Method and system for clinical marker screening after metabolic response profiling and applications

Abstract

The application provides a clinical marker screening method and system after metabolic response typing and application. Specifically, according to the change of metabolic indexes of normal weight people and overweight / obese people before and after meals, three clusters based on respiratory quotient (RQ) are found; the fasting fat oxidation capacity, postprandial metabolic response and postprandial blood glucose response of people belonging to different clusters also have different dynamic change patterns. In addition, key biomarkers distinguishing the three clusters are also found; the prediction model constructed by using the key biomarkers can accurately divide the subjects into three RQ clusters and can be used for predicting the results of weight loss intervention and life intervention guidance. The system constructed according to the clinical markers of the application has high sensitivity and accuracy, the indexes used are easy to obtain, and has good clinical application prospect.

Description

Technical Field

[0001] This invention relates to multi-omics analysis and energy metabolism prediction and typing, specifically to methods for screening clinical biomarkers and training models after metabolic response typing, and their applications. Background Technology

[0002] Obesity and metabolic-related diseases are prevalent globally, posing a serious public health problem. Routine clinical assessments of metabolic status primarily rely on static indicators such as fasting blood glucose, blood lipids, and single-point liver function tests. However, these static tests only reflect the body's metabolic level under resting conditions and are insufficient to reveal an individual's dynamic response to dietary stimuli and their postprandial homeostasis recovery ability, i.e., metabolic flexibility. The standard Mixed Macronutrient Tolerance Test (MMTT) is a standardized metabolic homeostasis test that simulates daily life and can reflect an individual's metabolic flexibility. Metabolic flexibility, as an important indicator of the body's substrate utilization and switching ability, is closely related to the development of diseases such as insulin resistance, diabetes, and obesity; therefore, a more suitable dynamic assessment method is urgently needed.

[0003] To overcome the limitations of static indicators, researchers have gradually introduced whole-room indirect calorimetry (WRIC) and portable indirect calorimetry methods. These technologies can monitor the body's oxygen consumption (VO2) and carbon dioxide output (VCO2) in real time, and calculate energy expenditure (EE), respiratory quotient (RQ), and carbohydrate oxidation (CarbOx) and fat oxidation (FatOx) based on these parameters. These indicators are widely used in research to assess an individual's metabolic response under different dietary loads. Existing research results show that the body's metabolic response patterns to different macronutrients such as fat, carbohydrates, and proteins differ. For example, a high carbohydrate load mainly causes drastic changes in blood glucose and insulin levels, while a high fat load has a greater impact on lipid metabolism. This difference suggests that a single nutrient load is insufficient to comprehensively reflect an individual's metabolic flexibility, and dynamic assessment across nutrient dimensions is necessary to reveal population metabolic differences and guide individualized dietary interventions.

[0004] Current technologies have not yet established an integrated approach and system from "mixed macronutrient load - dynamic metabolic response sequence - population typing" to clinical application. At the same time, clinical practice lacks low-cost, easily promoted, and simplified discrimination tools that can be operated under caloric-free conditions, making it difficult to effectively link limited clinical biochemical indicators or continuous glucose monitoring (CGM) data with individualized weight management outcomes.

[0005] Therefore, there is an urgent need in this field for a clinical biomarker for population metabolic response typing, as well as its screening methods, systems, and applications. Summary of the Invention

[0006] The purpose of this invention is to provide a method, system, and application for screening clinical biomarkers after metabolic response typing.

[0007] In a first aspect, the present invention provides the use of a clinical biomarker or its detection reagent for preparing a diagnostic reagent or diagnostic kit, said diagnostic reagent or diagnostic kit being used for: i) classifying the metabolic response of a subject; ii) predicting weight changes in a subject after a weight loss intervention;

[0008] The clinical biomarkers include: (A) Any biomarker selected from A1-A8, or a combination thereof: (A1) Respiratory quotient; (A2) Blood glucose level; (A3) Carbohydrate oxidation rate (CarbOx); (A4) Fat oxidation rate (FatOx); (A6) Energy expenditure rate (EE); (A7) Free fatty acids (FFA); (A8) Insulin / glucagon; (B) Markers shown in B1 and / or B2: (B1) Postprandial total free fatty acids (FFA) AUC; (B2) Fasting gamma-glutamyl transferase (GGT); (D) A combination of one or more markers from A1 to A8 and one or more markers from B1 to B2.

[0009] In another preferred embodiment, the total free fatty acid AUC after a meal includes the total free fatty acid AUC ≥ 2 hours after a meal, such as the total free fatty acid (5-h FAA) AUC 5 hours after a meal.

[0010] In another preferred embodiment, the subjects to be tested include people of normal weight, overweight / obese people, people with prediabetes, people with diabetes, and people with other metabolic diseases.

[0011] In another preferred embodiment, the diagnostic reagent or diagnostic kit method is used to predict weight changes after a weight loss intervention in overweight / obese individuals and / or prediabetic individuals.

[0012] In another preferred embodiment, the metabolic response includes respiratory quotient type, postprandial substrate conversion capacity, glycemic regulation capacity, free fatty acid level, and insulin / glucagon ratio.

[0013] In another preferred embodiment, the “classification” refers to classifying the test subject into two or more respiratory quotient cluster types.

[0014] In another preferred embodiment, the “classification” refers to classifying the test subject into respiratory quotient cluster type 1 or non-respiratory quotient cluster type 1.

[0015] In another preferred embodiment, the term "classification" refers to classifying the test subject into respiratory quotient cluster type 1, respiratory quotient cluster type 2, and respiratory quotient cluster type 3.

[0016] In another preferred embodiment, the “classification” refers to classifying the test subject into subjects with strong substrate switching ability after undergoing a steady-state challenge within the organism, subjects with moderate substrate switching ability after undergoing a steady-state challenge within the organism, and subjects with impaired substrate switching ability after undergoing a steady-state challenge within the organism.

[0017] In another preferred embodiment, the homeostatic challenges within the body include eating and exercise.

[0018] In another preferred embodiment, the blood glucose regulation capability includes the rate at which postprandial blood glucose reaches its peak, the rate at which blood glucose returns to fasting levels, or the peak blood glucose level.

[0019] In another preferred embodiment, the “classification” refers to classifying the subjects to be tested into subjects whose blood glucose reaches its peak value fastest after undergoing a homeostatic challenge, subjects whose blood glucose recovers slowest after undergoing a homeostatic challenge, and subjects whose blood glucose peak value is the highest after undergoing a homeostatic challenge.

[0020] In another preferred embodiment, the “prediction” refers to predicting the weight loss effect of the subject after weight loss intervention.

[0021] In another preferred embodiment, the “prediction” refers to predicting whether the subject can achieve clinically significant weight loss, such as ≥5% or ≥10%.

[0022] In another preferred embodiment, the weight loss effect refers to a reduction / loss of mass / level / percentage of weight.

[0023] In another preferred example, the test objects belonging to cluster 1 have three or more of the following characteristics: (c1.1) Low fasting RQ value; (c1.2) The postprandial RQ value reached its peak at 140±20 min; (c1.3) The postprandial RQ value rises rapidly and significantly before gradually declining; (c1.4) Blood glucose reached its peak at 30±10 min; (c1.5) Postprandial blood glucose reached fasting levels at 180±20 min; (c1.6) Postprandial blood glucose rebounded after 240±20 min; (c1.7) The postprandial FFA reached its trough at 120±10 min; (c1.8) Postprandial FFA rebounds within 120-300 min; (c1.9) Compared with clusters 2 and 3, the postprandial insulin / glucagon ratio reached a significantly lowest value at 180±20 min; (c1.10) Strong substrate switching ability after accepting the challenge of homeostasis in the organism; (c1.11) Blood glucose reaches its peak rate the fastest; (c1.12) Weight loss was effective after the weight loss intervention.

[0024] In another preferred embodiment, "strong substrate switching ability after accepting steady-state challenges in the body" means that the fasting RQ is low, the RQ reaches its peak value at the fastest rate after accepting steady-state challenges in the body, and the recovery rate of the RQ after reaching its peak value is fast.

[0025] In another preferred embodiment, "the rate at which the RQ reaches its peak value after undergoing a steady-state challenge within the body is the fastest" means that, compared to the rate S2 at which the postprandial RQ reaches its peak value in cluster 2, the rate S1 at which the postprandial RQ reaches its peak value in cluster 1 satisfies S1 / S2 > 100%, such as > 101% or > 102%; or compared to the rate S3 at which the postprandial RQ reaches its peak value in cluster 3, the rate S1 at which the postprandial RQ reaches its peak value in cluster 1 satisfies S1 / S3 > 150%, such as > 170%, > 190%, > 210%, > 230%, preferably > 290%.

[0026] In another preferred embodiment, the phrase "fast recovery rate of RQ after reaching peak value after accepting steady-state challenge within the body" means that, compared with the recovery rate J2 of RQ after reaching peak value after accepting steady-state challenge within the body in cluster 2, the rate J1 of RQ after reaching peak value after accepting steady-state challenge within the body in cluster 1 satisfies J1 / J2 > 105%, such as > 110%, > 115%, preferably > 120%.

[0027] In another preferred embodiment, "the fastest rate of reaching peak blood glucose" means that, compared with the rate of reaching peak blood glucose G2 of cluster 2, the rate of reaching peak blood glucose G1 of cluster 1 satisfies G1 / G2 > 200%, such as > 230%, > 260%, preferably > 290%; or compared with the rate of reaching peak postprandial blood glucose S3 of cluster 3, the rate of reaching peak postprandial blood glucose S1 of cluster 1 satisfies G1 / G3 > 150%, such as > 160%, > 170%, > 180%, preferably > 190%.

[0028] In another preferred embodiment, "good weight loss effect after weight loss intervention" means that, compared with the weight loss effect W2 of cluster 2, the weight loss effect W1 of cluster 1 satisfies W1 / W2 > 100%, such as > 110%, > 120%, preferably > 130%; or compared with the weight loss effect W3 of cluster 3, the weight loss effect W1 of cluster 1 satisfies W1 / W3 > 100%, such as > 120%, > 140%, > 160%, > 180%, > 200%, preferably > 215%.

[0029] In another preferred embodiment, the test objects belonging to cluster 2 have three or more of the following characteristics: (c2.1) Low fasting RQ value; (c2.2) The postprandial RQ value reached its peak at 210±20 min; (c2.3) The postprandial RQ value rose rapidly and significantly and then remained at a high level; (c2.4) Postprandial blood glucose reached its peak at 60±10 min; (c2.5) Postprandial blood glucose reached fasting levels at 240±20 min; (c2.6) Postprandial FFA reaches its trough at 120-180 min; (c2.7) Postprandial FFA rebounds within 170-300 min; (c2.8) The postprandial insulin / glucagon ratio remained stable between 40 and 140 min; (c2.9) The 5-h FFA AUC level is the lowest compared to clusters 1 and 3; (c2.10) Substrate switching ability is generally poor after accepting the challenge of homeostasis within the organism; (c2.11) Slowest rate of blood glucose recovery; (c2.12) The weight loss effect after the weight loss intervention was generally small.

[0030] In another preferred embodiment, the "lowest 5-h FFA AUC level" means that, compared with the 5-h FFA AUC level K1 of cluster 1, the 5-h FFA AUC level K2 of cluster 2 satisfies K2 / K1 < 90%, such as < 80%; or compared with the 5-h FFA AUC level K3 of cluster 3, the 5-h FFA AUC level K2 of cluster 2 satisfies K2 / K3 < 90%, such as < 80%, preferably < 75%.

