Method and system for screening intestinal flora and metabolite biomarkers with diet intervention weight loss effect of overweight or obese patient and application of intestinal flora and metabolite biomarkers

Abstract

The invention relates to an intestinal flora and metabolite biomarker screening method and system for weight loss effect of diet intervention of overweight or obese patients and application, and belongs to the technical field of microbiome, related metabolome and human health. The biomarker combination is designed and screened based on a random control intervention test, six strains and three excrement metabolites are obviously related to the weight reduction effect, the relative abundance of the six strains and the concentration of the two metabolites are in negative correlation with the weight reduction effect, and the concentration of one metabolite is in positive correlation with the weight reduction effect. Besides, an efficient weight loss effect prediction model is established by using the screened biomarker combination, a new strategy is provided for realizing effective weight loss, and personalized weight management can be guided.

Description

Technical Field

[0001] The present invention belongs to the field of microbiome and related metabolome and human health technology, and more specifically, to a method, system and application for screening biomarkers of intestinal flora and metabolites for the weight loss effect of dietary intervention in overweight or obese patients. Background Art

[0002] Obesity is a chronic disease caused by multiple factors and is a global public health challenge. The prevalence of overweight and obesity among adults in my country is about 46%, and the total number of overweight and obese people ranks first in the world (Wang et al., 2019; Pan et al., 2021). Overweight and obesity are risk factors for many diseases, including hypertension, diabetes, cardiovascular disease, and certain cancers (such as breast cancer, colorectal cancer, etc.) (Wang et al., 2021). Obesity not only increases the risk of adverse health outcomes for individuals, but also brings huge medical burdens to society. Studies have shown that a 5%-10% weight loss can help improve blood sugar control, lower blood pressure, and increase high-density lipoprotein cholesterol concentrations, thereby reducing the risk of related diseases (Elmaleh-Sachs et al., 2023).

[0003] Lifestyle intervention is the preferred treatment for overweight and obesity, and other treatments include weight loss drugs and bariatric surgery. "Clinical Management and Treatment of Obesity in China" points out that dietary intervention is the core component of lifestyle intervention in weight management (Zeng et al., 2021). Studies have shown that compared with conventional diets, diets such as energy-restricted diets, low-carbohydrate diets, time-restricted diets, and Mediterranean diets can effectively reduce the weight of overweight or obese people, but the weight loss effect is affected by multiple factors such as the environment and genetic background (Thom & Lean, 2017). Therefore, finding predictive factors that can be used to predict the effect of weight loss and establishing a predictive model for weight loss is crucial to preventing obesity, improving weight loss effects, and reducing the risk of obesity-related complications.

[0004] Metagenomics uses high-throughput sequencing technology to obtain all genomic information contained in microbial communities, which can more deeply study the structure and function of microorganisms in samples and provide new clues to disease mechanisms. However, metagenomics technology cannot directly measure the functional activity and key active molecules of microorganisms. Metabolomics is a qualitative and quantitative detection of small molecule metabolites in samples, which helps to clarify the pathogenesis of diseases. The combined analysis of metagenomics and metabolomics can overcome the limitations of single-omics research to a certain extent, help to fully understand the relationship between microorganisms, metabolites and diseases, and identify potential biomarkers and therapeutic targets.

[0005] The causes of obesity are complex and may include unhealthy lifestyles (high-calorie diets, low physical activity levels), genetic factors, and social environmental factors, among which intestinal flora and its metabolites may play an important role (Wu et al., 2021). Experimental and observational studies suggest (Le Chatelier et al., 2013; Ley et al., 2006; Turnbaugh et al., 2009) that the composition of intestinal flora is associated with obesity, and the diversity of intestinal flora in obese individuals is reduced. In addition, intestinal microorganisms may also regulate the host's metabolism and weight status through different mechanisms, such as directly affecting the amount of calories absorbed by the body from food, increasing intestinal permeability, and indirectly affecting the host's metabolism through metabolites produced by the flora or regulating diet and host-derived metabolites, thereby affecting the body's weight ( et al., 2004; Cardinelli et al., 2015; Van Hul & Cani, 2023).

[0006] In recent years, many studies have begun to look for biomarkers for predicting the effect of dietary weight loss. European patent application EP3464621 discloses a method of predicting the degree of weight loss obtained by applying one or more dietary interventions to an individual through small molecule RNA (miRNA). However, miRNA is greatly affected by ethnic and environmental factors, and expression may vary in different populations. Studies in the past 10 years have suggested that individual differences in weight loss may depend in part on the composition of the intestinal flora (Van Hul & Cani, 2023). Studies have suggested that intestinal flora is an important predictor of an individual's weight loss trajectory (Jie et al., 2021; Roager & Christensen, 2022). The Chinese patent with publication number CN112980945 discloses a method of using a neural network model to predict the weight loss effect of low-carb diet intervention by the relative abundance of Bacteroidetes before weight loss. However, the patent may be limited by the small sample size, and the prediction model is established without dividing the training set and the validation set. Only the baseline relative abundance of BacteroidaceaeBacteroides was found to be positively correlated with the weight loss effect, and the AUC value of the prediction model established by this bacterium was 73.2%, and the prediction ability needs to be improved. In addition, due to the limited amount of fecal samples, fecal metabolome testing was not performed.

