation scheme (see RESEARCH DESIGN AND METHODS).\u003c/p\u003e\n\u003cp\u003eIn the context of predicting GLUmax, the first model that relied solely on the amount of carbohydrates in a meal demonstrated the lowest correlation with PPGRs\u0026nbsp;(R = 0.35) and accounted for only 5% of the variation in glycemic response (Fig. 4A). The second model based on\u0026nbsp;clinically available parameters achieved a significantly higher correlation (R = 0.62) and explained 34 % of variance (Fig. 4B). Adding microbiome features (Fig. 4C) further increased the predictive ability with an R of 0.66 and a coefficient of determination of 42%.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eLikewise, in the prediction of iAUC120, a model based solely on the carbohydrate content of meals demonstrated a relatively weak correlation (R = 0.51) and explained only 26% of the variation in glycemic response (Fig. 5A). The addition of parameter groups, as described above, resulted in an increase in correlation between CGM-measured and predicted values (R = 0.71, R\u003csup\u003e2\u003c/sup\u003e = 0.50). Addition of microbial features to this model slightly increased the accuracy of prediction (R = 0.72, R\u003csup\u003e2\u003c/sup\u003e = 0.52) (Fig. 5B-5C).\u003c/p\u003e\n\u003cp\u003eBecause the performance of a model can also be affected by non-linear relationships between measured and predicted values, we also assessed MAE, MSE and RMSE for the models with higher performance (models 2-3, Table 2). As shown in Table 2, adding microbial features decreased MAE, MSE and RMSE for GLUmax prediction, but did not influence these parameters characterizing prediction of iAUC120.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExploring factors influencing the prediction of postprandial glycemic responses\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFollowing the examination of different models predicting PPGRs, our subsequent focus was on understanding the individual factors influencing prediction accuracy, including microbial features and other parameters comprising the full model. For this purpose, we conducted feature attribution analysis employing SHAP\u0026nbsp;[19].\u003c/p\u003e\n\u003cp\u003eThe features that exerted the greatest influence on iAUC120 prediction, as indicated by the highest mean absolute SHAP value, encompassed the carbohydrate content of the meal, glycemic load of the meal, amount of starch in the meal, and CGM-derived parameters characterizing glucose levels preceding the meal (glucose level 10 minutes before meal and glucose rise from 240 minutes before the meal to meal start) (Fig. 6A). The most influential parameters for the prediction of GLUmax were the glucose levels at the onset of the meal (GLU0), the carbohydrate content of the meal, glycemic load of the meal, RA of\u0026nbsp;\u003cem\u003eI. bartlettii\u003c/em\u003e,\u003cem\u003e\u0026nbsp;\u003c/em\u003eand the amount of protein consumed up to 6 hours before the meal\u003cem\u003e\u0026nbsp;\u003c/em\u003e(Fig. 6B).\u003c/p\u003e\n\u003cp\u003eAmong the 20 most influential parameters for the prediction of iAUC120 or GLUmax, the algorithm selected the RA of the following bacterial taxa: \u003cem\u003eI. bartlettii, \u0026ldquo;L. edouardi\u0026rdquo;, B. faecihominis\u003c/em\u003e (for iAUC120), and \u003cem\u003eI. bartlettii,\u003c/em\u003e\u003cem\u003eL. eligens\u003c/em\u003e (basonym: \u003cem\u003eEubacterium eligens\u003c/em\u003e), and \u003cem\u003eR. champanellensis\u003c/em\u003e (for GLUmax) (Fig. 6 A,B). Notably, \u003cem\u003eI. bartlettii\u003c/em\u003e ranked fourth among influential parameters for the prediction of GLUmax and was selected by the algorithm among the top parameters both for iAUC120 and for GLUmax prediction.\u003c/p\u003e\n\u003cp\u003eIn order to assess the cumulative influence of microbial composition and other feature groups on the model, we summed the SHAP values of associated features (Fig. 7). These examinations revealed that the meal composition had the most significant effect on prediction of iAUC120, followed by CGM-derived data, meal context, and microbial composition (Fig. 8). Оn the contrary, for the prediction of GLUmax the main predictor group was the CGM-derived data, followed by meal composition, meal context, and microbial data also taking the fourth place (Fig. 7).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eRecently, a high interpersonal difference in PPGRs was revealed, and the gut microbiota has been shown to be a factor underlying this variability. Furthermore, the gut microbiota has been used to enhance the accuracy of PPGR prediction in healthy volunteers and individuals with type 1 diabetes [\u003cspan additionalcitationids=\"CR8\" citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. However, to our knowledge, no study has explored the impact of the gut microbiome on PPGRs in pregnant women with or without GDM.\u003c/p\u003e \u003cp\u003ePregnancy is characterized by substantial alterations in all types of metabolism as well as by dynamic changes in gut microbial composition [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. This fact, in line with the importance of achieving target postprandial glucose levels for improved pregnancy outcomes, underscores the importance of comprehensive evaluation of factors underlying PPGRs in pregnant women for the construction of more accurate PPGR prediction models for personalized dietary advice in this specific population.\u003c/p\u003e \u003cp\u003eOur study has shown that microbiome features are among the top 10 most impactful individual parameters on the PPGR prediction in pregnant women with GDM. The cumulative influence of microbial composition was the fourth among the ten most impactful feature groups following CGM-derived data, meal composition and meal context for the prediction of GLUmax and iAUC120. Of note, adding microbiome features increased the predictive ability of the model with the increment of the coefficient of determination from 34\u0026ndash;42%.