[0031] In another preferred embodiment, "general substrate switching ability after undergoing steady-state challenge in the body" means that the fasting RQ is low, the rate at which the RQ reaches its peak value after undergoing steady-state challenge in the body is slow, and the recovery rate of the RQ after reaching its peak value after undergoing steady-state challenge in the body is slow.

[0032] In another preferred embodiment, "the rate at which RQ reaches its peak value after undergoing a steady-state challenge within the body is slow" means that, compared to the rate S1 of RQ reaching its peak value after undergoing a steady-state challenge within cluster 1, the rate S2 of RQ reaching its peak value after undergoing a steady-state challenge within cluster 2 satisfies S2 / S1 < 95%, such as < 93%, < 91%, preferably < 90%; or compared to the rate S3 of RQ reaching its peak value after undergoing a steady-state challenge within cluster 3, the rate S2 of RQ reaching its peak value after undergoing a steady-state challenge within cluster 2 satisfies S2 / S3 > 150%, such as > 170%, > 190%, > 210%, > 230%, preferably > 250%.

[0033] In another preferred embodiment, the phrase "slow recovery rate of RQ after reaching peak value after accepting steady-state challenge within the body" means that, compared with the recovery rate J1 of RQ after reaching peak value after accepting steady-state challenge within the body in cluster 1, the rate J2 of RQ after reaching peak value after accepting steady-state challenge within the body in cluster 2 satisfies J2 / J1 < 95%, such as < 90% or < 85%.

[0034] In another preferred embodiment, "slowest blood glucose recovery rate" means that, compared with the rate E1 of postprandial blood glucose recovery to fasting level in cluster 1, the rate E2 of postprandial blood glucose recovery to fasting level in cluster 2 satisfies E2 / E1 < 95%, such as < 90%, < 85%, preferably < 80%; or compared with the rate E3 of postprandial blood glucose recovery to fasting level in cluster 3, the rate E2 of postprandial blood glucose recovery to fasting level in cluster 2 satisfies E2 / E3 < 95%, such as < 90%.

[0035] In another preferred embodiment, "the weight loss effect after weight loss intervention is generally good" means that, compared with the weight loss effect W1 of cluster 1, the weight loss effect W2 of cluster 2 satisfies W2 / W1 < 95%, such as < 90%, < 85%, < 80%, preferably < 75%; or compared with the weight loss effect W3 of cluster 3, W2 / W3 > 100%, such as > 120%, > 140%, preferably > 160%.

[0036] In another preferred example, the test objects belonging to cluster 3 have three or more of the following characteristics: (c3.1) High fasting RQ value; (c3.2) The postprandial RQ value reached its peak at 140±20 min; (c3.3) The postprandial RQ value rises slowly and then remains at a high level; (c3.4) Postprandial blood glucose reached its peak at 60±10 min; (c3.5) Postprandial blood glucose reaches fasting level at 240±20 min; (c3.6) Postprandial blood glucose rebounded after 240±20 min; (c3.8) Postprandial FFA rebounds within 120-300 min; (c3.9) Compared to cluster 1 and cluster 2, the postprandial insulin / glucagon ratio remained high up to 180±20 min; (c3.10) Fasting GGT levels were highest compared to clusters 1 and 2; (c3.11) Impaired postprandial substrate switching ability after undergoing a challenge to homeostasis within the body; (c3.12) Highest blood glucose peak; (c3.13) The weight loss effect after weight loss intervention is poor.

[0037] In another preferred embodiment, "highest fasting GGT level" means that, compared with the fasting GGT level Q1 of cluster 1, the fasting GGT level Q3 of cluster 3 satisfies Q3 / Q1 > 120%, such as > 140%, > 160%, preferably > 180%; or compared with the fasting GGT level Q2 of cluster 2, the fasting GGT level Q3 of cluster 3 satisfies Q3 / Q2 > 140%, such as > 180%, > 220%, preferably > 260%.

[0038] In another preferred embodiment, "impaired substrate switching ability after undergoing a steady-state challenge in the body" means that the fasting RQ value is the highest, the rate at which the RQ reaches its peak value after undergoing a steady-state challenge in the body is the slowest, and the RQ fluctuation is gentle after undergoing a steady-state challenge in the body.

[0039] In another preferred embodiment, "highest fasting RQ value" means that, compared with the fasting RQ value K1 of cluster 1, the fasting RQ value K3 of cluster 3 satisfies K3 / K1>105%; or compared with the fasting RQ value K2 of cluster 2, the fasting RQ value K3 of cluster 3 satisfies K3 / K2>105%.

[0040] In another preferred embodiment, the phrase "RQ fluctuations are smooth after undergoing a steady-state challenge within the body" means that the peak value P1 of RQ after undergoing a steady-state challenge within the body and the fasting RQ level P0 satisfy 95%≤P1 / P0≤105%, such as 97%≤P1 / P0≤103%.

[0041] In another preferred embodiment, "highest blood glucose peak" means that, compared with the postprandial blood glucose peak P1 of cluster 1, the postprandial blood glucose peak P3 of cluster 3 satisfies P3 / P1>100%; or compared with the postprandial blood glucose peak P2 of cluster 2, the postprandial blood glucose peak P3 of cluster 3 satisfies P3 / P2>100%, such as>101%,>102%, preferably>103%.

[0042] In another preferred embodiment, "poor weight loss effect after weight loss intervention" means that, compared with the weight loss effect W1 of cluster 1, the weight loss effect W3 of cluster 3 satisfies W3 / W1 < 80%, such as < 70%, < 60%, < 50%, preferably < 45%; or compared with the weight loss effect W2 of cluster 2, the weight loss effect W3 of cluster 3 satisfies W3 / W2 < 90%, such as < 80%, < 70%, preferably < 65%.

[0043] In another preferred embodiment, the clinical biomarker further includes: (B) Any biomarker selected from B3-B11, or a combination thereof: (B3) Albumin AUC; (B4) Total cholesterol iAUC; (B5) Systolic blood pressure; (B6) Free triiodothyronine (FT3); (B7) Uric acid AUC; (B8) Fasting glucose; (B9) Age; (B10) Gamma-glutamyl transferase (GGT) AUC; (B11) Interferon-γ (IFN-γ) iAUC.

[0044] In another preferred embodiment, the clinical biomarker further includes: (C) Any marker selected from C1-C3, or a combination thereof: (C1) gender; (C2) BMI; (C3) intervention compliance.

[0045] In another preferred embodiment, the clinical biomarker further includes: (B) Any biomarker selected from B3-B11, or a combination thereof: (B3) Albumin AUC; (B4) Total cholesterol iAUC; (B5) Systolic blood pressure; (B6) Free triiodothyronine (FT3); (B7) Uric acid AUC; (B8) Fasting glucose; (B9) Age; (B10) Gamma-glutamyl transferase (GGT) AUC; (B11) Interferon-γ (IFN-γ) iAUC; (C) Any biomarker selected from C1-C3, or a combination thereof: (C1) sex; (C2) BMI; (C3) intervention compliance; (E) A combination of one or more markers from B3 to B11 and one or more markers from C1 to C3.

[0046] In another preferred embodiment, the diagnostic reagent or diagnostic kit method can also be used to evaluate the improvement of metabolic risk after weight loss intervention.

[0047] In another preferred embodiment, the indicators for evaluating the improvement of metabolic risk after the weight loss intervention include the difference in fasting blood glucose before and after the intervention, the difference in insulin resistance assessed by the homeostasis model (HOMA-IR) before and after the intervention, and the difference in fasting insulin before and after the intervention.

[0048] In another preferred embodiment, the subjects belonging to cluster 1 were evaluated as having a good effect on improving metabolic risk after weight loss intervention.

[0049] In another preferred embodiment, the subjects belonging to cluster 2 or cluster 3 were evaluated as having poor metabolic risk improvement after weight loss intervention.

[0050] In another preferred embodiment, "good improvement in metabolic risk after weight loss intervention" means that, compared with the HOMA-IR difference L2 before and after intervention in cluster 2, the HOMA-IR difference L1 before and after intervention in cluster 1 satisfies L1 / L2 > 120%, such as > 140%, > 160%, > 180%, preferably > 200%; or compared with the HOMA-IR difference L3 before and after intervention in cluster 3, L1 / L3 satisfies L1 / L3 > 100%, such as > 105%, > 110%.

[0051] In another preferred embodiment, "good improvement in metabolic risk after weight loss intervention" means that, compared with the difference in fasting insulin D2 before and after intervention in cluster 2, the difference in fasting insulin D1 before and after intervention in cluster 1 satisfies D1 / D2 > 105%, such as > 110%, > 115%, preferably > 120%; or compared with the difference in fasting insulin D3 before and after intervention in cluster 3, it satisfies D1 / D3 > 100%, such as > 105%.

[0052] In another preferred embodiment, the “poor improvement in metabolic risk after weight loss intervention” means that, compared with the HOMA-IR difference L1 before and after intervention in cluster 1, the HOMA-IR difference L2 before and after intervention in cluster 2 satisfies L2 / L1 < 90%, such as < 80%, < 70%, and preferably < 60%.

[0053] In another preferred embodiment, the “poor effect of metabolic risk improvement after weight loss intervention” means that, compared with the difference in fasting insulin D1 before and after intervention in cluster 1, the difference in fasting insulin D2 before and after intervention in cluster 2 satisfies D2 / D1 < 95%, such as < 90%, < 85%, preferably < 80%.

[0054] In another preferred embodiment, the “poor improvement in metabolic risk after weight loss intervention” means that, compared with the HOMA-IR difference L1 before and after intervention in cluster 1, the HOMA-IR difference L3 before and after intervention in cluster 3 satisfies L3 / L1 < 95%, preferably < 90%.

[0055] In another preferred embodiment, the “poor improvement in metabolic risk after weight loss intervention” means that, compared with the difference in fasting insulin D1 before and after intervention in cluster 1, the difference in fasting insulin D3 before and after intervention in cluster 3 satisfies D3 / D1 < 95%, preferably < 90%.

[0056] In another preferred embodiment, the diagnostic reagent or diagnostic kit can also be used for lifestyle intervention guidance.

[0057] In another preferred embodiment, lifestyle intervention guidance is provided based on the results of the typing and / or the results of the prediction.

[0058] In another preferred embodiment, if the classification result of the subject to be tested is respiratory quotient cluster type 2, then exercise intervention guidance is provided to the subject to be tested.

[0059] In another preferred embodiment, if the classification result of the subject to be tested is respiratory quotient cluster type 3, then dietary intervention guidance is provided to the subject to be tested.

[0060] In another preferred embodiment, the dietary intervention guidance includes recommendations to consume high-quality diets (such as the Mediterranean diet or the traditional Jiangnan diet).

[0061] A second aspect of the present invention provides a method for constructing a model for metabolic response typing and / or predicting weight change after a weight loss intervention, the method comprising the steps of: (Z1) Energy metabolism data, anthropometric data, continuous glucose monitoring (CGM) data, and multi-omics data were collected from subjects receiving a high-challenge diet; the energy metabolism data included raw time-series data of respiratory quotient (RQ) and various energy metabolism indicators; (Z2) Obtaining relevant biomarkers specifically includes the following steps: processing the energy metabolism data to obtain multiple representative respiratory quotient dynamic trajectory phenotypes; dividing the subjects into various representative respiratory quotient dynamic trajectories, detecting the dynamic trajectories of multiple energy metabolism indicators of the subjects in different clusters, and selecting energy metabolism indicators with different dynamic trajectories between different clusters as the relevant biomarkers. (Z3) Obtaining key biomarkers, specifically including the following steps: performing regression analysis on the anthropometric data, CGM data, and multi-omics data with the representative respiratory quotient dynamic trajectory phenotype to obtain the key biomarkers of the representative respiratory quotient dynamic trajectory phenotype; (Z4) The relevant biomarkers and key biomarkers are combined with the baseline model to construct a model for metabolic response typing and / or prediction of weight change after weight loss intervention.