[0007] In view of this, there is still a need to find biomarkers suitable for the Chinese population and to establish more accurate prediction models for individual weight loss effects. Summary of the invention

[0008] The present invention aims to screen out biomarkers related to intestinal flora and fecal metabolites in overweight or obese patients after dietary intervention. Finally, 9 biomarkers significantly related to weight loss effects were screened out, and combined with traditional clinical weight loss effect prediction models, existing weight loss strategies can be optimized. The present invention provides a scientific basis for evaluating the weight loss efficacy of dietary intervention and formulating personalized weight loss plans, and promotes the application of precision medicine in obesity management. In addition, the present invention also provides new ideas for the research and development of weight loss-related drugs.

[0009] According to a first aspect of the present invention, a method for screening intestinal flora and metabolite biomarkers for weight loss effects of dietary intervention in overweight or obese patients is provided, comprising the following steps:

[0010] (1) Collect baseline stool samples from patients for metagenomic sequencing, intestinal microbiome analysis, and fecal metabolome testing;

[0011] (2) The data of intestinal flora abundance and fecal metabolite concentrations were preprocessed. For missing values ​​below the detection limit, one-tenth of the minimum detection value of the analyte was used to fill in the missing values, and the concentration of each marker was log-transformed;

[0012] (3) Use a covariate-adjusted conditional logistic model to screen biomarkers associated with weight loss. The covariates in the model include age, sex, whether a healthy low-carbohydrate diet intervention is used, whether a time-restricted diet intervention is used, and baseline weight;

[0013] (4) The LASSO-logistic regression model with corrected confounding factors was used to reduce the dimensionality of the relevant biomarkers, and a ten-fold cross-validation was performed at each screening step, and finally a biomarker combination significantly associated with the weight loss effect after energy-reducing diet intervention was obtained; the biomarker combination included 9 biomarkers, and the 9 biomarkers were 6 bacterial species and 3 metabolites; the 6 bacterial species were: Actinomyces sp.ICM47, Blautia sp.AF19-10LB, Lachnospira sp.NSJ-4, Mediterraneibacter glycyrrhizinilyticus, Dysosmobacterwelbionis, and Bacteroides nordii. nordii]; the three metabolites are vanillic acid, hyocholic acid and isocaproic acid; the abundance of the six species was reduced in the weight loss group compared with the control group that did not lose weight; the concentrations of vanillic acid and hyocholic acid among the three metabolites were reduced in the weight loss group compared with the control group that did not lose weight, and the concentration of isocaproic acid among the three metabolites was increased in the weight loss group compared with the control group that did not lose weight.

[0014] Preferably, in step (1), the shotgun method is used for metagenomic sequencing and intestinal microbiome analysis.

[0015] Preferably, in step (1), the fecal metabolites are detected by ultra-high performance liquid chromatography-mass spectrometry.

[0016] According to another aspect of the present invention, a biomarker screening system for intestinal flora and metabolites for weight loss effects of overweight or obese patients after energy-reducing diet intervention is provided, comprising:

[0017] Data acquisition module: collect baseline stool samples from patients for metagenomic sequencing, intestinal microbiome analysis, and stool metabolome testing;

[0018] Data preprocessing module: preprocess the intestinal flora abundance and fecal metabolite concentration data. For missing values ​​below the detection limit, one-tenth of the minimum detection value of the analyte is used to fill in the missing values, and the concentration of each marker is log-transformed.

[0019] Candidate biomarker screening module: Use a covariate-corrected conditional logistic model to screen biomarkers related to weight loss effects. The covariates of the model include age, gender, whether a healthy low-carbohydrate diet intervention is adopted, whether a time-restricted diet intervention is adopted, and baseline weight;

[0020] Machine learning dimension reduction and biomarker screening module: The above candidate biomarkers were reduced in dimension using the LASSO-logistic regression model with correction for confounding factors. Each step of screening was subjected to ten-fold cross validation, and finally a biomarker combination significantly associated with the weight loss effect after energy-reducing diet intervention was obtained: the biomarker combination included 9 biomarkers, and the 9 biomarkers were 6 bacterial species and 3 metabolites; the 6 bacterial species were: Actinomyces sp.ICM47, Blautia sp.AF19-10LB, Lachnospira sp.NSJ-4, Mediterraneibacterglycyrrhizinilyticus, Dysosmobacter welbionis, and Bacteroides nordii; the 3 metabolites were: Vanillic acid The abundance of the six bacterial species was reduced in the weight-loss group compared with that in the control group that did not lose weight; the concentrations of vanillic acid and hydroxycholic acid among the three metabolites were reduced in the weight-loss group compared with that in the control group that did not lose weight, and the concentration of isocaproic acid among the three metabolites was increased in the weight-loss group compared with that in the control group that did not lose weight.