\u003c/p\u003e \u003cp\u003eThese findings underscore the potential role of the microbiome in the regulation of glycemic control even though the addition of microbiome features had lower impact on the accuracy metrics of iAUC120 prediction. The latter could be a consequence of relatively high accuracy of iAUC120 prediction even before microbiome addition.\u003c/p\u003e \u003cp\u003eThe postprandial glycemic predictions in our study (with R\u0026thinsp;=\u0026thinsp;0.72 for the model predicting iAUC120) closely resembled those documented by Zeevi et al. (with an R value of 0.70) in healthy subjects [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. The performance of our models was also superior to that of the models developed by Shilo et al. for patients with type 1 diabetes (R\u0026thinsp;=\u0026thinsp;0.72 vs 0.59 for iAUC120 prediction and R\u0026thinsp;=\u0026thinsp;0.66 vs 0.61 for GLUmax prediction) [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Furthermore, Shilo et al. included PPGRs from the same patient in training and validation datasets, while in our study, we separated participants between datasets so PPGRs of a participant from a training dataset could not be analyzed in the test or validation datasets. If Shilo et al. followed the same protocol, the difference in model performance might be even more pronounced. However, type 1 diabetes patients have much greater glucose variability and excursions, thus complicating the task of accurate PPGR prediction in this group of patients.\u003c/p\u003e \u003cp\u003eIn the biggest study of PPGRs in healthy individuals, to date, Berry et al. obtained the highest accuracy of iAUC120 prediction with R\u0026thinsp;=\u0026thinsp;0.75 in the validation cohort [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. They likely reached the maximum accuracy which could be anticipated judging by the correlation between PPGRs to repeated standard meals (intraindividual variability) of 0.7\u0026ndash;0.77 reported by Zeevi et al. [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. A potential reason for the lower performance of our model is almost 10-fold smaller sample size and a lower number of included genetic variants compared to the study by Berry et al. [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eCompared to our previous study on PPGRs prediction without implementing microbiome data [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e], our current model exhibited only a slight increase of R (0.72 vs 0.7). However, for the previous model, we had a larger dataset (3,240 records of meals and corresponding PPGRs) making direct comparisons inappropriate.\u003c/p\u003e \u003cp\u003eConcerning the gut microbiome, our study identified bacterial taxa differentially enriched in pregnant women with higher and lower mean PPGR. According to LefSe analyses, most taxa belong to Lachnospiraceae and Oscillospiraceae, and some families of Bacteroidia. These families are represented by the most functionally active bacteria involved in dietary fiber degradation and short-chain fatty acid (SCFA) biosynthesis [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. SCFAs, especially butyrate, are generally considered beneficial metabolites that reduce the risk of GDM. However, excess SCFAs can activate gluconeogenesis, leading to hyperglycemia and insulin resistance [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eEven though, in general, Lachnospiraceae and Oscillospiraceae are considered useful symbionts that interact beneficially with the host, among them, some taxa carry a \u0026ldquo;dual\u0026rdquo; function. For example, \u003cem\u003eAnaerobutyricum hallii\u003c/em\u003e and some \u003cem\u003eBlautia\u003c/em\u003e species, are considered pathobionts that can cause harm to the host [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. Further, among the taxa with higher abundance in women with higher iAUC120 or GLUmax, several are of interest for discussion as taxa potentially contributing to GDM pathogenesis. \u003cem\u003ePrevotella 9\u003c/em\u003e is now characterized as the new genus \u003cem\u003eSegatella\u003c/em\u003e with the type species \u003cem\u003eSegatella copri\u003c/em\u003e (basonym: \u003cem\u003ePrevotella copri\u003c/em\u003e). Although \u003cem\u003eS. copri\u003c/em\u003e is considered to be associated with health [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e], a significant positive association between increases in \u003cem\u003ePrevotella 9\u003c/em\u003e and higher GDM risk was identified [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e], and an increased abundance of \u003cem\u003ePrevotella\u003c/em\u003e was reported in GDM patients [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. \u003cem\u003eCoprococcus comes\u003c/em\u003e is a butyrate producer [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e], usually considered beneficial. However, in the FINRISK-2002 cohort, the strongest association with higher statin-associated new-onset type 2 diabetes risk was observed for \u003cem\u003eC. comes\u003c/em\u003e [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e], which aligns with our results. A possible explanation for this may be the ability of \u003cem\u003eC. comes\u003c/em\u003e to produce the highest butyrate levels [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e], which can lead to its excess.\u003c/p\u003e \u003cp\u003e \u003cem\u003eA. hallii\u003c/em\u003e (basonym: \u003cem\u003eEubacterium hallii\u003c/em\u003e) is also associated with health [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. However, GDM patients who failed to control glycemic levels were characterized by increased \u003cem\u003eA. hallii\u003c/em\u003e [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e], which corresponds to the results of our study.