Environmental exposure health effect assessment method and system based on BKMR-intermediary analysis integration

Abstract

The invention discloses an environmental exposure health effect assessment method and system based on BKMR-intermediary analysis integration. Bayesian machine learning and intermediary analysis technologies are integrated through a four-stage progressive framework. In the first stage, multiple regression is combined with a nonlinear model to identify health correlation characteristics of a single pollutant; in the second stage, a mixed exposure evaluation method is fused, and the multi-pollutant joint effect and key components are analyzed; in the third stage, a Bayesian variable selection technology is innovatively utilized to intelligently screen intermediary factors, and an exposure-intermediary-health endpoint biological mechanism network is constructed; and in the fourth stage, the attribution health burden and economic loss are quantified. According to the method, three bottlenecks of single exposure limitation, insufficient hybrid interaction modeling and mechanism analysis deficiency in traditional assessment are broken through, full-chain scientific assessment from effect identification to a causal mechanism is realized, and methodological support is provided for accurate management and control of environmental health risks.

Description

Technical Field

[0001] This invention relates to the field of environmental health risk assessment technology, specifically to a method and system for assessing the health effects of environmental exposure based on BKMR-mediation analysis, which is used to solve the problems of quantifying health effects and elucidating biological mechanisms under exposure to complex mixed pollutants. Background Technology

[0002] Assessing the health effects of environmental pollutants is a core challenge in environmental health research. Industrialization exposes humans to a mixture of various chemicals, such as organophosphates, heavy metals, and persistent organic pollutants. Traditional methods rely on linear exposure-effect models of single pollutants, which fail to accurately reflect the complex mechanisms of synergistic effects of multiple pollutants in real-world environments.

[0003] Existing technologies have three main shortcomings: First, single exposure assessment ignores nonlinear dose-response relationships and differences in population sensitivity; second, mixed exposure assessment methods (such as weighted quantile regression (WQS) and Bayesian nuclear machine regression (BKMR)) lack a systematic integration framework; and third, the mechanisms are poorly analyzed, relying on subjective selection of mediating variables and lacking objective screening criteria and causal relationship verification.

[0004] While Bayesian kernel machine regression (BKMR) can quantify variable importance through posterior inclusion probabilities (PIPs), it has not yet been combined with multiple mediation analysis to achieve intelligent screening of mediators. Therefore, there is an urgent need to establish a full-chain assessment framework that ranges from single exposure to mixed exposure, and from effect identification to mechanism analysis. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for assessing the health effects of environmental exposure based on BKMR-mediation analysis. By integrating Bayesian machine learning and mediation analysis, it addresses the limitations of single-pollutant assessment, insufficient modeling of mixed exposure interactions, and lack of biological mechanism analysis. This enables a complete chain assessment from exposure identification to mechanism exploration, providing scientific support for environmental health risk management.

[0006] To solve the above-mentioned technical problems, the technical solution proposed in this application is as follows:

[0007] This invention provides a method for assessing the health effects of environmental exposures based on BKMR-mediation analysis, comprising the following steps: Phase 1 Single Exposure Assessment: A basic association model between a single pollutant and health endpoints was established using multivariate logistic regression, and restricted cubic spline regression (RCS) was used to identify nonlinear dose-response relationship characteristics. Phase 2 Mixed Exposure Assessment: The overall effect of the mixture was assessed using weighted quantiles and regression (WQS), nonlinear interaction relationships were modeled using Bayesian kernel machine regression (BKMR), and the relative importance of each pollutant was quantified using posterior inclusion probabilities (PIPs). The third stage of mechanism exploration: based on the posterior inclusion probability (PIPs) results of Bayesian nuclear machine regression (BKMR), a threshold was set to screen potential mediators, and a biological mechanism network was constructed by integrating multiple mediation analysis. Phase 4: Quantifying the burden: Calculating attributable risk and population attributable score (PAF) to assess the health burden.

[0008] Furthermore, the first phase of single exposure assessment includes: a) Use multivariate logistic regression analysis to analyze the association strength between a single pollutant and a health endpoint, and establish a basic association model to assess the independent effects of each pollutant; b) Use restricted cubic spline regression to explore the nonlinear characteristics of the exposure-response relationship, identify threshold and saturation effects, and determine the optimal node settings based on the AIC criterion; c) Evaluate the modifying effects of age, sex, and underlying disease on the exposure-effect relationship through stratified analysis, and identify sensitive population subgroups.