[0021] According to another aspect of the present invention, there is provided an application of intestinal flora and metabolite biomarkers in constructing a prediction model for the weight loss effect of dietary intervention in overweight or obese patients, wherein at least one of the screened intestinal flora and metabolite biomarkers is added to the weight loss effect prediction model after traditional clinical intervention to reconstruct a new weight loss effect prediction model;

[0022] The biomarkers are 6 bacterial species and 3 metabolites; the 6 bacterial species are: Actinomyces sp.ICM47, Blautia sp.AF19-10LB, Lachnospira sp.NSJ-4, Mediterraneibacterglycyrrhizinilyticus, Dysosmobacter welbionis and Bacteroides nordii; the 3 metabolites are: vanillic acid, hyocholic acid and isocaproic acid;

[0023] The abundance of the six bacterial species was reduced in the weight loss group compared with the control group that did not lose weight; the concentrations of vanillic acid and hydroxycholic acid among the three metabolites were reduced in the weight loss group compared with the control group that did not lose weight, and the concentration of isocaproic acid among the three metabolites was increased in the weight loss group compared with the control group that did not lose weight.

[0024] Preferably, the traditional prediction model is based on age, gender, intervention measures and baseline initial weight factors; the new weight loss effect prediction model incorporates dietary characteristic indicators on the basis of the traditional prediction model, and then further incorporates the newly identified biomarker combination as a predictor into the model; the predictive performance of the model is evaluated and compared by the area under the receiver operating characteristic curve and the DeLong test.

[0025] Preferably, the dietary characteristic indicators include baseline total energy intake and baseline total dietary fiber intake.

[0026] In summary, the above technical solutions conceived by the present invention have the following technical advantages compared with the prior art:

[0027] (1) Based on high-quality clinical research: The present invention is based on a randomized controlled intervention study on the effects of a healthy low-carbohydrate diet and a time-restricted diet on the improvement of body weight and blood sugar in overweight or obese people. This study minimized potential bias by randomly assigning subjects, ensuring the reliability and objectivity of the results, and providing a solid scientific basis for the present invention.

[0028] (2) The biomarker screening process is rigorous and comprehensive: First, the innovative use of metagenomics and metabolomics combined analysis can overcome the limitations of single-omics research to a certain extent, help to fully understand the relationship between microorganisms, metabolites and diseases, and identify potential biomarkers and therapeutic targets. Secondly, patients are grouped according to whether they have achieved clinically significant weight loss, and their intestinal microorganisms and fecal metabolites are analyzed. Then, compared with the method of screening biomarkers using only conditional logistic models, screening using combined covariate-corrected conditional logistic models, machine learning dimensionality reduction technology, and cross-validation methods can take into account the collinearity between biomarkers, retain important features, and improve the stability and reliability of the results. Finally, through the above screening methods, 9 biomarkers significantly associated with weight loss effects were screened. These markers provide a scientific basis for the formulation of personalized weight loss strategies in the future and promote the application of precision medicine in obesity management.

[0029] (3) Significantly improve the performance of the prediction model and have broad clinical application potential: The present invention constructs a prediction model based on the above-screened biomarker combination. By evaluating and comparing the prediction performance of the model, it is found that the biomarker combination significantly improves the prediction accuracy of the traditional clinical prediction model. In addition, the prediction model has good interpretability, verifiability and clinical practicality, can provide personalized intervention suggestions for obese patients, improve the success rate of weight loss, and has broad clinical application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 The clinical experimental design and the statistical methods involved are the sources of the present invention. Biomarkers significantly correlated with the weight loss effect of 12-week dietary intervention were screened through this process, and a prediction model for the weight loss effect of 12-week dietary intervention was established based on this.

[0031] Figure 2 The clinical experimental design and weight loss effect of 12-week dietary intervention: A. The research design of 12-week dietary intervention, which provides subjects with an isoenergetic diet with about 25% energy reduction, male: 1600kcal / day, female 1300kcal / day; B. The line graph shows the weight changes of the four groups during the 12-week intervention. Compared with the baseline, the weight of the subjects after the 12-week dietary intervention was significantly reduced. The present invention is divided into groups according to whether the weight loss effect of the 12-week dietary intervention reaches clinically significant weight loss (5% of own body weight, calculated by the formula weight loss percentage = weight loss kg / initial body weight * 100%). The response group is a weight loss that achieves clinical significance, that is, the weight loss percentage ≥5%, and the non-response group is a weight loss that does not achieve clinical significance (i.e. <5%).