\u003c/p\u003e \u003cp\u003eIn women with lower iAUC120 or GLUmax, some taxa also had higher abundance, conversely suggesting a protective effect against higher PPGR.\u003c/p\u003e \u003cp\u003e \u003cem\u003eOscillospiraceae UCG-002\u003c/em\u003e, previously \u003cem\u003eRuminococcaceae UCG-002\u003c/em\u003e, was more abundant in the normal glucose tolerance group than in GDM. Previous research found it was reduced in early pregnancy in women with subsequent GDM and was negatively correlated with fasting blood glucose levels [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]. \u003cem\u003eOscillospiraceae UCG-002\u003c/em\u003e was also negatively associated with the homeostasis model assessment of insulin resistance (HOMA-IR) index and served as a marker of intestinal phytoestrogen enterolactone production [\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e].\u003c/p\u003e \u003cp\u003e \u003cem\u003eChristensenellaceae R-7 group\u003c/em\u003e is a beneficial genus: elevated abundance was associated with reduced visceral adipose tissue and a healthier metabolic profile [\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. \u003cem\u003eParabacteroides distasonis\u003c/em\u003e may protect against inflammation and obesity; however, increased abundance of \u003cem\u003eP. distasonis\u003c/em\u003e was previously reported in GDM [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. \u003cem\u003eSellimonas\u003c/em\u003e is an acetate producer, associated with a reduced type 2 diabetes risk [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e] and has been linked to low polycystic ovary syndrome (PCOS) risk [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e].\u003c/p\u003e \u003cp\u003e \u003cem\u003eEisenbergiella tayi\u003c/em\u003e produces butyrate, lactate, acetate, and succinate and is thought to be potentially beneficial. However, \u003cem\u003eE. tayi\u003c/em\u003e was associated with the disease state [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. Women who developed GDM showed a significantly higher abundance of \u003cem\u003eEisenbergiella\u003c/em\u003e in early pregnancy, and \u003cem\u003eEisenbergiella\u003c/em\u003e was also positively correlated with fasting blood glucose levels [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e], which contradicts our results. \u003cem\u003eBilophila wadsworthia\u003c/em\u003e is associated with the metabolism of fatty acid esters of hydroxy fatty acids, which improves glucose homeostasis, stimulates insulin sensitivity, and has anti-inflammatory effects [\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eAfter including bacterial taxa distinguished by LefSe as input variables for creation of PPGR prediction models, microbiome features were categorized as either advantageous or disadvantageous. As the RA of these taxa increased, the algorithm projected a decrease or increase in postprandial glucose response, respectively. Among bacterial features, the greatest contribution to iAUC120 prediction was made by the RAs of \u003cem\u003eI. bartlettii\u003c/em\u003e, \u003cem\u003eB. faecihominis\u003c/em\u003e, and \u003cem\u003e\u0026ldquo;L. edouardi\u0026rdquo;\u003c/em\u003e. The most impactful bacterial features for the prediction of GLUmax were \u003cem\u003eI. bartlettii, R. champanellensis\u003c/em\u003e, and \u003cem\u003eL. eligens\u003c/em\u003e. A higher abundance of these bacteria was associated with lower PPGRs.\u003c/p\u003e \u003cp\u003eNotably, \u003cem\u003eI. bartlettii\u003c/em\u003e was selected by the algorithm both for the prediction of iAUC120 and GLUmax among the top 20 parameters. \u003cem\u003eI. bartlettii\u003c/em\u003e can produce indoleacetic and phenylacetic acids, acetate, isovalerate, and isobutyrate. Due to the latter's production, \u003cem\u003eIntestinibacter\u003c/em\u003e might be beneficial to host lipid and glucose metabolism and intestinal barrier integrity, which may explain the inverse association of \u003cem\u003eIntestinibacter\u003c/em\u003e with diabetes [\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e].\u003c/p\u003e \u003cp\u003e \u003cem\u003eL. eligens\u003c/em\u003e produces butyrate, acetate, and lactate, and promotes the production of the anti-inflammatory cytokine IL-10. \u003cem\u003eL. eligens\u003c/em\u003e was reduced in early pregnancy in women with subsequent GDM [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]. The abundance of \u003cem\u003eL. eligens\u003c/em\u003e was significantly higher in the healthy controls than in the obese individuals [\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e]. Additionally, \u003cem\u003eL. eligens\u003c/em\u003e was positively associated with adherence to a Mediterranean diet [\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e]. \u003cem\u003eB. faecihominis\u003c/em\u003e is a butyrate producer and was included in the stool-derived microbial ecosystem therapeutics to combat \u003cem\u003eClostridioides difficile\u003c/em\u003e infection as a beneficial bacterium [\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e]. \u003cem\u003eR. champanellensis\u003c/em\u003e is a cellulose-degrading bacterium [\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e]. The strongest association with lower statin-associated new-onset type 2 diaberes risk was observed for \u003cem\u003eR. champanellensis\u003c/em\u003e [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. \u003cem\u003e\u0026ldquo;L. edouardi\u0026rdquo;\u003c/em\u003e was associated with an increased risk of GDM [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e] and heightened type 2 diabetes risk [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe limitation of our study is a relatively small sample size. Further studies in other cohorts and populations of pregnant women are needed to confirm our findings concerning certain bacterial taxa associated with PPGRs. There is further room for improvement, such as conducting more comprehensive assessments of contextual factors than those employed in the current study. For example, including data on physical activity preceding meals and integrating extensive 'omics' data could improve the predictive capacity of these algorithms.