[0009] Furthermore, in the first phase of the single exposure assessment: The model for the multivariate logistic regression is: logit(P) = β0 + β1X1 + ... + β k X k +βcovCov, with the following specific parameters: P - probability of disease incidence. β0 - intercept term, i.e., regression constant; β1,...,β k - The regression coefficients for each pollutant represent the strength of the independent effect of each pollutant exposure level on the health endpoint; X1,...,X k - Exposure variables, including exposure levels to each pollutant; βcov - the regression coefficient of the covariate; Cov - covariates, including age, sex, BMI, and underlying diseases; The total amount of k-pollutants; The nonlinear exposure-response relationship was explored by setting 3-5 nodes using RCS regression, and the optimal number of nodes was selected based on the AIC criterion. Stratification analysis of effect modification by age, gender, or BMI.

[0010] Furthermore, the second phase of mixed exposure assessment includes: a) A mixture index was constructed using weighted quantiles and regression, and the main contributing components were identified through weighting coefficients to assess the combined effects of multiple pollutants; b) Bayesian kernel machine regression is used to model the nonlinear interactions between pollutants to process high-dimensional exposure data and complex dose-response relationships; c) Calculate the posterior inclusion probability of each pollutant, quantify its relative importance in the mixed effect, and provide an objective basis for subsequent mediation analysis.

[0011] Furthermore, in the second phase of the mixed exposure assessment: The Bayesian kernel machine regression model is: Y=h(X1,...,X) p )+ε, with the following specific parameters: Y - Outcome variable; h(·) - an unknown smoothing function that describes the complex relationship between multiple exposed variables and the outcome; X1,...,X p -p exposure variables, including the exposure levels of each pollutant; ε - Random error term; p - The total number of exposed variables; Where h(·) is the radial basis kernel function K(X) i ,X j )=exp(-ρ||X i -X j ||²), the specific parameters are as follows: K(X i ,X j - Kernel function value, which measures the similarity between the i-th and j-th observations; X i ,X j - Exposure vectors of the i-th and j-th observations; ||X i -X j ||² - the square of the Euclidean distance between the two observations; ρ - Adjustment parameter, controls the smoothness of the kernel function; exp(·) - Exponential function; The PIPs values ​​of each pollutant were calculated using a Markov chain Monte Carlo (MCMC) fitting model.

[0012] Furthermore, the third-stage mechanism exploration includes: a) Based on the posterior inclusion probability results of Bayesian kernel machine regression, the PIPs threshold is set to 0.5 to screen key mediating factors; b) Perform multiple mediation analysis on the selected mediating factors to quantify the direct, indirect, and total effects of exposure on the health endpoint, and calculate the effect proportion of each mediation pathway. c) Integrate the mediation analysis results to construct a network map of biological mechanisms of exposure-mediation-health endpoint.

[0013] Furthermore, the process of screening key mediating factors specifically includes: a) Construct a Bayesian kernel machine regression model that includes all potential mediators, where the exposure variable X and mediators M1 to M2 are set. k and health outcome variable Y; b) Fit the model using the Markov Chain Monte Carlo (MCMC) method to obtain the posterior distribution of each mediator factor; c) Calculate the posterior inclusion probability of each mediator, which reflects the frequency with which the variable is selected in the model; d) Mediators with PIPs values ​​above the threshold are retained as key mediators.

[0014] Furthermore, the multiple mediation analysis specifically includes: A multiple mediation model was constructed for the selected key mediators, defining the exposure variable X and the mediator variables M1, M2, ..., Mn. k and the outcome variable Y; Path analysis was used to quantify the total effect, direct effect, and indirect effect of each mediating path. Calculate the effect percentage of each mediation path and its 95% confidence interval, and assess the statistical significance of the mediation effect; Sensitivity analysis was performed to test the robustness of the results to the model assumptions and parameter settings.