[0032] Figure 3The conditional logistic model was used to analyze the difference between the baseline intestinal flora species level in the response group and the non-response group and calculate the difference multiple between the two groups. It is displayed in the form of a volcano chart, with the horizontal axis being the difference multiple and the vertical axis being the negative logarithm of the P value. When the difference multiple is 0, the right side of the vertical dotted line is the level of intestinal flora species that is significantly upregulated in the response group, and the left side is the level of intestinal flora species that is significantly downregulated. The horizontal dotted line represents the significant P value, the upper part represents the level of intestinal flora species that has changed significantly, and the lower part represents the level of intestinal flora species that has no significant difference.

[0033] Figure 4 The conditional logistic model was used to analyze the differences between the baseline fecal metabolites in the response group and the non-response group and to calculate the difference fold between the two groups. It is presented in the form of a volcano plot, with the horizontal axis being the difference fold and the vertical axis being the negative logarithm of the P value. When the difference fold is 0, the right side of the vertical dotted line is the fecal metabolite that is significantly upregulated in the response group, and the left side is the fecal metabolite that is significantly downregulated. The horizontal dotted line represents the significant P value, with the above representing the fecal metabolites with significant changes and the below representing the fecal metabolites with no significant differences.

[0034] Figure 5 The penalty parameter lambda in LASSO regression is selected based on 10-fold cross validation. Figure 5 A in the figure shows the changing trajectory of the coefficients of each independent variable in the model as lambda changes. Figure 5 Figure B shows the lambda minimum and lambda within 1 standard error found based on the cross-validated mean square error.

[0035] Figure 6 The receiver operating characteristic curve (ROC) analysis was used to compare the newly constructed prediction model with the traditional model that only included age, gender, whether the intervention method was implemented, and baseline weight, and the model that added baseline dietary characteristics indicators for the prediction of the weight loss effect of the 12-week dietary intervention. The prediction efficiency of the traditional regression model was 0.672 (95% CI 0.557, 0.788); when dietary characteristics indicators (baseline total energy intake and baseline total dietary fiber intake) were added, the prediction efficiency of the prediction model reached 0.722 (95% CI 0.614, 0.830); after further adding 9 newly identified biomarkers as predictors, the prediction efficiency was significantly improved to 0.939 (95% CI 0.891, 0.988). DETAILED DESCRIPTION

[0036] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0037] Figure 1 The clinical trial design and the statistical methods involved are the sources of the present invention. The biomarkers significantly correlated with the weight loss effect of 12-week dietary intervention were screened through this process, and a prediction model for the weight loss effect of 12-week dietary intervention was established based on this.

[0038] Example 1

[0039] 1. Experimental Methods

[0040] 1 Research subjects

[0041] The subjects of this study were from a prospective clinical intervention study conducted by the School of Public Health of Huazhong University of Science and Technology: the control group, time-restricted diet group, healthy low-carbohydrate diet group, and healthy low-carbohydrate diet combined with time-restricted diet intervention group in the "Randomized controlled intervention study on the improvement of weight and blood sugar in overweight or obese people with a healthy low-carbohydrate diet and time-restricted diet" (hereinafter referred to as RCT). This study complies with the ethical guidelines of the Declaration of Helsinki, was approved by the Ethics Committee of Tongji Medical College of Huazhong University of Science and Technology, and was registered in advance with the China Clinical Trial Registry (Clinical Trial Approval Number: ChiCTR2200056363). The subjects voluntarily participated in this study and signed a written informed consent notice. The trial period is from March 2022 to April 2023. The clinical trial design ideas of this study are as follows. Figure 2 shown.

[0042] 2 Diagnostic criteria

[0043] The diagnosis of overweight / obesity is based on the diagnostic criteria of the "Guidelines for the Prevention and Treatment of Overweight and Obesity in Chinese Adults" (2003). Overweight: 24.0 kg / m 2 ≤BMI<28.0kg / m 2 , Obesity: BMI ≥ 28.0 kg / m 2 .

[0044] 2.1 Inclusion criteria

[0045] 1) Age 20-60 years old;

[0046] 2) Overweight or obese: Body mass index ≥ 24 kg / m 2 ;

[0047] 3) Engage in light physical activity.

[0048] 2.2 Exclusion criteria

[0049] 1) Working night shift (all night shift: 8:00 pm to 8:00 am the next day) more than once a week

[0050] 2) Fast regularly (fast for 15 hours per day or complete 12 24-hour fasts in the past year)

[0051] 3) Those who have adopted a carbohydrate-restricted diet in the past three months, or have taken a special or prescribed diet for other reasons

[0052] 4) Significant weight change in the past 3 months (≥3kg)

[0053] 5) Currently participating in a weight loss or weight management program or taking any medications known to affect appetite

[0054] 6) Suffering from severe digestive system diseases, history of gastrointestinal surgery, impaired nutrient absorption or eating disorders

[0055] 7) Suffering from diabetes or severe cardiovascular, kidney, liver, lung or nervous system diseases, acute or chronic diseases, infectious diseases, cancer, etc.