\u003c/p\u003e \u003cp\u003eIt is essential to delve deeper into understanding the functional roles of bacterial taxa that were the most influential for PPGR prediction in our study. The insights gained from this data could pave the way for the future advancement of probiotic or autoprobiotic therapies aimed at enhancing glycemic regulation. Probiotics with metabolic effects that target functionally active bacteria, predominantly belonging to Clostridia and Bacterodia, which play a key role in maintaining the balance of the intestinal microbiota, seem promising [\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e].\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eOur study highlights the emerging role of the gut microbiota in the interpersonal variability of PPGRs. While previous research extensively utilized microbiota for PPGR prediction in healthy individuals and those with type 1 diabetes, our study fills an important gap by examining its impact on PPGRs in pregnant women, particularly those with GDM.\u003c/p\u003e \u003cp\u003eOur findings indicate that microbiome features rank among the top parameters influencing PPGR prediction in pregnant women with GDM. Specifically, certain bacterial taxa were identified as significantly associated with variations in PPGRs. Incorporating microbiome data enhanced the accuracy of our predictions, highlighting the potential of microbiota-based interventions for optimizing glycemic control.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThis work was financially supported by the Ministry of Science and Higher Education of the Russian Federation (Agreement No. 075-15-2022-301) and the Ministry of Innovation, Science \u0026amp; Technology, Israel. The funders played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eP.P., E.P., E.G., O.K. and E.S. were involved in the conception, design, and conduct of the study and the analysis and interpretation of the results. A.I., A.R., and S.S. designed and conducted the analyses, interpreted the results, and wrote the manuscript. A.A., E.V., A.T., I.N., Ek. S., A.E., El.S. and T. P. provided data and interpreted the results, M.K. and E.V. performed laboratory analysis and interpreted the results., S.Z., E.R., C.E., and S.T. performed fecal specimen processing, sample sequencing and interpreted the results. L.V. performed blood specimen processing, genotyping assays and interpreted the results, P.P., A.I., A.R., and S.S. wrote the first draft of the manuscript, and all authors edited, reviewed, and approved the final version of the manuscript. P.P. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe datasets generated and analyzed in the current study are available in a github repository [https://github.com/artemisak/MicrobesAndGlucouseAnalysis?tab=readme-ov-file].\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSacks DA, et al. Frequency of gestational diabetes mellitus at collaborating centers based on IADPSG consensus panel-recommended criteria. Diabetes Care 2012;35:526-8.https://doi.org/10.2337/dc11-1641\u003c/li\u003e\n\u003cli\u003eHAPO Study Cooperative Research Group; Metzger BE, et al. Hyperglycemia and adverse pregnancy outcomes. N Engl J Med 2008;358:1991-2002.https://doi.org/10.1056/NEJMoa0707943\u003c/li\u003e\n\u003cli\u003eHajj NE, Schneider E, Lehnen H, Haaf T. Epigenetics and life-long consequences of an adverse nutritional and diabetic intrauterine environment. Reproduction 2014;148(6):R111-20.https://doi.org/10.1530/REP-14-0334\u003c/li\u003e\n\u003cli\u003ePopova P, Castorino K, Grineva EN, Kerr D. Gestational diabetes mellitus diagnosis and treatment goals: measurement and measures. Minerva Endocrinol 2016;41(4):421-32. PMID: 26824326.\u003c/li\u003e\n\u003cli\u003eAmerican Diabetes Association Professional Practice Committee. 15. Management of Diabetes in Pregnancy: Standards of Care in Diabetes-2024. Diabetes Care 2024;47(Suppl 1):S282-S294.https://doi.org/10.2337/dc24-S015\u003c/li\u003e\n\u003cli\u003eKoning SH, et al. Neonatal and obstetric outcomes in diet- and insulin-treated women with gestational diabetes mellitus: a retrospective study. BMC Endocr Disord 2016;16(1):52.https://doi.org/10.1186/s12902-016-0136-4\u003c/li\u003e\n\u003cli\u003eZeevi D, et al. Personalized Nutrition by Prediction of Glycemic Responses. Cell 2015;163(5):1079-1094.https://doi.org/10.1016/j.cell.2015.11.001\u003c/li\u003e\n\u003cli\u003eBerry SE, et al. Human postprandial responses to food and potential for precision nutrition. Nat Med 2020;26(6):964-973.https://doi.org/10.1038/s41591-020-0934-0\u003c/li\u003e\n\u003cli\u003eShilo S, et al. Prediction of Personal Glycemic Responses to Food for Individuals With Type 1 Diabetes Through Integration of Clinical and Microbial Data. Diabetes Care 2022;45(3):502-511.https://doi.org/10.2337/dc21-1048\u003c/li\u003e\n\u003cli\u003eShilo S, et al. The gut microbiome of adults with type 1 diabetes and its association with the host glycemic control. Diabetes Care 2022;45(3):555-563.https://doi.org/10.2337/dc21-1656\u003c/li\u003e\n\u003cli\u003ePustozerov EA, et al. Machine learning approach for postprandial blood glucose prediction in gestational diabetes mellitus. IEEE Access 2020;8:219308-219321.https://doi.org/10.1109/ACCESS.2020.3042483\u003c/li\u003e\n\u003cli\u003ePopova P, et al. A randomised, controlled study of different glycaemic targets during gestational diabetes treatment: Effect on the level of adipokines in cord blood and ANGPTL4 expression in human umbilical vein endothelial cells. Int J Endocrinology 