[0062] In another preferred embodiment, the high-challenge diet is: a high-fat diet, a high-carbohydrate diet, a high-protein diet, or a combination thereof.

[0063] In another preferred embodiment, the high-challenge diet is a high-fat diet.

[0064] In another preferred embodiment, the subjects to be tested are people of normal weight, overweight / obese people, and people with prediabetes.

[0065] In another preferred embodiment, the anthropometric data, CGM data, and multi-omics data include: demographic characteristics (e.g., sex, age), lifestyle indicators (e.g., MET, energy intake, macronutrient ratio), anthropometric indicators (e.g., BMI, waist circumference, hip circumference, waist-to-hip ratio, FM, FFM), cardiovascular function indicators (e.g., systolic blood pressure, diastolic blood pressure, heart rate), and blood glucose control and CGM indicators (e.g., blood glucose, HbA1c, HOMA-IR, insulin, C-peptide, glucagon, CGM mean blood glucose, variable blood glucose, etc.). The parameters include: differential coefficient, time to reach target and time to hypoglycemia, lipid indicators (such as TCH, TG, LDL-C, HDL-C, FFA), organ function indicators (ALT, AST, GGT, ALP, Cr, estimated glomerular filtration rate, ALB, UA), inflammatory markers (such as CRP, IFN-γ, various interleukins), thyroid hormones (TSH, FT3, FT4) and metabolic hormones (Amylin, Ghrelin, GIP, GLP-1, Leptin, Adiponectin).

[0066] In another preferred embodiment, the anthropometry data, CGM data, and multi-omics data are preprocessed anthropometry data, CGM data, and multi-omics data.

[0067] In another preferred embodiment, the preprocessing includes: (s0.1) Standardize the continuous variable (mean 0, standard deviation 1); and (s0.2) Evaluate the correlation between variables and remove highly correlated variable pairs (e.g., |r|>0.80).

[0068] In another preferred embodiment, step (Z2) includes the following steps: (Z2.1) The original time series data of the respiratory quotient of each subject were transformed by wavelet transform and smoothing method to obtain the smoothed time series data of the respiratory quotient of each subject. (Z2.2) Cluster the smoothed time series data of each subject's respiratory quotient to obtain a representative dynamic trajectory phenotype of the respiratory quotient; (Z2.3) The subjects are divided into representative respiratory quotient dynamic trajectories, and the dynamic trajectories of various energy metabolism indicators of the subjects in different clusters are detected. The energy metabolism indicators with different dynamic trajectories between different clusters are selected as the clinical biomarkers.

[0069] In another preferred embodiment, in step (Z2.1), the wavelet transform is a discrete wavelet transform.

[0070] In another preferred embodiment, in step (Z2.2), the smoothing method is Locally Weighted Regression Scatter Smoothing (LOESS).

[0071] In another preferred embodiment, the span parameter is set to 0.7 ± 0.1 in LOESS.

[0072] In another preferred embodiment, step (Z2.1) specifically includes the following steps: (z2.1.1) The raw time-series data of the respiratory quotient of each subject is decomposed using Haar wavelet basis functions, and a three-level decomposition scale is set; high-frequency fluctuations are identified by a threshold based on median absolute deviation (MAD); thus obtaining the transformed respiratory quotient time-series data; and (z2.1.2) Use LOESS to smooth the transformed respiratory quotient time series data of each subject, and set the span parameter to 0.7; thus obtaining the smoothed respiratory quotient time series data of each subject.

[0073] In another preferred embodiment, the relevant markers include: Selected from any biomarker from A1-A8, or a combination thereof: (A1) respiratory quotient; (A2) blood glucose level; (A3) carbohydrate oxidation rate (CarbOx); (A4) fat oxidation rate (FatOx); (A6) energy expenditure rate (EE); (A7) free fatty acids (FFA); (A8) insulin / glucagon.

[0074] In another preferred embodiment, step (Z2.2) specifically includes the following steps: (Z2.2.1) The smoothed time series data of the respiratory quotient of each subject is modeled using B-spline basis functions to obtain a continuous smooth function of the respiratory quotient trajectory of each subject. (Z2.2.2) Based on the continuous smooth function, the similarity of the respiratory quotient trajectory of each test object is calculated using the dynamic time warping algorithm to obtain the distance matrix; (Z2.2.3) An unsupervised algorithm is used to perform cluster analysis on the distance matrix to obtain a representative dynamic trajectory phenotype of the respiratory quotient.

[0075] In another preferred embodiment, the dynamic time warping algorithm includes: derivative-based dynamic time warping (DDTW) and weighted dynamic time warping (WDTW).

[0076] In another preferred embodiment, in step (Z3), the regression learning includes linear regression and logistic regression.

[0077] In another preferred embodiment, the linear regression is a hierarchical multiple regression.

[0078] In another preferred embodiment, the logistic regression is elastic network logistic regression.

[0079] In another preferred embodiment, the unsupervised algorithm includes an algorithm based on the shape centroid.

[0080] In another preferred embodiment, the key markers include: The markers indicated by B1 and / or B2 are: (B1) AUC of total free fatty acids (5-h FFA) 5 hours after a meal; (B2) fasting gamma-glutamyl transferase (GGT).

[0081] In another preferred embodiment, the key marker further includes: (B) Any biomarker selected from B3-B11, or a combination thereof: (B3) Albumin AUC; (B4) Total cholesterol iAUC; (B5) Systolic blood pressure; (B6) Free triiodothyronine (FT3); (B7) Uric acid AUC; (B8) Fasting glucose; (B9) Age; (B10) Gamma-glutamyl transferase (GGT) AUC; (B11) Interferon-γ (IFN-γ) iAUC.

[0082] In another preferred embodiment, the baseline model includes gender and age.

[0083] In another preferred embodiment, the metabolic response includes respiratory quotient type, postprandial substrate conversion capacity, and glycemic regulation capacity.

[0084] In another preferred embodiment, the metabolic response typing includes typing the respiratory quotient type.

[0085] In another preferred embodiment, the term "classification" refers to classifying the test subject into respiratory quotient cluster type 1, respiratory quotient cluster type 2, and respiratory quotient cluster type 3.

[0086] The third invention provides a system for typing the metabolic response of a test subject and / or predicting weight changes after a weight loss intervention, the system comprising: An input unit is configured to input data, the input data including clinical biomarker data of the subject; the clinical biomarker data includes values ​​or change curves of the clinical biomarkers; the clinical biomarker data includes: (A) Any biomarker selected from A1-A8, or a combination thereof: (A1) Respiratory quotient; (A2) Blood glucose level; (A3) Carbohydrate oxidation rate (CarbOx); (A4) Fat oxidation rate (FatOx); (A6) Energy expenditure rate (EE); (A7) Free fatty acids (FFA); (A8) Insulin / glucagon; (B) Markers selected from B1 and / or B2: (B1) Postprandial total free fatty acids (FFA) AUC; (B2) Fasting gamma-glutamyl transferase (GGT); (D) A combination of one or more markers from A1 to A8 and one or more markers from B1 to B2; The typing and prediction unit is configured as a model that types the metabolic response of the subject to be tested and / or predicts weight changes after weight loss intervention based on the clinical biomarker data, thereby obtaining results; the model is constructed using the method described in the second aspect of the present invention. An output unit is configured to output the results of the fractal and prediction unit.

[0087] In another preferred embodiment, the total free fatty acid AUC after a meal includes the total free fatty acid AUC ≥ 2 hours after a meal, such as the total free fatty acid (5-h FAA) AUC 5 hours after a meal.

[0088] In another preferred embodiment, the clinical biomarker further includes: (B) Any biomarker selected from B3-B11, or a combination thereof: (B3) Albumin AUC; (B4) Total cholesterol iAUC; (B5) Systolic blood pressure; (B6) Free triiodothyronine (FT3); (B7) Uric acid AUC; (B8) Fasting glucose; (B9) Age; (B10) Gamma-glutamyl transferase (GGT) AUC; (B11) Interferon-γ (IFN-γ) iAUC; (C) Any biomarker selected from C1-C3, or a combination thereof: (C1) sex; (C2) BMI; (C3) intervention compliance; (E) A combination of one or more markers from B3 to B11 and one or more markers from C1 to C3.

[0089] In another preferred embodiment, the system further includes a diagnostic unit configured to provide diagnostic results for the object to be tested based on the results of the typing and prediction unit.

[0090] In another preferred embodiment, the output unit is further configured to output the diagnostic results of the diagnostic unit.

[0091] A fourth aspect of the present invention provides a method for screening key biomarkers, the method comprising the steps of: (C1) Collect energy metabolism data, anthropometric data, continuous glucose monitoring (CGM) data, and multi-omics data from subjects receiving a high-challenge diet; the energy metabolism data includes raw time-series data of respiratory quotient (RQ) and various energy metabolism indicators; (C2) The original time series data of the respiratory quotient of each subject were transformed by wavelet transform and smoothing method to obtain the smoothed time series data of the respiratory quotient of each subject; the smoothed time series data of the respiratory quotient of each subject were clustered to obtain a representative dynamic trajectory phenotype of the respiratory quotient. (C3) The anthropometric data, CGM data and multi-omics data are subjected to regression analysis with the representative respiratory quotient dynamic trajectory phenotype to obtain the key biomarkers of the representative respiratory quotient dynamic trajectory phenotype.

[0092] In another preferred embodiment, the high-challenge diet is: a high-fat diet, a high-carbohydrate diet, a high-protein diet, or a combination thereof.

[0093] In another preferred embodiment, the high-challenge diet is a high-fat diet.

[0094] In another preferred embodiment, the subjects to be tested are people of normal weight, overweight / obese people, and people with prediabetes.

[0095] In another preferred embodiment, the anthropometric data, CGM data, and multi-omics data include: demographic characteristics (e.g., sex, age), lifestyle indicators (e.g., MET, energy intake, macronutrient ratio), anthropometric indicators (e.g., BMI, waist circumference, hip circumference, waist-to-hip ratio, FM, FFM), cardiovascular function indicators (e.g., systolic blood pressure, diastolic blood pressure, heart rate), and blood glucose control and CGM indicators (e.g., blood glucose, HbA1c, HOMA-IR, insulin, C-peptide, glucagon, CGM mean blood glucose, variable blood glucose, etc.). The parameters include: differential coefficient, time to reach target and time to hypoglycemia, lipid indicators (such as TCH, TG, LDL-C, HDL-C, FFA), organ function indicators (ALT, AST, GGT, ALP, Cr, estimated glomerular filtration rate, ALB, UA), inflammatory markers (such as CRP, IFN-γ, various interleukins), thyroid hormones (TSH, FT3, FT4) and metabolic hormones (Amylin, Ghrelin, GIP, GLP-1, Leptin, Adiponectin).

[0096] In another preferred embodiment, the anthropometry data, CGM data, and multi-omics data are preprocessed anthropometry data, CGM data, and multi-omics data.

[0097] In another preferred embodiment, the preprocessing includes: (s0.1) Standardize the continuous variable (mean 0, standard deviation 1); and (s0.2) Evaluate the correlation between variables and remove highly correlated variable pairs (e.g., |r|>0.80).