[0056] 8) Underwent surgery in the past 3 months or plans to undergo surgery within 3 months

[0057] 9) Taking antibiotics or any drugs that may affect the study endpoints such as blood sugar, blood lipids and body composition metabolism (such as anti-diabetic drugs, steroids, beta-blockers, adrenergic stimulants, etc.) in the past 3 months

[0058] 10) Have you taken dietary supplements (including n-3 unsaturated fatty acids, prebiotics, probiotics, etc.) in the past three months?

[0059] 11) Suffering from severe mental impairment or unable to understand or follow instructions

[0060] 12) Daily high-intensity physical activity

[0061] 13) Heavy drinking habit (more than 100g of liquor or more than 5 bottles of beer per day) or heavy smoking habit (more than 2 packs per day)

[0062] 14) Women who are pregnant, planning to become pregnant within three months, or breastfeeding

[0063] 15) Participating in other clinical trials

[0064] 3 Clinical trial research methods

[0065] The research method followed the RCT study protocol.

[0066] 3.1 Study Groups

[0067] First, the volunteers were randomly divided into 4 groups at a ratio of 1:1:1:1 by computer-generated random numbers, and the randomization was performed by personnel not participating in the study.

[0068] (1) Control group: Volunteers were provided with a normal diet prepared in accordance with the Dietary Guidelines for Chinese Residents without restricting their total daily energy intake.

[0069] (2) Healthy low-carbohydrate diet group: A healthy low-carbohydrate diet was provided without restrictions on eating time.

[0070] (3) Time-restricted diet group: provided with normal meals and a time-restricted diet.

[0071] (4) Combined intervention group: provided with a healthy low-carbohydrate diet and a time-restricted diet.

[0072] 3.2 Dietary intervention program

[0073] The intervention is mainly to provide meals (on weekdays) and supplemented by dietary guidance (on weekends and rest days).

[0074] a) General diet: diet prepared in accordance with the Dietary Guidelines for Chinese Residents, without limiting total energy intake;

[0075] b) Healthy low-carbohydrate diet: Whole grains are the main source of carbohydrates, accounting for 35% of the total daily energy, and fat and protein provide 45-50% and 15-20% of the energy respectively. The fat is mainly unsaturated fatty acids, and the protein is mainly plant protein with a certain proportion of animal protein.

[0076] c) Time-restricted diet: 14 hours of fasting and 10 hours of eating. Three meals are distributed regularly, and daily energy intake is concentrated in the eating period. During the fasting period, no high-calorie food can be eaten. Volunteers are recommended to drink water, or consume 1-2 servings of chewing gum, mints, tea, black coffee, 0-calorie beverages, etc., each with less than 4kcal.

[0077] d) Non-time-restricted diet: that is, there is no restriction on eating time. In addition to the three meals, volunteers can eat snacks such as fruits, but they need to record it.

[0078] 3.3 Research Plan

[0079] According to the inclusion and exclusion criteria of the research subjects, 96 overweight or obese patients who met the inclusion and exclusion criteria were selected. After signing the informed consent, baseline measurements including 24-hour dietary records, questionnaires, biological measurements and physical examinations were performed, and blood and stool samples were collected. After the baseline measurement is completed and the randomization is completed (4 groups, 24 people in each group), the subjects will receive a dietary overview booklet. During the formal intervention period, all volunteers will be provided with three meals a day on weekdays and dietary guidance on rest days. The intervention will last for 12 weeks, and endpoint measurements (consistent with baseline measurements) will be performed after the intervention.

[0080] 3.4 Clinical data collection

[0081] (1) Body weight was measured using a standard electronic weighing scale at baseline (before intervention) and at the end point (after intervention). The present invention uses the baseline and end point body weight measurement data.

[0082] (2) Stool samples were collected at baseline (before intervention) and endpoint (after intervention) by the subjects themselves.

[0083] (3) Shotgun metagenomic sequencing and gut microbiome analysis: Microbial DNA was extracted from 758 fecal samples using the Mag Pure Fast Stool DNAKFKit B kit, and the extracted DNA was subjected to shotgun metagenomic 100bp paired-end (PE) sequencing using the BGI platform. The quality control process of the whole-genome shotgun sequencing data used KneadData (version 0.7.2), Trimmomatic (version 0.33), and Bowtie2 (version 2.3.4.3). Human reads and rDNA reads were filtered by aligning the reads to the human reference genome (GRCh37) database and the SILVA 128 database. After quality control, an average of 36.9 million (minimum 14.8 million, maximum 54.1 million) high-quality reads were obtained for each sample. Taxonomic features were determined by MetaPhlan (version 3.0.3), and microbial functional features, including MetaCyc pathways and enzyme commission gene families, were determined by HUMAnN (version 3.0.1). Microbial species with a relative abundance lower than 0.01% and not present in more than 90% of the samples were excluded from subsequent analyses. Finally, a total of 205 microbial species were included.