2018;2018:6481658.https://doi.org/10.1155/2018/6481658\u003c/li\u003e\n\u003cli\u003eInternational Association of Diabetes and Pregnancy Study Groups Consensus Panel; Metzger BE, et al. International association of diabetes and pregnancy study groups recommendations on the diagnosis and classification of hyperglycemia in pregnancy. Diabetes Care 2010;33(3):676-682.https://doi.org/10.2337/dc09-1848\u003c/li\u003e\n\u003cli\u003ePustozerov E, et al. Development and evaluation of a mobile personalized blood glucose prediction system for patients with gestational diabetes mellitus. JMIR Mhealth Uhealth 2018;6(1):e6.https://doi.org/10.2196/mhealth.9236\u003c/li\u003e\n\u003cli\u003ePustozerov E, et al. The role of glycemic index and glycemic load in the development of real-time postprandial glycemic response prediction models for patients with gestational diabetes. Nutrients 2020;12(2):302.https://doi.org/10.3390/nu12020302\u003c/li\u003e\n\u003cli\u003eCaporaso JG, et al. Ultra-high-throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. ISME J 2012;6(8):1621-4.https://doi.org/10.1038/ismej.2012.8\u003c/li\u003e\n\u003cli\u003eAndrews, S. (2010). FastQC: A Quality Control Tool for High Throughput Sequence Data [Online]. Available online at: http://www.bioinformatics.babraham.ac.uk/projects/fastqc/\u003c/li\u003e\n\u003cli\u003eEwels, P. et al. MultiQC: summarize analysis results for multiple tools and samples in a single report, Bioinformatics, Volume 32, Issue 19, October 2016, Pages 3047-3048, https://doi.org/10.1093/bioinformatics/btw354\u003c/li\u003e\n\u003cli\u003eBolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics. 2014 Aug 1;30(15):2114-20. doi: 10.1093/bioinformatics/btu170. \u003c/li\u003e\n\u003cli\u003eCallahan BJ, et al. (2016). \u0026quot;DADA2: High-resolution sample inference from Illumina amplicon data.\u0026quot; Nature Methods, 13, 581-583. doi: 10.1038/nmeth.3869.https://doi.org/10.1038/nmeth.3869\u003c/li\u003e\n\u003cli\u003eMirdita M, Steinegger M, Breitwieser F, S\u0026ouml;ding J, Levy Karin E. Fast and sensitive taxonomic assignment to metagenomic contigs. Bioinformatics. 2021 Sep 29;37(18):3029-3031. doi: 10.1093/bioinformatics/btab184. PMID: 33734313; PMCID: PMC8479651.https://doi.org/10.1093/bioinformatics/btab184\u003c/li\u003e\n\u003cli\u003eQuast, C. et al. The SILVA ribosomal RNA gene database project: Improved data processing and web-based tools. Nucleic Acids Res. 41, D590-D596 (2012).https://doi.org/10.1093/nar/gks1219\u003c/li\u003e\n\u003cli\u003eOksanen, F.J., et al. (2017) Vegan: Community Ecology Package. R package Version 2.4-3. https://CRAN.R-project.org/package=vegan\u003c/li\u003e\n\u003cli\u003eMcMurdie, P. J., \u0026amp; Holmes, S. (2013). phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data. PloS one, 8(4), e61217. https://doi.org/10.1371/journal.pone.0061217\u003c/li\u003e\n\u003cli\u003eYang Cao et al. MicrobiomeMarker: an R/Bioconductor package for microbiome marker identification and visualization, Bioinformatics, Volume 38, Issue 16, August 2022, Pages 4027-4029, https://doi.org/10.1093/bioinformatics/btac438\u003c/li\u003e\n\u003cli\u003eB\u0026uuml;hling KJ, et al. Optimal timing for postprandial glucose measurement in pregnant women with diabetes and a non-diabetic pregnant population evaluated by the Continuous Glucose Monitoring System (CGMS). J Perinat Med 2005;33:125-31.https://doi.org/10.1515/JPM.2005.024\u003c/li\u003e\n\u003cli\u003eTukey, J. W. (1977). Exploratory Data Analysis. Addison-Wesley\u003c/li\u003e\n\u003cli\u003eKe, G., et al. (2017). Lightgbm: A highly efficient gradient boostin decision tree. Advances in Neural Information Processing Systems, 30, 3146-3154.\u003c/li\u003e\n\u003cli\u003eAkiba T, Sano S, Yanase T, Ohta T, Koyama M. 2019. Optuna: a next-generation hyperparameter optimization framework. KDD \u0026apos;19: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery \u0026amp; Data; p. 2623-2631. doi: 10.1145/3292500.3330701.\u003c/li\u003e\n\u003cli\u003eEfron, B. (1979). Bootstrap methods: Another look at the jackknife. The Annals of Statistics, 7(1), 1-26https://doi.org/10.1214/aos/1176344552\u003c/li\u003e\n\u003cli\u003ePopova PV, et al. Effect of gene-lifestyle interaction on gestational diabetes risk. Oncotarget 2017;8:112024-112035.https://doi.org/10.18632/oncotarget.22999\u003c/li\u003e\n\u003cli\u003eLundberg SM, Erion G, Chen H, et al. From local explanations to global understanding with explainable AI for trees. Nat Mach Intell 2020;2:56-67.https://doi.org/10.1038/s42256-019-0138-9\u003c/li\u003e\n\u003cli\u003eNuriel-Ohayon M, Neuman H, Koren O. Microbial Changes during Pregnancy, Birth, and Infancy. Front Microbiol 2016;7:1031.https://doi.org/10.3389/fmicb.2016.01031\u003c/li\u003e\n\u003cli\u003eAbdugheni R, et al. Metabolite profiling of human-originated Lachnospiraceae at the strain level. iMeta 2022;1(4):e58.https://doi.org/10.1002/imt2.58\u003c/li\u003e\n\u003cli\u003eHu R, et al. Gut Microbiota and Critical Metabolites: Potential Target in Preventing Gestational Diabetes Mellitus? Microorganisms 2023;11(7):1725.https://doi.org/10.3390/microorganisms11071725\u003c/li\u003e\n\u003cli\u003eGiliberti R, Cavaliere S, Mauriello IE, Ercolini D, Pasolli E. Host phenotype classification from human microbiome data is mainly driven by the presence of microbial taxa. PLoS Comput Biol. 2022;18(4):e1010066.https://doi.org/10.1371/journal.pcbi.1010066\u003c/li\u003e\n\u003cli\u003eWu X, et al. Investigating causal associations among gut microbiota, gut microbiota-derived metabolites, and gestational diabetes mellitus: a bidirectional Mendelian randomization study. Aging (Albany NY) 2023;15(16):8345-8366.https://doi.org/10.18632/aging.204973\u003c/li\u003e\n\u003cli\u003ePonzo V, et al. Diet-Gut Microbiota Interactions and Gestational Diabetes Mellitus (GDM). Nutrients 2019;11(2):330.https://doi.org/10.3390/nu11020330\u003c/li\u003e\n\u003cli\u003eKoponen