[0098] In another preferred embodiment, in step (C2), the wavelet transform is a discrete wavelet transform.

[0099] In another preferred embodiment, in step (C2), the smoothing method is Locally Weighted Regression Scatter Smoothing (LOESS).

[0100] In another preferred embodiment, the span parameter is set to 0.7 ± 0.1 in LOESS.

[0101] In another preferred embodiment, step (C2) specifically includes the following steps: (c2.1) The original time series data of the respiratory quotient of each subject is decomposed using the Haar wavelet basis function, and a three-level decomposition scale is set; high-frequency fluctuations are identified by a threshold based on median absolute deviation (MAD); thus obtaining the transformed respiratory quotient time series data. (c2.2) Use LOESS to smooth the transformed respiratory quotient time series data of each subject, and set the span parameter to 0.7; thereby obtaining the smoothed respiratory quotient time series data of each subject. (c2.3) Use B-spline basis functions to model the smoothed time series data of the respiratory quotient of each subject to obtain a continuous smooth function of the respiratory quotient trajectory of each subject to obtain the continuous smooth function. (c2.4) Based on the continuous smooth function, the similarity of the respiratory quotient trajectory of each test object is calculated using the dynamic time warping algorithm to obtain the distance matrix; (c2.5) An unsupervised algorithm is used to perform cluster analysis on the distance matrix to obtain a representative dynamic trajectory phenotype of the respiratory quotient.

[0102] In another preferred embodiment, in step (C3), the regression learning includes linear regression and logistic regression.

[0103] In another preferred embodiment, the linear regression is a hierarchical multiple regression.

[0104] In another preferred embodiment, the logistic regression is elastic network logistic regression.

[0105] In another preferred embodiment, the unsupervised algorithm includes an algorithm based on the shape centroid.

[0106] In another preferred embodiment, the key markers include: The markers indicated by B1 and / or B2 are: (B1) AUC of total free fatty acids (5-h FFA) 5 hours after a meal; (B2) fasting gamma-glutamyl transferase (GGT).

[0107] In another preferred embodiment, the key marker further includes: (B) Any biomarker selected from B3-B11, or a combination thereof: (B3) Albumin AUC; (B4) Total cholesterol iAUC; (B5) Systolic blood pressure; (B6) Free triiodothyronine (FT3); (B7) Uric acid AUC; (B8) Fasting glucose; (B9) Age; (B10) Gamma-glutamyl transferase (GGT) AUC; (B11) Interferon (IFN)-γ iAUC.

[0108] It should be understood that, within the scope of this invention, the above-described technical features of this invention and the technical features specifically described below (such as in the embodiments) can be combined with each other to form new or preferred technical solutions. Due to space limitations, they will not be described in detail here. Attached Figure Description

[0109] Figure 1 The overall scheme of this study is shown.

[0110] Figure 2 This table shows characteristic comparisons between study groups (n=112) before and after a 12-week lifestyle intervention in the EMPOWER study, under fasting conditions. Continuous variables are expressed as mean (standard deviation) or median [1st quartile, 3rd quartile], and categorical variables are expressed as n (%). Differences between groups were compared using Student's t-test (continuous variables) or Wilcoxon's test (categorical variables), supplemented by Pearson's χ² test. P < 0.05. For weight-corrected energy metabolism biomarkers, inter-group differences were compared using analysis of covariance (ANCOVA) after adjusting for sex, age, weight, and seasonality. The estimated mean and standard error are reported in the table. For analyses of variable change, additional adjustment to baseline levels using ANCOVA is required. Demographics and Anthropometrics; Resting Energy Metabolism Markers; Glycemic Traits; Lipid Profiles; Organ Functions; Inflammatory Markers; Hormones and Adipokines; ALB; ALP; ALT; AST; BMI (Body Mass Index, calculated as weight (kg) / (height [m])²); CarbOx; CGM CV; CGM LBGI (Continuous Glycemic Index for Low Glycemic Range, applicable only when blood glucose <3.9 mmol / L); CGM TBR (Continuous Glycemic Time in Low Range); CGM TIR (Continuous Glycemic Time in Normal Range); CR (Creative Creatinine). DBP, diastolic blood pressure; EE, energy expenditure; eGFR, estimated glomerular filtration rate, calculated using the simplified CKD-EPI formula: 141 × min (creatinine / κ, 1). α × max(creatinine / κ, 1) -1.209 × 0.993 年龄× 1.018 (female) × 1.159 (Black), where κ = 0.7 (female) / 0.9 (male), α = -0.329 (female) / -0.411 (male); FFA, free fatty acids; FatOx, fat oxidation rate; FM, fat mass; FFM, fat-free mass; FT3, free triiodothyronine; FT4, free thyroxine; GGT, gamma-glutamyl transferase; GLP-1, glucagon-like peptide-1; GIP, glucose-dependent insulinotropic peptide; HbA1c, glycated hemoglobin A1c; HDL-C, high-density lipoprotein cholesterol; HOMA-β, β-cell function assessed by the homeostasis model, calculated as: (20 × insulin concentration (μIU / mL)) / (blood glucose concentration (mmol / L) - 3.5); HOMA-IR, homeostasis model assessment of insulin resistance, calculated as (insulin (μIU / mL) × glucose (mmol / L)) / 22.5; HS-CRP, high-sensitivity C-reactive protein; IFN-γ, interferon-γ; IL-1β, interleukin-1β; IL-18, interleukin-18; IL-6, interleukin-6; IL-8, interleukin-8; LM, lean body mass; LDL-C, low-density lipoprotein cholesterol; MET, metabolic equivalent; RQ, respiratory quotient; SBP, systolic blood pressure; TC, total cholesterol; TG, triglycerides; TSH, thyroid-stimulating hormone; UA, uric acid; VATM, visceral adipose tissue mass.

[0111] Figure 3 The changes in energy metabolism biomarkers are shown. (a) shows the time-coordinated dynamic changes in the energy, glucose, and lipid metabolism systems during the postprandial phase of the MMTT; (b) shows the trajectory of fasting energy metabolism biomarkers in individuals who achieved clinically meaningful weight loss goals (≥10% of baseline body weight).

[0112] Figure 4This study presents postprandial substrate utilization patterns and characteristics of blood glucose, lipolysis kinetics, weight changes, and the degree of improvement in metabolic risk for three metabolic clusters based on the response rate (RQ). (a) Changes in RQ values, carbohydrate oxidation rate (CarbOx), fat oxidation rate (FatOx), glucose levels, and free fatty acid levels before and after the MMTT in the three clusters. (b) Changes in RQ values, carbohydrate oxidation rate (CarbOx), fat oxidation rate (FatOx), glucose levels, and free fatty acid levels before and after eating in the three clusters after a 12-week weight loss intervention. (c) Comparison of differences in weight loss among overweight and obese individuals receiving a restricted-energy diet lifestyle intervention in different clusters. (d) Comparison of differences in fasting glucose, HOMA-IR, and fasting insulin levels among overweight and obese individuals receiving a restricted-energy diet lifestyle intervention in different clusters.

[0113] Figure 5 The dynamic trajectories of three RQ clustering patterns are shown under a high-carbohydrate diet (carbohydrate:fat:protein = 55:30:15).

[0114] Figure 6 The results show the performance evaluation of the RQ-based clustering method in predicting weight loss achieved through lifestyle interventions.

[0115] Figure 7a The key biomarkers distinguishing Cluster 1, Cluster 2, and Cluster 3 are shown. The top two figures use hierarchical multivariate regression to screen biomarkers that differentiate between clusters at the physiological function level; the bottom two figures reproduce the above results using machine learning methods.

[0116] Figure 7b The main driving factors in different clusters are shown.

[0117] Figure 7c The study demonstrates how multivariate logistic regression can be used to assess the discrimination and prediction of respiratory quotient clustering based on determinants such as age, sex, FFA AUC, and fasting GGT.

[0118] Figure 7d The results show that in external validation, the CGM and OGTT blood glucose curves of the populations clustered based on age, sex, FFA AUC, and fasting GGT all exhibited clustering trends similar to those of the respiratory quotient.

[0119] Figure 7e The study showed differences in weight loss and blood glucose levels among clusters of externally validated populations.

[0120] Figure 8 The baseline characteristics of the externally validated population clusters are shown.

[0121] Figure 9 This study demonstrated the effect of acute exercise on substrate selection for energy metabolism 3-5 hours postprandial in the MMTT among individuals with different physical activity levels in Cluster2.

[0122] Figure 10 This study showed changes in weight loss and glucose homeostasis indicators among clusters after 3 months of Mediterranean diet intervention.

[0123] Figure 11 This study showed changes in weight loss and glucose metabolism homeostasis indicators among clusters after 3 months of traditional Jiangnan dietary intervention.

[0124] Figure 12 The study showed the dynamic changes in CGM blood glucose among different dietary groups within clusters before and after intervention, as well as the dynamic changes in CGM blood glucose among different dietary groups within different clusters. Detailed Implementation

[0125] Through extensive and in-depth research, the inventors, based on changes in metabolic indicators before and after meals in normal-weight and overweight / obese individuals, discovered for the first time that three clusters based on the respiratory quotient (RQ) are correlated with weight loss and the degree of improvement in metabolic risk after weight loss intervention. Multiple biomarkers related to these three cluster characteristics also exhibit three types of dynamic change patterns. The inventors further discovered key biomarkers that distinguish these three clusters. Predictive models constructed using these key biomarkers can accurately classify subjects into the three clusters, thus predicting weight loss intervention outcomes and guiding personalized stratification strategies for weight loss intervention. Adding additional key biomarkers can construct even more accurate predictive models. Based on these findings, this invention was completed.

[0126] It should be understood that the specific methods and experimental conditions of the invention described below in varying degrees of detail are intended to provide a substantive understanding of the invention. Definitions of certain terms used in this specification are provided below. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0127] the term As used herein, the terms “containing” or “including (comprise)” can be open-ended, semi-closed, or closed-ended. In other words, the terms also include “consistently made of” or “made of”.

[0128] As used herein, the term “and / or” refers to and covers any and all possible combinations of one or more of the related listed items.

[0129] As used in this article, the term "significant" means that, in hypothesis testing, the observed effect (such as the difference between the experimental and control groups) is unlikely to be caused solely by random error. A hypothesis test includes: the null hypothesis (H0), which assumes that the observed effect does not exist (such as no difference between the experimental and control groups); the p-value, which is the probability of observing the current or more extreme effect when H0 is true; and the significance threshold (α). The significance threshold is typically used to determine whether a hypothesis test is significant. Generally, the significance threshold is 0.05. If the p-value ≤ α, then H0 is rejected, meaning the observed effect exists, and the result is called "significant." The term "marginal significance" refers to a situation where the p-value is close to, but does not reach, the significance level.

[0130] As used herein, the term "standardized mixed macronutrient tolerance test (MMTT)" is a test that monitors changes in blood glucose, insulin, C-peptide, and other indicators after consuming a standardized mixed meal (containing carbohydrates, protein, and fat). It serves as a standardized metabolic homeostasis test that simulates daily life and reflects an individual's metabolic flexibility. The term "acceptable meal in a standardized mixed macronutrient tolerance test" refers to a nutritionally balanced meal commonly used in MMTT trials. Those skilled in the art are familiar with how to formulate or select meals for MMTT trials.

[0131] As used in this paper, the term "EMPOWER (Energy metabolism profiles over weight-loss and eating responses)" is the name of this study.