[0084] (4) Fecal metabolome detection: Ultra-high performance liquid chromatography-mass spectrometry was used for detection. A total of 222 intestinal flora-related fecal metabolites were identified. After excluding metabolites with a missing ratio of more than 30% or a coefficient of variation >30% in all samples, a total of 217 fecal microbial host co-metabolites were quantified and included in the data analysis. In addition, the missing values ​​of the metabolites included in the analysis were estimated using 1 / 10 of the minimum detectable value. The 217 metabolites include 40 amino acids, 42 fatty acids, 19 carbohydrates, 34 bile acids, 6 indoles, 10 short-chain fatty acids, and other categories of metabolites, including benzoic acid, organic acids, phenols, etc.

[0085] (5) The present invention only uses baseline intestinal microbiome species abundance and baseline fecal metabolite detection data.

[0086] 2. Statistical analysis and results

[0087] 1. After the 12-week intervention, a total of 88 subjects completed the 12-week dietary intervention. After excluding subjects with poor compliance and missing intestinal flora measurement data, a total of 84 subjects' baseline intestinal flora metagenomic data, fecal metabolomics data, baseline dietary survey data, and clinical and physical examination data at baseline and endpoint were included in the present invention. The 12-week dietary intervention produced a good weight loss effect on all subjects ( Figure 2 B in the figure). Previous studies have shown that a weight loss percentage of 5% is considered to be clinically significant, and the calculation formula is weight loss percentage = weight loss kg / initial body weight*100%. Therefore, in order to facilitate the promotion of the invention, the present invention uses whether the weight loss effect of the 12-week dietary intervention has achieved clinically significant weight loss (5% of one's own body weight) as the standard for screening biological indicators of the present invention and as a predictive outcome. And according to whether the weight loss effect of the 12-week dietary intervention has achieved clinically significant weight loss, the responding group has achieved clinically significant weight loss, that is, the weight loss percentage is ≥5%, while the non-responding group has not achieved clinically significant weight loss (that is, <5%).

[0088] 2. Biomarkers significantly associated with the weight loss effect of the 12-week dietary intervention were screened using common epidemiological statistical methods and machine learning algorithms. All the following analyses were completed in SAS 9.4 (SAS Institute, Cary, NC) and R software 4.3.1. The analysis process is detailed as follows:

[0089] 2.1 Normalization of Potential Biomarker Data

[0090] For the abundance data of 205 intestinal flora species and the concentration data of 217 fecal intestinal flora-related metabolites, the missing values ​​below the detection limit will be filled with one-tenth of the minimum detection value of the analyte. Since the above markers are usually skewed, the concentrations must be log-transformed before analysis. The above markers will be used as potential biomarkers for the next screening module

[0091] 2.2 Screening of candidate biomarkers

[0092] Multivariate regression analysis was used to screen potential biomarkers, and 0.05 was set as the statistically significant threshold. Specifically, a conditional logistic model with covariate correction was used to screen biomarkers related to weight loss effect one by one. The covariates of the model included age, gender, whether a healthy low-carbohydrate diet intervention was adopted, whether a time-restricted diet intervention was adopted, and baseline weight. The analysis results are as follows Figure 3 and Figure 4As shown in the figure, 6 bacterial species and 10 fecal metabolites were found to be significantly different between those who achieved clinically significant weight loss and those who did not. Compared with the control group that did not achieve clinically significant weight loss, the baseline abundance of 6 bacterial species in the group that achieved clinically significant weight loss was significantly downregulated, mainly including Actinomyces sp.ICM47, Blautia sp.AF19-10LB, Lachnospira sp.NSJ-43, Mediterraneibacterglycyrrhizinilyticus, Dysosmobacter welbionis, and Bacteroides nordii. Among the 10 fecal metabolites, the concentrations of 8 metabolites that achieved clinically significant weight loss were significantly upregulated, including β-D-galactofuranose (Beta_D_Fucose), N-acetylneuraminic acid (N-Acetylneuraminic acid), N-acetyl-L-aspartic acid (N-Acetyl-L-aspartic acid), L-malic acid (L-Malic acid), dodecanoic acid (Dodecanoic acid), myristic acid (Myristic acid), pyroglutamic acid (Pyroglutamic acid) and isocaproic acid (Isocaproic acid), and 2 metabolites were significantly downregulated (Vanillic acid and hydroxycholic acid). Finally, 16 biological metabolites associated with weight loss were screened and further screened as candidate biomarkers through machine learning algorithms.

[0093] 2.3 Machine Learning Dimensionality Reduction Screening and Verification

[0094] Metagenomics and metabolomics data are characterized by high dimensionality and high collinearity. The Least Absolute Shrinkage and Selection Operator (LASSO) regression model is a regularization method used to avoid overfitting and achieve dimensionality reduction. This design uses the LASSO-logistic regression method with correction for confounding factors to reduce data dimensionality. Each step of screening is repeated with cross-validation, and the above 16 candidate biomarkers are further screened.