K, et al. Role of Gut Microbiota in Statin-Associated New-Onset Diabetes-A Cross-Sectional and Prospective Analysis of the FINRISK 2002 Cohort. Arterioscler Thromb Vasc Biol 2024;44(2):477-487.https://doi.org/10.1161/ATVBAHA.123.319458\u003c/li\u003e\n\u003cli\u003eYe G, et al. The Gut Microbiota in Women Suffering from Gestational Diabetes Mellitus with the Failure of Glycemic Control by Lifestyle Modification. J Diabetes Res 2019;2019:6081248.https://doi.org/10.1155/2019/6081248\u003c/li\u003e\n\u003cli\u003eMa S, et al. Alterations in Gut Microbiota of Gestational Diabetes Patients During the First Trimester of Pregnancy. Front Cell Infect Microbiol 2020;10:58.https://doi.org/10.3389/fcimb.2020.00058\u003c/li\u003e\n\u003cli\u003eAtzeni A, et al. Taxonomic and Functional Fecal Microbiota Signatures Associated With Insulin Resistance in Non-Diabetic Subjects With Overweight/Obesity Within the Frame of the PREDIMED-Plus Study. Front Endocrinol (Lausanne) 2022;13:804455.https://doi.org/10.3389/fendo.2022.804455\u003c/li\u003e\n\u003cli\u003eTavella T, et al. Elevated gut microbiome abundance of Christensenellaceae, Porphyromonadaceae and Rikenellaceae is associated with reduced visceral adipose tissue and healthier metabolic profile in Italian elderly. Gut Microbes 2021;13(1):1-19.https://doi.org/10.1080/19490976.2021.1880221\u003c/li\u003e\n\u003cli\u003eAlcazar M, et al. Gut microbiota is associated with metabolic health in children with obesity. Clin Nutr 2022;41(8):1680-1688.https://doi.org/10.1016/j.clnu.2022.06.007\u003c/li\u003e\n\u003cli\u003eSong S, Zhang Q, Zhang L, Zhou X, Yu J. A two-sample bidirectional Mendelian randomization analysis investigates associations between gut microbiota and type 2 diabetes mellitus. Front Endocrinol (Lausanne) 2024;15:1313651.https://doi.org/10.3389/fendo.2024.1313651PMid:38495787 PMCid:PMC10940336\u003c/li\u003e\n\u003cli\u003eLiang Y, et al. Gut microbiome and reproductive endocrine diseases: a Mendelian randomization study. Front Endocrinol (Lausanne) 2023;14:1164186.https://doi.org/10.3389/fendo.2023.1164186\u003c/li\u003e\n\u003cli\u003eFolz J, et al. Human metabolome variation along the upper intestinal tract. Nat Metab 2023;5(5):777-788.https://doi.org/10.1038/s42255-023-00777-z\u003c/li\u003e\n\u003cli\u003eLuo K, et al. Metabolic and inflammatory perturbation of diabetes associated gut dysbiosis in people living with and without HIV infection. Genome Med 2024;16(1):59.https://doi.org/10.1186/s13073-024-01336-1\u003c/li\u003e\n\u003cli\u003eHu X, et al. Integrative metagenomic analysis reveals distinct gut microbial signatures related to obesity. BMC Microbiol 2024;24(1):119.https://doi.org/10.1186/s12866-024-03278-5\u003c/li\u003e\n\u003cli\u003eGhosh TS, et al. Mediterranean diet intervention alters the gut microbiome in older people reducing frailty and improving health status: the NU-AGE 1-year dietary intervention across five European countries. Gut. 2020;69(7):1218-1228.https://doi.org/10.1136/gutjnl-2019-319654\u003c/li\u003e\n\u003cli\u003eCarlucci C, et al. Effects of defined gut microbial ecosystem components on virulence determinants of Clostridioides difficile. Sci Rep 2019;9(1):885.https://doi.org/10.1038/s41598-018-37547-x\u003c/li\u003e\n\u003cli\u003eMora\u0026iuml;s S, et al. Cryptic diversity of cellulose-degrading gut bacteria in industrialized humans. Science 2024;383(6688):eadj9223.https://doi.org/10.1126/science.adj9223\u003c/li\u003e\n\u003cli\u003ePokrotnieks J, Sitkin S. He who controls Clostridia and Bacteroidia controls the gut microbiome: The concept of targeted probiotics to restore the balance of keystone taxa in irritable bowel syndrome. Neurogastroenterol Motil. 2024:e14805.https://doi.org/10.1111/nmo.14805\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTable 1\u0026nbsp;\u0026ndash; Characteristics of the participants\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"623\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003eGDM\u003c/p\u003e\n \u003cp\u003eN=77\u003c/p\u003e\n \u003cp\u003e(mean\u0026plusmn;SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003eHealthy pregnant women\u003c/p\u003e\n \u003cp\u003eN=28\u003c/p\u003e\n \u003cp\u003e(mean\u0026plusmn;SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003ep-value\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003eAge, years\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e32.2 \u0026plusmn; 4.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e31.4 \u0026plusmn; 4.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.392\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003eGestational age (weeks)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e30.1\u0026plusmn; 4.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e30.0\u0026plusmn; 2.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.947\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003eBMI before pregnancy (kg/m\u003csup\u003e2\u003c/sup\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e24.7 \u0026plusmn; 5.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e22.1 \u0026plusmn; 3.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.017\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003eFasting PG (mmol/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e5.1 \u0026plusmn; 0.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e4.4 \u0026plusmn; 0.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e1-h postload glucose (mmol/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e9.5 \u0026plusmn; 1.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e6.7 \u0026plusmn; 1.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e2-h postload glucose (mmol/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e8.3 \u0026plusmn; 1.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e6.0 \u0026plusmn;1.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003eHbA1C, %\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e5.0 \u0026plusmn; 0.