[0132] As used in this article, the term "respiratory quotient" is the ratio of the body's carbon dioxide production (VCO2) to its oxygen consumption (VO2). It reflects the body's oxidation rate of carbohydrates and fats; the higher the RQ, the higher the body's carbohydrate oxidation capacity; conversely, the lower the RQ, the higher the fat oxidation capacity.

[0133] As used herein, the term "AUC" refers to the area under the curve. In this invention, the AUC of a certain indicator refers to the AUC of the indicator over a period of time, obtained by measuring the level of the indicator over that period and obtaining the curve of the level change of the indicator during that time, thereby allowing the calculation of the AUC of the indicator within that time period. Preferably, the AUC of each indicator in this invention is the AUC 5 hours after a meal. The term "iAUC" is used to assess the fluctuation of an indicator over a specific time period. Those skilled in the art are familiar with methods for calculating the AUC value or iAUC of a specific indicator.

[0134] The main advantages of this invention include: (1) This invention identifies three significantly different clusters based on the dynamic trajectory of respiratory quotients before and after MMTT in normal weight, overweight / obese subjects.

[0135] (2) The present invention further discovered that people belonging to different clusters have different dynamic change patterns in fasting fat oxidation capacity, postprandial metabolic response, and postprandial blood glucose response.

[0136] (3) This invention has discovered key biomarkers for distinguishing different clusters. The predictive model constructed using only two key biomarkers, FFA AUC and GGT, can accurately predict the type of blood glucose dynamic change that a subject belongs to, and this type of blood glucose dynamic change is highly consistent with the RQ cluster type, and can also predict weight changes after receiving meal intervention. By adding additional indicators, the sensitivity and accuracy of the classification can be further improved.

[0137] (4) The clinical biomarkers used in the method and system of the present invention are low-cost, simplified clinical prediction tools with the characteristics of convenient operation, universality and clinical application.

[0138] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Experimental methods in the following embodiments, unless otherwise specified, are generally performed under conventional conditions, such as those described in Sambrook et al., Molecular Cloning: A Laboratory Manual (New York: Cold Spring Harbor Laboratory Press, 1989), or as recommended by the manufacturer. Unless otherwise stated, percentages and parts are weight percentages and parts by weight.

[0139] Data sources and methods: 1. Study Design and Population: 1.1 Study Design: From March 2023 to January 2024, a 100-person precision intervention study on nutrition and lifestyle was conducted in Hangzhou, China. This study was an exploratory pre- and post-control intervention study. Volunteers were initially screened through an online recruitment questionnaire and then underwent physical examinations in person. Volunteers who met the inclusion criteria completed a questionnaire survey and physical examination. Information on demographics and physical activity was collected through the questionnaire. Anthropometric data such as height, weight, blood pressure, and heart rate were collected through the physical examination. Body composition and distribution data were collected using dual-energy X-ray absorptiometry (DXA). Metabolic homeostasis experiments were conducted in a metabolic chamber under resting and exercise conditions. Energy metabolism data, including EE, RQ, CarbOx, and FatOx, were collected through monitoring in the metabolic chamber. Clinical biochemical indicators were detected using blood samples collected at multiple time points (0, 0.5, 1, 2, 3, 4, and 5 hours) during the MMTT in the chamber. Subsequently, overweight and obese volunteers underwent a 12-week weight loss intervention based on a restricted-energy balanced diet and lifestyle. Volunteers of normal weight received guidance on healthy eating and lifestyle recommendations according to the *Chinese Dietary Guidelines*. After the intervention, volunteers underwent questionnaires, physical examinations, and metabolic homeostasis experiments. This project has been reviewed by the Ethics Committee of Sir Run Run Shaw Hospital, affiliated with Zhejiang University School of Medicine (Approval No.: Sir Run Run Shaw Hospital Ethics Review 2022 Research No. 0302) and registered on the ClinicalTrials.gov clinical trial registry website (Registration No.: NCT05785221). All volunteers signed informed consent forms before participating in the project.

[0140] 1.2 Study Population: A total of 112 volunteers were included in this project, including 84 with a BMI ≥ 24.0 kg / m². -2 The study included overweight and obese volunteers and 28 volunteers with a BMI of 18.5 ≤ BMI < 24.0 kg·m². -2 Normal body weight recombinant volunteers. The inclusion criteria for volunteers were: (1) age 20-70 years old; (2) BMI ≥ 18.5 kg·m -2 (3) Has not participated in other research projects in the present or in the three months prior to the study; (4) Is able to take care of themselves, has good cognitive and memory abilities, and is able to sign informed consent forms on their own. See Table 1 for detailed inclusion criteria.

[0141] Table 1. Intake and Exhaustion Criteria in the EMPOWER Study

[0142] 1.3 Lifestyle intervention: Overweight and obese volunteers will receive a 12-week weight loss intervention based on a restricted-energy balanced diet. At the start of the intervention, the nutritionist will determine the energy content based on the formula (kcal·d).-1 = (Height (cm) – 105) (kg) × Calorie requirement per unit body weight (kcal·kg) -1 ·d -1 The energy level range for weight loss in volunteers was calculated, where height – 105 was used to calculate ideal weight. For calorie requirements per unit body weight, since the study population consisted primarily of students, researchers, office workers, and other individuals engaged in light physical labor, a value of 20 or 25 kcal / kg was adopted. -1 ·d -1 Nutritionists provided guidance to volunteers by selecting recipes corresponding to lower energy levels within the energy range, and conducted standardized online monitoring and communication weekly (including randomized dietary records for two days, one on a weekday and one on a rest day). The weight loss intervention diet used in the overweight and obese group included six energy levels: 1000, 1200, 1400, 1600, 1800, and 2000 kcal. The macronutrient energy supply ratios were the same for all six energy levels: carbohydrates 40-50%, fat 30-35%, and protein 20-25%.

[0143] Volunteers in the control group will receive a 12-week routine nutritional lifestyle intervention. At the start of the intervention, a nutritionist will guide volunteers to develop healthier and more reasonable eating habits based on the latest version of the Chinese Dietary Guidelines and their three-day dietary records. Standardized online monitoring and communication will be conducted every four weeks (including two random days of dietary recording, one on a weekday and one on a rest day). During the lifestyle intervention, if volunteers had a regular exercise routine before participating in the program, they will continue to maintain their original exercise habits and intensity; if volunteers did not have a regular exercise routine before participating in the program, they will be advised to exercise 6,000 steps per day. All volunteers' exercise, heart rate, sleep, weight changes, and blood pressure during the intervention will be recorded using fitness trackers, smart scales, and blood pressure monitors, respectively.

[0144] 2. Laboratory testing and evaluation: 2.1 Questionnaires, CGM, and Physical Examination: Questionnaires and physical examinations were conducted 2–9 days prior to baseline and the first metabolic homeostasis experiment after intervention. Face-to-face interviews were conducted by trained researchers using a standard questionnaire (general survey questionnaire) to obtain information on demographic characteristics, lifestyle, physical activity, health status, and nutritional supplements. A three-day dietary assessment (including two workdays and one rest day) was conducted by a professional nutritionist to evaluate volunteers' dietary intake. Volunteers were required to wear a CGM (FreeStyle® Libre 2™; Abbott Diabetes Care, Alameda, CA) continuously for 14 days both at baseline and after intervention, with measurements covering the three days of baseline diet and the metabolic homeostasis experiment. Physical examination included measuring height and weight (Seca, 704s), waist and hip circumference (Seca, 201), and blood pressure and heart rate (OMRON, J750). Body fat mass (FM), fat-free mass (FFM), and distribution were measured using a DXA (GE, Lunar Prodigy Pro scanner), including upper limb mass, lower limb mass, trunk mass, upper limb FM, lower limb FM, trunk FM, upper limb FFM, lower limb FFM, and trunk FFM, as well as visceral adipose tissue mass (VATM). FFM = weight - FM. All data was entered using Epidata software (version 3.1), and double entry was used to ensure accuracy. Waist-to-hip ratio (WHR) is the ratio of waist circumference to hip circumference. Fat mass percentage (FM%) is the percentage of body weight that is fat mass (FM); fat-free mass percentage (FFM%) is the percentage of body weight that is fat-free (FFM). BMI is calculated by dividing weight by the square of height; Fat Mass Index (FMI) is calculated by dividing FM by the square of height; Fat-Free Mass Index (FFMI) is calculated by dividing FFM by the square of height. The units for BMI, FMI, and FFMI are all kg·m.-2 Metabolic equivalent task (MET) is calculated as activity factor × duration (H) × frequency (days), with units of H·W. -1 The activity coefficients for light, moderate, and heavy physical activity and walking were 2.5, 4.0, 8.0, and 3.3, respectively.

[0145] 2.2 Metabolic Homeostasis Experiment: Volunteers underwent one "resting day" metabolic homeostasis experiment and one "exercise day" metabolic homeostasis experiment in the metabolic chamber at baseline and after intervention, for a total of 4 experiments. In the "resting day" metabolic homeostasis experiment, volunteers entered the chamber after fasting for at least 10 hours overnight, lay supine on a recliner for 1 hour, and had fasting venous blood collected after 1 hour. They then ate a standard mixed meal (which had to be consumed within 5 minutes), and continued to lie supine on the recliner. Venous blood was collected at 0.5, 1, 2, 3, 4, and 5 hours after the start of the standard mixed meal (7 time points before and after the MMTT). Sleeping and significant movement were avoided throughout the experiment. In the "Exercise Day" metabolic homeostasis experiment, volunteers entered the chamber after fasting for at least 10 hours overnight, lay supine in a reclining chair for 2 hours, and had fasting venous blood collected 2 hours later. They then consumed a standard mixed meal (which had to be consumed within 5 minutes) and continued to lie supine in the reclining chair. Two hours after the meal, they engaged in 30 minutes of moderate-intensity cycling (emotion FITNESS, Torqualizer 600 med) (target heart rate value was 65% of the estimated maximum heart rate, i.e., 65% × (220 - age)). Venous blood was collected at 0.5, 1, 2, 3, 4, and 5 hours after the meal. Sleeping or significant movement was avoided throughout the experiment. Blood samples from all time points in the four metabolic homeostasis experiments were collected using coagulation tubes containing SST II separating gel. Each sample was centrifuged at 4°C and 2400 rcf for 15 minutes and then stored at -80°C for long-term analysis. Whole blood was collected in anticoagulant tubes containing dipotassium ethylenediaminetetraacetate (K2EDTA) at two additional fasting time points during the baseline and post-intervention "rest day" experiments for glycated hemoglobin (HbA1c) testing. The standard mixed meals consumed during the metabolic homeostasis experiment were prepared by By-Health Co., Ltd. based on the "standard mixed meal" from internationally published phenotypic plasticity studies, and were guaranteed to meet HACCP standards. The energy and macronutrient content of the formula strictly adhered to GB29922 requirements. Each standard mixed meal was 500ml in volume, containing 75g of carbohydrates, 60g of fat, and 20g of protein, with a total energy of approximately 3900 kJ (approximately 920kcal). The energy contribution ratio of the three nutrients was approximately 33%: 59%: 8%.