[0095] To prevent overfitting, the LASSO penalty parameter, lambda, is determined by 10-fold cross validation and using the cross-validated mean square error as the loss function. Specifically, all samples are randomly divided into 10 parts of equal size. For each different lambda value, 9 of the 10 parts are used as training data sets to fit the model parameters, and the remaining 1 part is used as a validation data set to calculate the mean square error of the parameters. This process is repeated 10 times for each lambda, and different parts are used as validation data sets in turn. The mean square error of the model can be estimated each time, and the average of the 10 mean square errors is taken after 10-fold cross validation. The minimum mean square error of cross validation corresponds to the minimum value of lambda. In order to obtain a prediction model with a relatively concise number of predictors, we choose lambda.min as the penalty parameter. Figure 5 A in the figure shows the change trajectory of each independent variable coefficient with the lambda value in a LASSO regression. The ordinate is the value of the independent variable coefficient, the lower abscissa is log(lambda), and the upper abscissa is the number of variables with non-zero coefficients in the model for each lambda. Since the covariates need to be corrected in the model, the covariates are not penalized, and all covariates are assigned non-zero regression coefficients and are always retained in the model. When the variable coefficient is equal to zero, it means that the variable is not included in the model when corresponding to the lambda parameter. Figure 5 B in the figure shows how the mean square error of the model changes with lambda in a LASSO regression, and the optimal model is selected based on the mean square error of the cross-validation. Considering that different ways of segmenting the data set in the cross-validation will affect the results and easily show high variance, the LASSO regression with 10-fold cross-validation is repeated 100 times, and the regression coefficient is non-zero and the reproducibility is 100% selected as the biomarker.

[0096] The analysis screened for a biomarker combination that was significantly associated with the weight loss effect of the 12-week dietary intervention; the biomarker combination included 9 biomarkers, composed of 6 bacterial species (Actinomyces sp. ICM47, Blautia sp. AF19-10LB, Lachnospira sp. NSJ-4, Mediterraneibacter glycyrrhizinilyticus, Dysosmobacter welbionis, and Bacteroides nordii) and 3 fecal metabolites (vanillic acid, hyocholic acid, and isocaproic acid).

[0097] This analysis can be performed using the R package “glmnet”.

[0098] 3. Establishment and evaluation of prediction models

[0099] In the present invention, the biomarker combination of the weight loss effect of the 12-week dietary intervention obtained by the above screening is applied to the optimization and improvement of the weight loss effect prediction model after the traditional clinical intervention. The traditional prediction model of the weight loss effect after the intervention reported at present is mainly established by factors such as age, gender, intervention measures and baseline initial weight. Improved model 1 incorporates dietary characteristic indicators (baseline total energy intake and baseline total dietary fiber intake) on the basis of the traditional prediction model. Improved model 2, on the basis of improved model 1 (traditional predictors and dietary characteristic indicators), further incorporates the newly identified 9 biomarker combinations as predictors into the model. The present invention uses the receiver operating characteristic curve (ROC) to evaluate the accuracy of the prediction model. The horizontal axis X axis represents (1-specificity), that is, the false positive rate; the vertical axis Y axis represents sensitivity, that is, the true positive rate. The area under the curve Area UnderCurve (AUC) is used to indicate the accuracy of the prediction, with a value between 0 and 1. The larger the AUC, the higher the accuracy of the prediction. AUC can be obtained by the trapezoidal rule. Usually, an AUC of 0 indicates that the method has no ability to distinguish between responders and non-responders. 0.7-0.8 is considered to have good discrimination, and above 0.8 is considered to have excellent discrimination. When evaluating the predictive ability of a binary logistic regression model, the value of C-statistic is equal to AUC. The DeLong test was used to compare the statistical differences in AUC between models. In this design, the traditional prediction model for weight loss effect is composed of age, gender, intervention measures, and baseline initial weight, with a C-statistic of 0.672 (95% CI 0.557, 0.788). When dietary trait indicators (baseline total energy intake and baseline total dietary fiber intake) are added, the C-statistic increases to 0.722 (95% CI 0.614, 0.830) ( Figure 6 ). Nine newly identified biomarkers were added as predictors to the above model, which further improved the C-statistic to 0.939 (95% CI 0.891, 0.988). The DeLong test showed that the predictive efficiency of the newly established prediction model was significantly improved compared with the traditional model and clinical model: ΔAUC was 26.7% and 21.7%, respectively (P<0.001).