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e4.8 \u0026plusmn; 0.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.020\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" colspan=\"4\" valign=\"top\"\u003e\n \u003cp\u003eReal-time meal logging\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003eDays logged per participant\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e7.2 \u0026plusmn; 0.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e7.0 \u0026plusmn; 0.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003eMeals logged per participant\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e26.5 \u0026plusmn; 3.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e25.3 \u0026plusmn; 4.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.72\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003eEnergy intake per meal (kcal)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e334.8 \u0026plusmn; 99.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e346.3 \u0026plusmn; 95.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.597\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003eCarbohydrate intake per meal (g)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e28.4 \u0026plusmn; 10.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e36.6 \u0026plusmn; 10.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003eGlycaemic load per meal (g)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e14.8 \u0026plusmn; 7.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e20.4 \u0026plusmn; 6.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003eFat intake per meal (g)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e16.4 \u0026plusmn; 5.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e15.3 \u0026plusmn; 5.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.413\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003eProtein intake per meal (g)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e17.0 \u0026plusmn; 5.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e13.8 \u0026plusmn; 2.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.003\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" colspan=\"4\" valign=\"top\"\u003e\n \u003cp\u003eCGM \u0026ndash; derived indices\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003eMean GLUmax, mmol/L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e6.2 \u0026plusmn; 0.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e6.4 \u0026plusmn; 0.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.083\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003eMean iAUC120, mmol/L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.52 \u0026plusmn; 0.29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.63 \u0026plusmn; 0.28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.092\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eNotes: Comparisons were performed using the bootstrap hypothesis test.\u003c/p\u003e\n\u003cp\u003eTable 2 \u0026ndash; Accuracy of the models for predicting PPGRs and peak postprandial glycaemic levels based on clinical data with and without the addition of bacterial features\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"622\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"17.84565916398714%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"40.836012861736336%\" colspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003eGLUmax\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"41.31832797427653%\" colspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003ePPGR (iAUC120)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"17.84565916398714%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003eWithout microbiome\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003eWith\u003c/p\u003e\n \u003cp\u003emicrobiome\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003eWithout microbiome\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.90032154340836%\" valign=\"top\"\u003e\n \u003cp\u003eWith\u003c/p\u003e\n \u003cp\u003emicrobiome\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"17.84565916398714%\" valign=\"top\"\u003e\n \u003cp\u003eMAE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003e0.49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003e0.46\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003e0.30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.90032154340836%\" valign=\"top\"\u003e\n \u003cp\u003e0.30\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"17.84565916398714%\" valign=\"top\"\u003e\n \u003cp\u003eMSE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003e0.38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003e0.33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003e0.15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.90032154340836%\" valign=\"top\"\u003e\n \u003cp\u003e0.15\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"17.84565916398714%\" valign=\"top\"\u003e\n \u003cp\u003eRMSE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003e0.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003e0.39\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.90032154340836%\" valign=\"top\"\u003e\n \u003cp\u003e0.39\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"17.84565916398714%\" valign=\"top\"\u003e\n \u003cp\u003ePearson R\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003e0.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003e0.71\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.90032154340836%\" valign=\"top\"\u003e\n \u003cp\u003e0.72\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"17.84565916398714%\" valign=\"top\"\u003e\n \u003cp\u003eR\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003e0.34\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003e0.42\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.418006430868168%\" valign=\"top\"\u003e\n \u003cp\u003e0.