[0146] 2.3 Energy Metabolism Detection: Energy metabolism was measured using an indirect calorimetry metabolic chamber (Sable Systems International, LasVegas NV). The chamber temperature was maintained at 20–24°C. O2 concentration, CO2 concentration, water vapor pressure (WVP), and flow rate (FR) were measured using sensors, and VO2 and VCO2 per minute were calculated using the Melanson method. RQ is defined as the ratio of VCO2 to VO2 during respiration. When carbohydrates, especially glucose, are used as the energy substrate for complete oxidation and decomposition, 6 molecules of oxygen are consumed and 6 molecules of carbon dioxide are produced for every 1 molecule of glucose consumed, therefore RQ is 1. When fats are used as the energy substrate for complete oxidation and decomposition, more oxygen is required, resulting in an RQ of approximately 0.7. RQ varies slightly depending on the chain length and degree of saturation of fatty acids. EE was calculated based on the Weir formula, i.e., EE (kcal·min -1 = 3.941×VO2(L·min) -1 ) + 1.106×VCO2(L·min -1 CarbOx and FatOx are calculated based on the Kelly formula, i.e., CarbOx (g·min⁻¹) -1 = -3.226×VO2(L·min) -1 ) + 4.585×VCO2(L·min -1 ) -0.461×0.066. FatOx (g·min -1 = 1.695×VO2(L·min -1 ) - 1.701×VCO2(L·min -1 -0.319×0.066.

[0147] 2.4 Blood Biochemical Indicator Detection: Venous blood samples were collected using two types of blood collection tubes: tubes containing a coagulant were used for serum preparation, and tubes containing ethylenediaminetetraacetic acid (EDTA) were used for plasma preparation. Incretins and gastrointestinal peptides, including glucagon-like peptide-1 (GLP-1), were detected. 1 (Glucagon-Like Peptide-1, GLP 1) Glucagon, Active Ghrelin, and Amylin are administered vials pre-filled with peptide stabilizers, and the corresponding protease inhibitor, DPP, is added immediately after blood collection. IV inhibitors (GLP) 1) A mixture of apratinine (Glucagon), AEBSF (ActiveGhrelin), and a broad-spectrum protease inhibitor (Amylin). All samples were centrifuged at 4°C, aliquoted, and then... Store frozen at 80°C for further analysis.

[0148] HbA1c was detected using an immunoturbidimetric assay (Cobas C513 Roche). Insulin and C-peptide were detected using a Roche kit on a fully automated biochemical analyzer (Cobas E602 Roche). Free triiodothyronine (FT3), free thyroxine (FT4), and thyroid-stimulating hormone (TSH) were detected using Wako reagents on a fully automated biochemical analyzer (Cobas C702, Roche). Free fatty acids (FFA) were detected using a Meikang kit on a fully automated biochemical analyzer (AU680, Beckman). Routine biochemical indicators included glucose, total cholesterol (TCH), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C). LDL-C, alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), gamma-glutamyltransferase (GGT), creatinine (Cr), uric acid (UA), C-reactive protein (CRP), and albumin (ALB) were detected using a Roche kit on a fully automated biochemical analyzer (C702, Roche).Multiple cytokines and metabolic hormones, including amylin, ghrelin, glucose-dependent insulinotropic peptide (GIP), glucagon-like peptide-1 (GLP-1), glucagon, leptin, interferon-γ (IFN-γ), interleukin-1β (IL-1β), interleukin-6 (IL-6), interleukin-8 (IL-8), and interleukin-18 (IL-18), were measured on the Luminex platform using the Milliporesigma kit. To ensure the accuracy and consistency of the test results, all surplus samples collected during the project were combined into a single Quality Control (QC) sample before the testing began, and this sample was inserted into the sample testing sequence multiple times. The coefficient of variation for all QC samples was <10%.

[0149] 2.5 Compliance Assessment: Compliance was defined as the extent to which participants completed the pre-defined core tasks. Due to differences in task requirements between different study groups (overweight / obese group and normal weight group), compliance was calculated separately for each group. For the overweight / obese group, the core tasks included: (1) self-weight measurement; (2) completion of online micro-courses; (3) dietary record keeping; (4) maintaining interaction and communication with the dietitian; (5) physical activity monitoring; and (6) dietary quality assessment. The weekly compliance rate for each task was calculated as follows: the number of weeks in which the required frequency was achieved divided by the total number of weeks of intervention (e.g., weight measurement compliance = number of weeks in which the required measurement frequency was achieved / total number of effective weeks). The normal weight group did not need to undergo dietary quality assessment, and the remaining tasks were the same as those for the overweight / obese group. For each task for each participant, compliance was divided into two categories based on the cohort median: participants who reached or exceeded the median received 1 point, otherwise 0 points. The total compliance score is the unweighted sum of the scores for each task. The score is 0–6 for the overweight / obese group and 0–5 for the normal weight group. A higher score indicates better adherence to the study protocol.

[0150] 3. Statistical methods: 3.1 Data Smoothing: To reduce measurement noise and preserve true physiological dynamics, the original time-series data were processed using a two-stage smoothing method. In the first stage, Discrete Wavelet Transform (DWT) was used for signal decomposition, specifically employing the Haar wavelet basis function and setting a three-level decomposition scale. High-frequency fluctuations were identified using a threshold based on Median Absolute Deviation (MAD). In the second stage, Locally Estimated Scatterplot Smoothing (LOESS) was applied with a span parameter of 0.7 to generate a smoothed trajectory estimation curve for each subject. This method can adapt to local variations in the data without relying on a pre-defined global function form, making it particularly suitable for capturing the complex dynamic response patterns during postprandial metabolism.

[0151] 3.2 RQ Dynamic Trajectory Clustering: The smoothed RQ time series data were further modeled using B-spline basis functions (15 cubic splines covering the entire observation time window), representing individual metabolic trajectories as continuous smooth functions, thus more comprehensively characterizing their overall morphology and temporal variation features. Subsequently, the trajectory similarity between individuals was calculated using a distance metric based on derivative dynamic time warping (DTW). This method can correct for differences in reaction rates and phases between individuals through nonlinear matching of the time axis, thereby more accurately capturing differences in dynamic metabolic patterns. Based on the DTW distance matrix, an algorithm based on shape centroids was used to perform unsupervised clustering analysis on the subjects. Considering both the clinical interpretability of the clusters and the sample size balance for subsequent statistical analysis, three representative RQ dynamic trajectory phenotypes (k=3) were finally determined. The shape-based centroid of each cluster was used to characterize the typical RQ change pattern of that group of subjects, reflecting their energy metabolism response characteristics. To maintain consistency in the interpretation of results, the re-clustering results of the overweight and obese groups were re-labeled after the clustering process, referring to the phenotypic characteristics of the entire sample.

[0152] 3.3 RQ Clustering for Weight Loss Prediction: A nested logistic regression model was used to evaluate the incremental predictive value of baseline RQ clustering in predicting weight loss exceeding 10%. The base model included conventional clinical predictors (gender, age, baseline BMI, intervention adherence), while the full model further incorporated RQ clustering as an independent variable. Both models employed 5-fold cross-validation to obtain unbiased performance evaluation. The model's discriminative power was measured by the area under the receiver operating characteristic curve (AUC-ROC), and the difference in AUC between the two models was compared using the DeLong test. Furthermore, to further quantify the improvement in risk classification and probability discrimination after incorporating RQ clustering information, the Net Reclassification Improvement (NRI) and Integrated Discrimination Improvement (IDI) indices were calculated.

[0153] 3.4 Screening of key biomarkers for RQ clustering: The determinants of RQ clustering were assessed using two complementary strategies: (1) hierarchical multiple regression at both the physiological function and specific indicator levels; and (2) elastic-Net logistic regression to assess the joint signal of multiple indicators. Candidate variables were pre-divided into several domains based on physiological function before analysis, including: • Demographic characteristics (sex, age); • Lifestyle indicators (MET, energy intake, macronutrient ratio). • Anthropometry metrics (BMI, waist circumference, hip circumference, WHR, FM, FFM); • Cardiovascular function indicators (systolic blood pressure (SBP), diastolic blood pressure (DBP), heart rate); • Blood glucose control and CGM indicators (Glucose, HbA1c, HOMA-IR, Insulin, C-peptide, Glucagon; CGM mean blood glucose, coefficient of variation, time in range (TIR), and time above range (TBR)). • Blood lipid indicators (TCH, TG, LDL-C, HDL-C, FFA); • Organ function indicators (ALT, AST, GGT, ALP, Cr, estimated glomerular filtration rate (eGFR), ALB, UA); • Inflammatory markers (CRP, IFN-γ, various interleukins); • Thyroid hormones (TSH, FT3, FT4); • Metabolic hormones (Amylin, Ghrelin, GIP, GLP-1, Leptin, Adiponectin).

[0154] Unless otherwise specified, all biomarkers were collected in fasting and postprandial states. Postprandial assessment indicators were the total area under the curve (AUC) and incremental area under the curve (AUC) at 5 hours. Complete data were used for analysis, and continuous variables were standardized (mean 0, standard deviation 1). To reduce intradimensional collinearity, highly correlated variable pairs (|r|>0.80) were removed, prioritizing those with fewer missing values. For the physiological dimension model, AUC and McFadden's pseudo-R² were reported; for the univariate model, AUC (95% confidence interval) and pseudo-R² were reported. Variables with cross-validation AUC ≤ 0.50 were not discussed in detail. After identifying candidate variables through exploratory univariate screening, resilient network models were constructed for RQ clustering Cluster1 vs. Cluster3 (primary comparison) and Cluster1 vs. Cluster2 (secondary comparison), respectively. 10-fold cross-validation was used, and ROC curve optimization was employed. The gender variable was coded as male = 0 and female = 1, and no other covariates were included. The cross-validation AUC and its 95% confidence interval were calculated based on retention prediction. Finally, the non-zero elasticity network coefficients were summarized by refitting the model, and the contribution of each variable was presented as a percentage of relative absolute value (%|β| / Σ|β|).

[0155] 3.5 Validation of RQ Clustering: First, two binary logistic regression models were constructed, each containing a key predictor and age and gender variables. The model AUC was optimized through repeated 10-fold cross-validation (3 times). To further achieve ternary RQ clustering, a multinomial logistic regression model including all four predictor variables was established. 10-fold cross-validation was performed using Cohen's kappa coefficient as the optimization objective, and the one-to-many AUC was calculated based on the confusion matrix. To evaluate the effectiveness of the weight loss intervention, the trained model was applied to post-intervention data, and RQ clustering was performed again. The RQ clustering model was validated and its response to the weight loss intervention was evaluated based on population data from the Dietary Pattern and Metabolic Health (DPMH) project. Details of the experimental protocol can be found in published literature. Given that the DPMH population only had four oral glucose tolerance test (OGTT) blood collection time points (0, 30, 60, and 120 minutes), this study used the 2-hour FFA area under the curve instead of the 5-hour FFA area under the curve in the original model. To analyze the specific metabolic response of RQ clusters to weight loss intervention, a multivariable linear regression model was used to compare changes in body weight and glucose homeostasis parameters for each cluster during the 3-month intervention period, and adjustments were made for age, sex, baseline body weight, and baseline values ​​of the corresponding parameters. All analyses were performed using R 4.4.2, and the statistical significance was set at two-sided α = 0.05.

[0156] Example 1: Study Design Process and Baseline Characteristics of the Population The overall design scheme of the research is as follows: Figure 1 As shown. The overweight / obese and normal-weight groups were well-matched in age and sex distribution, but showed significant differences in anthropometric parameters: compared with the normal-weight group, the overweight / obese group had significantly higher BMI and waist circumference. In terms of energy metabolism, the overweight / obese group showed higher absolute resting energy expenditure (EE) and carbOx, but lower fatOx after adjusting for body weight. Figure 2 ).

[0157] Example 2: Temporal substrate switching characteristics of postprandial energy metabolism This embodiment involves exploring the temporal characteristics of postprandial energy metabolism biomarkers in a standard mixed macronutrient tolerance test.