[0100] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for screening biomarkers of intestinal flora and metabolites for weight loss effects of dietary intervention in overweight or obese patients, characterized in that: The following steps are involved: (1) Collect baseline stool samples from patients for metagenomic sequencing, intestinal microbiome analysis, and fecal metabolome testing; (2) The data of intestinal flora abundance and fecal metabolite concentrations were preprocessed. For missing values ​​below the detection limit, one-tenth of the minimum detection value of the analyte was used to fill in the missing values, and the concentration of each marker was log-transformed; (3) Use a covariate-adjusted conditional logistic model to screen biomarkers associated with weight loss. The covariates in the model include age, sex, whether a healthy low-carbohydrate diet intervention is used, whether a time-restricted diet intervention is used, and baseline weight; (4) The LASSO-logistic regression model with corrected confounding factors was used to reduce the dimensionality of the relevant biomarkers, and a ten-fold cross-validation was performed at each screening step, and finally a biomarker combination significantly associated with the weight loss effect after energy-reducing diet intervention was obtained; the biomarker combination included 9 biomarkers, and the 9 biomarkers were 6 bacterial species and 3 metabolites; the 6 bacterial species were: Actinomyces sp.ICM47, Blautia sp.AF19-10LB, Lachnospira sp.NSJ-4, Mediterraneibacter glycyrrhizinilyticus, Dysosmobacterwelbionis, and Bacteroides nordii. nordii]; the three metabolites are vanillic acid, hyocholic acid and isocaproic acid; the abundance of the six species was reduced in the weight loss group compared with the control group that did not lose weight; the concentrations of vanillic acid and hyocholic acid among the three metabolites were reduced in the weight loss group compared with the control group that did not lose weight, and the concentration of isocaproic acid among the three metabolites was increased in the weight loss group compared with the control group that did not lose weight.

2. The biomarker screening method according to claim 1, characterized in that: In step (1), the shotgun method was used for metagenomic sequencing and intestinal microbiome analysis.

3. The biomarker screening method according to claim 1, characterized in that: In step (1), fecal metabolites are detected by ultra-high performance liquid chromatography-mass spectrometry.

4. A biomarker screening system for intestinal flora and metabolites for weight loss effects of overweight or obese patients after energy-reducing diet intervention, characterized in that: include: Data acquisition module: collect baseline stool samples from patients for metagenomic sequencing, intestinal microbiome analysis, and stool metabolome testing; Data preprocessing module: preprocess the intestinal flora abundance and fecal metabolite concentration data. For missing values ​​below the detection limit, one-tenth of the minimum detection value of the analyte is used to fill in the missing values, and the concentration of each marker is log-transformed. Candidate biomarker screening module: Use a covariate-corrected conditional logistic model to screen biomarkers related to weight loss effects. The covariates of the model include age, gender, whether a healthy low-carbohydrate diet intervention is adopted, whether a time-restricted diet intervention is adopted, and baseline weight; Machine learning dimension reduction and biomarker screening module: The LASSO-logistic regression model with corrected confounding factors was used to reduce the dimension of the above candidate biomarkers. Each step of the screening was subjected to ten-fold cross-validation, and finally a biomarker combination significantly associated with the weight loss effect after energy-reducing diet intervention was obtained: the biomarker combination included 9 biomarkers, and the 9 biomarkers were 6 bacterial species and 3 metabolites; the 6 bacterial species were: Actinomyces p. ICM47, Blautia sp. AF19-10LB, Lachnospira sp. NSJ-4, Mediterraneibacter glycyrrhizinilyticus, Dysosmobacter welbionis, and Bacteroides nordii; the 3 metabolites were: Vanillic acid The abundance of the six bacterial species was reduced in the weight-loss group compared with that in the control group that did not lose weight; the concentrations of vanillic acid and hydroxycholic acid among the three metabolites were reduced in the weight-loss group compared with that in the control group that did not lose weight, and the concentration of isocaproic acid among the three metabolites was increased in the weight-loss group compared with that in the control group that did not lose weight.

5. Application of intestinal flora and metabolite biomarkers in constructing a prediction model for weight loss effects of dietary intervention in overweight or obese patients, characterized in that: Adding at least one of the screened intestinal flora and metabolite biomarkers to the weight loss effect prediction model after traditional clinical intervention to reconstruct a new weight loss effect prediction model; The biomarkers are 6 bacterial species and 3 metabolites; the 6 bacterial species are: Actinomyces p. ICM47, Blautia sp. AF19-10LB, Lachnospira sp. NSJ-4, Mediterraneibacter glycyrrhizinilyticus, Dysosmobacter welbionis and Bacteroides nordii; the 3 metabolites are: vanillic acid, hyocholic acid and isocaproic acid; The abundance of the six bacterial species was reduced in the weight loss group compared with the control group that did not lose weight; the concentrations of vanillic acid and hydroxycholic acid among the three metabolites were reduced in the weight loss group compared with the control group that did not lose weight, and the concentration of isocaproic acid among the three metabolites was increased in the weight loss group compared with the control group that did not lose weight.

6. The use according to claim 5, characterized in that The traditional prediction model is based on age, gender, intervention measures and baseline initial weight factors; the new weight loss effect prediction model incorporates dietary characteristic indicators on the basis of the traditional prediction model, and then further incorporates the newly identified biomarker combination as a predictor into the model; the predictive performance of the model is evaluated and compared by the area under the receiver operating characteristic curve and the DeLong test.

7. The use according to claim 6, characterized in that The dietary characteristic indicators include baseline total energy intake and baseline total dietary fiber intake.

Images