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.90032154340836%\" valign=\"top\"\u003e\n \u003cp\u003e0.52\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eNotes: CGM \u0026ndash; continuous glucose monitoring, GLUmax \u0026ndash; peak postprandial glycaemic level, PPGR \u0026ndash; postprandial glycemic response, iAUC120 \u0026ndash; incremental area under glucose curve during 120 minutes after meal, MAE \u0026ndash; mean absolute error, MSE \u0026ndash; mean squared error, RMSE \u0026ndash; root mean squared error, R \u0026ndash; coefficient of correlation for predicted and observed values, R\u003csup\u003e2\u003c/sup\u003e \u0026ndash; coefficient of determination.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"npj-biofilms-and-microbiomes","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjbiofilms","sideBox":"Learn more about [npj Biofilms and Microbiomes](http://www.nature.com/npjbiofilms/)","snPcode":"41522","submissionUrl":"https://submission.springernature.com/new-submission/41522/3","title":"npj Biofilms and Microbiomes","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"glycemic response, personalized nutrition, microbiome, postprandial glycemia, Intestinibacter bartlettii, Butyricicoccus faecihominis, “Lachnoclostridium edouardi”, Ruminococcus champanellensis, Lachnospira eligens, continuous glucose monitoring.","lastPublishedDoi":"10.21203/rs.3.rs-4850670/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4850670/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eWe aimed to develop a prediction model for postprandial glycemic response (PPGR) in pregnant women with gestational diabetes mellitus (GDM) and to explore the influence of gut microbial data on prediction accuracy.\u003c/p\u003e \u003cp\u003eWe enrolled 105 pregnant women (70 GDM and 35 healthy). Participants underwent continuous glucose monitoring (CGM) for 7 days and provided detailed food diaries. Stool samples were collected at 28.8\u0026thinsp;\u0026plusmn;\u0026thinsp;3.6 gestational weeks, followed by 16S rRNA gene sequence analysis. We developed machine learning algorithms for predicting PPGR, incorporating CGM measurements, meal content, lifestyle factors, biochemical parameters, anthropometrics, and gut microbiota data. The accuracy of the models with and without gut microbiota were compared.\u003c/p\u003e \u003cp\u003ePPGR prediction models were created based on 2,706 meals with measured PPGRs. The integration of microbiome data in models increased the explained variance in peak glycemic levels (GLUmax) from 34\u0026ndash;42% and the explained variance in the incremental area under the glycemic curve 120 minutes after meal start (iAUC120) from 50\u0026ndash;52%. The final model performed better than the model based solely on carbohydrate count in terms of correlation between predicted and measured PPGRs (r\u0026thinsp;=\u0026thinsp;0.72 vs r\u0026thinsp;=\u0026thinsp;0.51 for iAUC120 and r\u0026thinsp;=\u0026thinsp;0.66 vs r\u0026thinsp;=\u0026thinsp;0.35 for GLUmax). After summing the SHAP values of associated features, the microbiome emerged as the fourth most impactful parameter for GLUmax and iAUC120 prediction, following meal composition, CGM measurements, and meal context.\u003c/p\u003e \u003cp\u003eMicrobiome features rank among the top 5 most impactful parameters in predicting PPGR in women with GDM.\u003c/p\u003e","manuscriptTitle":"Personalized Prediction of Glycemic Responses to Food in Women with Gestational Diabetes: Gut Microbiota Matters","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-10-15 13:41:34","doi":"10.21203/rs.3.rs-4850670/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2024-11-04T15:16:19+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-11-04T04:46:51+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"260030635620760568259954619842256473479","date":"2024-10-28T22:03:53+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-10-09T21:52:45+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"38589697122812189479109506519727983659","date":"2024-10-01T03:29:54+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"174695669838715949464653734144533455781","date":"2024-09-27T01:47:27+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-08-26T10:01:06+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-08-21T11:36:51+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-08-16T16:18:14+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Biofilms and Microbiomes","date":"2024-08-02T22:43:24+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-biofilms-and-microbiomes","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjbiofilms","sideBox":"Learn more about [npj Biofilms and Microbiomes](http://www.nature.com/npjbiofilms/)","snPcode":"41522","submissionUrl":"https://submission.springernature.com/new-submission/41522/3","title":"npj Biofilms and Microbiomes","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"439ce43d-ba9f-4aec-a79a-30e0e2033a89","owner":[],"postedDate":"October 15th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":36885696,"name":"Biological sciences/Microbiology"},{"id":36885697,"name":"Health sciences/Health care"}],"tags":[],"updatedAt":"2025-02-10T16:04:21+00:00","versionOfRecord":{"articleIdentity":"rs-4850670","link":"https://doi.org/10.1038/s41522-025-00650-9","journal":{"identity":"npj-biofilms-and-microbiomes","isVorOnly":false,"title":"npj Biofilms and Microbiomes"},"publishedOn":"2025-02-07 15:58:04","publishedOnDateReadable":"February 7th, 2025"},"versionCreatedAt":"2024-10-15 13:41:34","video":"","vorDoi":"10.1038/s41522-025-00650-9","vorDoiUrl":"https://doi.org/10.1038/s41522-025-00650-9","workflowStages":[]},"version":"v1","identity":"rs-4850670","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4850670","identity":"rs-4850670","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}