[0158] In the postprandial phase of the MMTT, the energy, glucose, and lipid metabolism systems all exhibited time-coordinated dynamic changes. Figure 3a) In the general population, EE exhibits a rapid thermal effect response, peaking around 60 minutes and gradually declining, but remaining above baseline throughout the 5-hour sampling period. Concurrently, RQ rises early, and the CarbOx peak (around 42 minutes) coincides with the CGM blood glucose peak (around 46 minutes), indicating a preferential carbohydrate utilization pattern in the early postprandial period. FatOx shows the opposite pattern, being suppressed during the peak of glucose utilization and then gradually recovering as blood glucose levels decrease. FFA shows an inhibitory decline postprandially, with its trough coinciding with the CarbOx peak, and rebounding as FatOx recovers.

[0159] The temporal coupling trajectories of EE, RQ, substrate oxidation, glucose, and FFA reveal a complete metabolic response system: the body maintains energy homeostasis after nutrient load by dynamically adjusting substrate preferences.

[0160] Example 3: The effects of weight loss lifestyle interventions on energy metabolism This embodiment involves investigating the changes in individual energy metabolism indicators and biomarkers after a weight loss lifestyle intervention.

[0161] During the 12-week intervention period, physical examination parameters of overweight and obese volunteers receiving energy-restricted lifestyle interventions changed significantly, while those of normal-weight controls receiving weight-maintaining lifestyle guidance remained stable. Figure 2 The overweight and obese volunteers achieved an average weight loss of 6% of their baseline weight, with significant reductions in waist circumference and free fat mass (FFM). Post-intervention resting energy metabolism assessment showed that, compared to the normal-weight control group, the overweight and obese volunteers had increased fasting FatOx and decreased fasting CarbOx, indicating improved fat utilization under fasting conditions. No significant changes were observed in resting energy expenditure (EE), suggesting no adaptive downregulation of basal metabolic rate during weight loss. Blood glucose and endocrine indicators showed a simultaneous improvement trend: fasting insulin and HOMA-IR levels significantly decreased in the overweight and obese volunteers, while metabolic indicators remained relatively stable in the normal-weight control group. Figure 2 ).

[0162] Despite consistent positive changes at the group level, significant heterogeneity in metabolic responses remained among individuals. Analysis of individuals who achieved clinically meaningful weight loss (≥10% of baseline body weight) revealed highly differentiated trajectories in fasting energy metabolism markers. Figure 3 b): Cases ranging from significant improvements in lipid oxidation capacity to almost no change were observed. Individual-level analysis further showed that even individuals with limited weight loss (5–10% of baseline weight) could still experience improvements in metabolic adaptation, suggesting that optimization of energy metabolism may be achieved independently of the magnitude of weight loss. This heterogeneity fully demonstrates the significant differences in individual metabolic responses to energy-restricted lifestyle interventions.

[0163] Example 4: RQ-based metabolic clustering reveals different postprandial substrate utilization patterns To further explore the heterogeneity of energy metabolism response in non-fasting states, individual differences in postprandial metabolic responses were analyzed.

[0164] Although all volunteers received the exact same MMTT, functional data analysis combined with derivative dynamic time warping algorithm identified three significantly different clusters in the baseline MMTT based on the RQ dynamic trajectory. Figure 4 a). These three clusters (hereinafter referred to as Cluster 1, Cluster 2, and Cluster 3, respectively) exhibited unique substrate utilization characteristics: Cluster 1 (45 people) showed lower fasting RQ values, with a rapid and significant increase after meals followed by a gradual decline, suggesting that this group had stronger fasting lipid oxidation capacity and more efficient postprandial substrate conversion capacity; Cluster 2 (52 people) showed a delayed RQ peak that remained at a high level and failed to recover to baseline within the sampling time, indicating that their postprandial metabolic conversion process was prolonged; Cluster 3 (15 people) had higher fasting RQ values ​​and a relatively flat postprandial trajectory, suggesting impaired fasting lipid oxidation capacity and decreased postprandial metabolic response capacity. Figure 4 a).

[0165] After adopting a high-carbohydrate diet (carbohydrate:fat:protein = 55:30:15), RQ dynamic trajectory clustering also showed three significantly different clustering patterns. Figure 5 The dynamic trajectories of the three RQ clustering types are the same as the clustering results after receiving MMTT, indicating that the RQ-based clustering results do not change with dietary changes.

[0166] Example 5: RQ-based metabolic clusters exhibit consistent glycemic and lipolysis kinetics. The three clusters exhibited different postprandial blood glucose response patterns consistent with their RQ changes. Figure 4 a). All clusters showed a rapid rise in blood glucose after the MMTT, peaking within 30-60 minutes, but the magnitude of the peak and the recovery varied significantly. Cluster 1 exhibited the best blood glucose regulation: a peak blood glucose level of approximately 7.0 mmol / L (peak time ≈ 30 minutes), and the fastest recovery, reaching near fasting levels (approximately 4.8 mmol / L) by 180 minutes; Cluster 3 showed the highest and most delayed peak blood glucose level (approximately 7.1 mmol / L, peak time 60 minutes); Cluster 2, while showing a moderate peak (approximately 6.8 mmol / L, peak time 60 minutes), had the slowest overall recovery, maintaining the highest blood glucose level in the late postprandial period. Figure 4a). FFA levels were similar across groups in the fasting state, and all groups showed a decrease after meals, reaching their trough at 120–180 minutes, with Cluster 2 exhibiting the highest degree of inhibition. Figure 4 a). During the recovery phase, Cluster 1 rebounded earliest and with the largest magnitude, while Cluster 3 showed a milder and delayed rebound. From 180 minutes to 300 minutes, the FFA levels of both Cluster 1 and Cluster 3 surpassed those of Cluster 2, indicating that Cluster 1 had the fastest lipolytic recovery, while Cluster 2 had the most delayed recovery. The insulin / glucagon ratio showed significant cluster-specific differences in both amplitude and kinetics (a). Figure 4 a) Starting from similar fasting baselines, all clusters showed a rapid increase in postprandial ratios, but with distinct patterns: Cluster 3 reached its highest peak and maintained a high value until 180 minutes, indicating the longest duration of insulin dominance, consistent with its sustained delayed metabolic transition and FFA mobilization; Cluster 1 reached a moderate peak but declined the fastest, reaching its lowest ratio at 180 minutes, reflecting an earlier recovery from catabolic state, consistent with its excellent metabolic flexibility; Cluster 2 exhibited a decaying, plateau-like response pattern with a moderate rate of decline. These differentiated insulin / glucagon kinetics provide a hormonal explanation for the observed differences in cluster-specific substrate allocation and metabolic recovery.

[0167] Example 6: Acute exercise enhances postprandial energy metabolism To assess the short-term effects of exercise on postprandial metabolism, this example compares the differences in the "recovery period" 3-5 hours after a meal in the MMTT between resting and exercise days, and analyzes the area under the curves of EE, FatOx, and the contribution rate of fat oxidation (FatOx / EE).

[0168] Compared to the resting state, both EE and FatOx were significantly increased on exercise days (P<0.001), and the values ​​for Cluster 2 were consistently lower than those for Cluster 1 across different date comparisons. Notably, only Cluster 2 exhibited a change in substrate allocation pattern: its FatOx / EE significantly increased from 0.41 on resting days to 0.49 on exercise days (P<0.001), while Cluster 1 and Cluster 3 remained stable.

[0169] These data suggest that acute exercise can significantly enhance postprandial energy metabolism and selectively increase the lipid oxidation capacity of Cluster 2.

[0170] Example 7: RQ clustering reveals differential weight loss outcomes In overweight and obese individuals receiving a restricted-energy lifestyle intervention, RQ clustering was strongly associated with differences in intervention outcomes. Multiple linear modeling results, adjusted for age, sex, baseline BMI, intervention adherence, season, and energy intake during the intervention, showed that the percentage weight loss in Cluster 1 was significantly higher than in Clusters 2 and 3 (P=0.049 and P=0.012, respectively). This differential response remained consistent across individuals with different baseline weights. Figure 4 c). The results indicate that baseline RQ dynamics can serve as a potential stratification marker for weight loss responsiveness to lifestyle interventions.

[0171] Besides changes in weight, different RQ clusters also showed differences in the degree of improvement in metabolic risk: regarding blood glucose improvement, there were overall marginally significant differences in fasting blood glucose among clusters, while HOMA-IR showed significant differences. Cluster 1 showed greater improvement than Cluster 2, with Cluster 3 in between. Figure 4 d).

[0172] Example 8: Integrating RQ clustering features to improve weight loss prediction performance Integrating RQ clustering features significantly improved the predictive power for achieving clinically significant weight loss (≥10%) after lifestyle interventions. Figure 6 After adding clustering information to the baseline model that included age, gender, BMI, and intervention adherence, the model's discriminative power was significantly improved (AUC increased from 0.52 to 0.79, P=0.0004). Figure 6 This improvement in predictive accuracy was further validated by significant improvements in the net weight classification improvement index (NRI=0.34, 95%CI: 0.06-0.62; P=0.02) and the comprehensive discriminant improvement index (IDI=0.13, 95%CI: 0.03-0.23; P=0.008). Figure 6 ).

[0173] The results show that RQ-based clustering methods have the potential to become a personalized stratification strategy to guide weight loss interventions.

[0174] Example 9: Metabolic convergence and cluster redistribution after weight loss intervention After 12 weeks of intervention, the postprandial RQ trajectories of different clusters showed a convergence: the peak time was closer to the recovery process, and the difference in magnitude was significantly reduced. Figure 4(b) The RQ dynamic trajectories of Cluster 2 and Cluster 3 converged with those of Cluster 1, showing lower fasting RQ values ​​and more synchronized postprandial dynamics, suggesting improved substrate conversion ability in some participants. However, idiosyncratic differences remained: Cluster 1 maintained the fastest baseline recovery rate, Cluster 3 maintained the most sustained RQ elevation, while Cluster 2 remained at an intermediate level. Other indicators also showed similar characteristics of partial normalization with preservation of the original order: CarbOx peaks were more synchronized, lipid oxidation during recovery was enhanced (Cluster 2 showed the most significant improvement), blood glucose fluctuations were reduced, and FFA rebound occurred earlier.

[0175] These findings collectively demonstrate that interventions, while promoting metabolic convergence, still retain the specific characteristics of each cluster.

[0176] Example 10: Key clinical indicators reveal individual differences in the dynamic changes of postprandial RQ. The key factors determining RQ clustering can ultimately be attributed to a few core biomarkers. Preliminary screening of physiological function dimensions suggests that postprandial indicators are generally better than fasting indicators. Figure 7a While individual indicator analysis precisely pinpointed the specific biomarkers that played a dominant role, postprandial FFA response was the most significant differentiating indicator when distinguishing between the most metabolically flexible Cluster 1 and Cluster 2: their 5-hour total FFA AUC had the highest AUC and pseudo-R² values ​​among all indicators, exhibiting a clear intergroup separation—Cluster 1's values ​​were significantly higher than Cluster 2's (P<0.001). In contrast, the distinction between Cluster 1 and Cluster 3 was primarily driven by GGT: fasting GGT, reflecting early hepatocyte stress and γ-glutamyl cycle activity, was significantly elevated in Cluster 3 (P<0.001). Figure 7b ).

[0177] Another machine learning approach replicated the above findings using a resilient network logistic regression with 10-fold cross-validation: in the comparison between Cluster 1 and Cluster 2, both coefficient weights and retention ROC performance were concentrated on FFA AUC; while in the comparison between Cluster 1 and Cluster 3, the focus was on fasting GGT ( Figure 7a These results collectively reveal two biological axes of RQ trajectory heterogeneity: Cluster 2 showed impaired postprandial lipid metabolism after a high-fat diet, while Cluster 3 showed enhanced liver function signals indicated by fasting GGT, which is consistent with the characteristics of this population's slow metabolic response to dietary fat and long-term reduced lipid oxidation capacity.

[0178] Example 11: Validating RQ clustering in an independent population A binary logistic regression model adjusted for sex and age was established, and key determinants were paired with demographic indicators for analysis. The model including the 5-h FFA AUC was able to effectively distinguish between Cluster 1 and Cluster 2 (AUC = 0.72). Figure 7c The model including fasting GGT successfully distinguished between Cluster 1 and Cluster 3 (AUC = 0.76). Figure 7c ).

[0179] Subsequently, multinomial logistic regression was used to evaluate whether gender, age, and the two determinants could jointly predict the three RQ clusters. Ten-fold cross-validation determined a simplified and clinically feasible model with the following AUCs for cross-validation: Cluster 1: 0.70, Cluster 2: 0.65, and Cluster 3: 0.65. Figure 7c Based on the single-factor models of FFA and GGT, adding other key determinants also yielded good differentiation results. Figure 7c ).

[0180] The Dietary Pattern and Metabolic Health (DPMH) study is a randomized controlled meal intervention trial. It includes three intervention diets: the Mediterranean Diet (MD), the Traditional Jiangnan Diet (TJD), and the Control Diet (CD). RQ clustering was validated based on this study (n=200). Baseline characteristics are summarized below. Figure 8 The postprandial blood glucose dynamics clustering of this study population highly matched the three RQ clustering phenotypes: after consuming a standard breakfast with relatively high fat content (Table 2), the CGM blood glucose curves showed a trend of gradually increasing peak values ​​and gradually slowing recovery from Cluster 1 to Cluster 3, with Cluster 1 recovering to baseline levels the fastest. Figure 7d The baseline OGTT glycemic response reproduced the same ranking pattern. Figure 7d These patterns indicate that the substrate switching ability is gradually impaired from Cluster 1 to Cluster 3, which corroborates the results of RQ-based classification.

[0181] Table 2. Standard breakfast during the DPMH study intervention (n=67)

[0182] In the control diet group, after 3 months of dietary intervention (37% of energy from dietary fat), Cluster 1 showed better weight loss than Cluster 3 (8.7% vs 6.2%, P = 0.02), and visceral fat was significantly reduced. 44 cm² vs 12 cm², P = 0.005). Insulin sensitivity was significantly improved: the decrease in fasting insulin was greater ( 6.9 μIU / mL vs 4.1 μIU / mL, P = 0.03), and the decrease in HOMA-IR was also more significant. 2.2 vs 1.2, P = 0.009) Figure 7e However, there were no significant differences between the Mediterranean diet and the traditional Jiangnan diet in terms of clusters in terms of weight loss and glucose homeostasis. Figure 10 and Figure 11 ).

[0183] Example 12: Personalized Lifestyle Intervention Guided by RQ Clustering This study found that RQ clustering can guide personalized lifestyle interventions.

[0184] Analysis of the found population, Cluster 2, showed that exercise can significantly increase the proportion of fat oxidation for energy (FatOx / EE) in individuals with a moderate to vigorous physical activity (MVPA) habit. Figure 9 This suggests that exercise is an important modifiable factor in improving the metabolic adaptation of Cluster 2, and its benefits are particularly pronounced for those who already have a habit of moderate to high-intensity exercise.

[0185] For the validation population, Cluster 3, the improvement in weight loss and glucose homeostasis indicators (such as fasting insulin, HOMA-IR, and mean CGM blood glucose) under the control diet (CD) intervention was lower than that in Cluster 1. However, there was no difference in weight loss and glucose homeostasis improvement among the clusters under high-quality dietary patterns such as the Mediterranean diet (MD) and the traditional Jiangnan diet (TJD). Figure 10-12 This indicates that improving dietary quality is an important modifiable factor in enhancing Cluater 3 metabolic adaptation.

[0186] All documents mentioned in this invention are incorporated herein by reference as if each document were individually incorporated by reference. Furthermore, it should be understood that after reading the foregoing teachings of this invention, those skilled in the art can make various alterations or modifications to this invention, and these equivalent forms also fall within the scope defined by the appended claims.

Claims

1. The use of a clinical biomarker or its detection reagent, characterized in that, This is used to prepare diagnostic reagents or diagnostic kits, which are used for: i) classifying the metabolic response of the subject; ii) predicting weight changes in the subject after a weight loss intervention; The clinical biomarkers include: (A) Any biomarker selected from A1-A8, or a combination thereof: (A1) Respiratory quotient; (A2) Blood glucose level; (A3) Carbohydrate oxidation rate; (A4) Fat oxidation rate; (A6) Energy expenditure rate; (A7) Free fatty acids (FFA); (A8) Insulin / glucagon; (B) Markers indicated by B1 and / or B2: (B1) Postprandial total free fatty acid AUC; (B2) Fasting gamma-glutamyl transferase; (D) A combination of one or more markers from A1 to A8 and one or more markers from B1 to B2.

2. The use as described in claim 1, characterized in that, The clinical biomarkers also include: (B) Any biomarker selected from B3-B11, or a combination thereof: (B3) Albumin AUC; (B4) Total cholesterol iAUC; (B5) Systolic blood pressure; (B6) Free triiodothyronine; (B7) Uric acid AUC; (B8) Fasting glucose; (B9) Age; (B10) Gamma-glutamyl transferase AUC; (B11) Interferon-γ iAUC; (C) Any biomarker selected from C1-C3, or a combination thereof: (C1) sex; (C2) BMI; (C3) intervention compliance; (E) A combination of one or more markers from B3 to B11 and one or more markers from C1 to C3.

3. A method for constructing a model for metabolic response typing and / or predicting weight change after a weight loss intervention, characterized in that, The method includes the following steps: (Z1) Energy metabolism data, anthropometric data, continuous glucose monitoring (CGM) data, and multi-omics data were collected from subjects receiving a high-challenge diet; the energy metabolism data included raw time-series data of respiratory quotient (RQ) and various energy metabolism indicators; (Z2) Obtaining relevant biomarkers specifically includes the following steps: processing the energy metabolism data to obtain multiple representative respiratory quotient dynamic trajectory phenotypes; dividing the subjects into various representative respiratory quotient dynamic trajectories, detecting the dynamic trajectories of multiple energy metabolism indicators of the subjects in different clusters, and selecting energy metabolism indicators with different dynamic trajectories between different clusters as the relevant biomarkers. (Z3) Obtaining key biomarkers, specifically including the following steps: performing regression analysis on the anthropometric data, CGM data, and multi-omics data with the representative respiratory quotient dynamic trajectory phenotype to obtain the key biomarkers of the representative respiratory quotient dynamic trajectory phenotype; (Z4) The relevant biomarkers and key biomarkers are combined with the baseline model to construct a model for metabolic response typing and / or prediction of weight change after weight loss intervention.

4. The method as described in claim 3, characterized in that, The baseline model includes gender and age.

5. A system for classifying the metabolic response of a subject and / or predicting weight changes after a weight loss intervention, characterized in that, The system includes: An input unit is configured to input data, the input data including clinical biomarker data of the subject; the clinical biomarker data includes values ​​or change curves of the clinical biomarkers; the clinical biomarker data includes: (A) Any biomarker selected from A1-A8, or a combination thereof: (A1) Respiratory quotient; (A2) Blood glucose level; (A3) Carbohydrate oxidation rate; (A4) Fat oxidation rate; (A6) Energy expenditure rate; (A7) Free fatty acids; (A8) Insulin / glucagon; (B) Markers indicated by B1 and / or B2: (B1) Postprandial total free fatty acid AUC; (B2) Fasting gamma-glutamyl transferase; (D) A combination of one or more markers from A1 to A8 and one or more markers from B1 to B2; The typing and prediction unit is configured as a model that types the metabolic response of the subject to be tested based on the clinical biomarker data and / or predicts weight changes after weight loss intervention, thereby obtaining results; the model is constructed using the method of claim 2. An output unit is configured to output the results of the fractal and prediction unit.

6. The system as described in claim 5, characterized in that, The clinical biomarkers also include: (B) Any biomarker selected from B3-B11, or a combination thereof: (B3) Albumin AUC; (B4) Total cholesterol iAUC; (B5) Systolic blood pressure; (B6) Free triiodothyronine (FT3); (B7) Uric acid AUC; (B8) Fasting glucose; (B9) Age; (B10) Gamma-glutamyl transferase AUC; (B11) Interferon-γ iAUC; (C) Any biomarker selected from C1-C3, or a combination thereof: (C1) sex; (C2) BMI; (C3) intervention compliance; (E) A combination of one or more markers from B3 to B11 and one or more markers from C1 to C3.

7. A method for screening key biomarkers, characterized in that, The method includes the following steps: (C1) Collect energy metabolism data, anthropometric data, continuous glucose monitoring data, and multi-omics data from subjects receiving a high-challenge diet; the energy metabolism data includes raw time-series data of the respiratory quotient and various energy metabolism indicators; (C2) The original time series data of the respiratory quotient of each subject were transformed by wavelet transform and smoothing to obtain the smoothed time series data of the respiratory quotient of each subject. Clustering of the smoothed time series data of each subject's respiratory quotient was performed to obtain a representative dynamic trajectory phenotype of the respiratory quotient; (C3) The anthropometric data, continuous blood glucose monitoring data, and multi-omics data are subjected to regression analysis with the representative respiratory quotient dynamic trajectory phenotype to obtain the key biomarkers of the representative respiratory quotient dynamic trajectory phenotype.

8. The method as described in claim 7, characterized in that, The anthropometric data, continuous glucose monitoring data, and multi-omics data include: demographic characteristics, lifestyle indicators, anthropometric indicators, cardiovascular function indicators, blood glucose control and continuous glucose monitoring indicators, blood lipid indicators, organ function indicators, inflammatory markers, thyroid hormones, and metabolic hormones.

9. The method as described in claim 7, characterized in that, In step (C2), the wavelet transform is a discrete wavelet transform; the smoothing method is a local weighted regression scatter smoothing method; in the local weighted regression scatter smoothing method, the span parameter is set to 0.7±0.

1.

10. The method as described in claim 7, characterized in that, Step (C2) specifically includes the following steps: (c2.1) The original time series data of the respiratory quotient of each subject is decomposed using the Haar wavelet basis function, and a three-level decomposition scale is set; high-frequency fluctuations are identified by a threshold based on the absolute deviation of the median; thus obtaining the transformed respiratory quotient time series data. and (c2.2) The transformed respiratory quotient time series data of each subject was smoothed using the local weighted regression scatter smoothing method, with the span parameter set to 0.7; thus obtaining the smoothed time series data of the respiratory quotient of each subject. (c2.3) Use B-spline basis functions to model the smoothed time series data of the respiratory quotient of each subject to obtain a continuous smooth function of the respiratory quotient trajectory of each subject to obtain the continuous smooth function. (c2.4) Based on the continuous smooth function, the similarity of the respiratory quotient trajectory of each test object is calculated using the dynamic time warping algorithm to obtain the distance matrix; (c2.5) An unsupervised algorithm is used to perform cluster analysis on the distance matrix to obtain a representative dynamic trajectory phenotype of the respiratory quotient.

Images