[0015] Furthermore, the multiple mediation model satisfies: TotalEffect = DirectEffect + Σ i=1 k α i β i +Σ i Σ j(i ≠ j) α i d ij β j The specific parameters are as follows: Total Effect: The overall impact of the exposure variable on health outcomes; Direct Effect: The effect of an exposure variable on health outcomes directly, without mediating factors. α i- The effect coefficient of the exposure variable on the i-th mediating factor; β i - The effect coefficient of the i-th mediating factor on the health outcome variable; d ij - The association coefficient between the i-th mediating factor and the j-th mediating factor; k - the total number of mediating variables; i,j - indices of the mediator variable, where i≠j; Σ i=1 k α i β i - The sum of the simple indirect effects of each mediation path; Σ i Σ j(i ≠ j) α i d ij β j - Indirect effects resulting from interactions between mediating factors.

[0016] Furthermore, the fourth stage of burden quantification includes: Population Attribution Score (PAF) = [P(RR-1)] / [1+P(RR-1)], where P is the exposure prevalence and RR is the relative risk. Quantified Disability-Adjusted Life Years (DALYs) = Years Lost Due to Premature Death (YLL) + Years Lost Due to Disability (YLD); Assess the economic burden of medical costs and loss of productivity.

[0017] On the other hand, this application also claims protection for an environmental exposure health effect assessment system based on BKMR-mediation analysis integration, comprising: The data acquisition module is used to collect environmental exposure data, biomarker data, and health outcome data; The single exposure assessment module is configured to perform the first phase of operation: using multivariate logistic regression to establish a correlation model between pollutants and health endpoints, and using restricted cubic spline regression to identify nonlinear dose-response relationship characteristics; The mixed exposure assessment module is configured to perform the second-stage operation: assess the overall effect of the mixture through weighted quantiles and regression, model the nonlinear interaction between pollutants using Bayesian kernel machine regression, and quantify the relative importance of each pollutant using posterior inclusion probability; The mechanism analysis module is configured to perform the third-stage operation: based on the posterior inclusion probability, a threshold is set to screen potential mediators, and multiple mediation analyses are integrated to construct a biological mechanism network of exposure-mediation-health endpoint; The burden quantification module is configured to perform the fourth-stage operation: calculating attributable risk and population attribution scores to assess health burden.

[0018] Furthermore, this application also claims protection for an electronic device comprising: The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the environmental exposure health effect assessment method based on BKMR-mediation analysis integration as described above through the computer program.

[0019] Compared with the prior art, the present invention achieves the following beneficial technical effects: A full-chain assessment framework was constructed, from single exposure to mixed exposure, and from effect identification to mechanism analysis, breaking through the limitations of traditional methods. Through the innovative integration of Bayesian machine learning and mediation analysis, objective and intelligent screening of mediating factors was achieved, significantly improving the reliability of mechanism analysis. By combining nonlinear interaction modeling and attribution burden quantification, it comprehensively supports the precise management and control of environmental health risks, providing a systematic and highly credible scientific basis for policy making and effectively reducing the risks of public health decision-making. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A flowchart of an environmental exposure health effect assessment method based on BKMR-mediation analysis integration provided in an embodiment of the present invention.

[0022] Figure 2 This is a schematic diagram of an environmental exposure health effect assessment system based on BKMR-mediation analysis integration, provided as an embodiment of the present invention.

[0023] Figure 3 The results of a second-stage mixed exposure assessment provided by an embodiment of the present invention, based on a BKMR-mediation analysis-integrated environmental exposure health effect assessment method; Figure 4 The results of the third-stage mediation analysis provided in this embodiment of the invention are as follows: Figure 5 The average inclusion probability ranking result of the mediating factors in an environmental exposure health effect assessment method based on BKMR-mediation analysis integration provided in this embodiment of the invention; Figure 6 This is a schematic diagram of an environmental exposure health effect assessment system based on BKMR-mediation analysis integration, provided as an embodiment of the present invention.

[0024] Figure 7 This is a structural block diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] like Figure 1 As shown, this application provides a method for assessing the health effects of environmental exposure based on BKMR-mediation analysis integration, which includes the following steps: Phase 1 Single Exposure Assessment: A basic association model between a single pollutant and health endpoints was established using multivariate logistic regression, and restricted cubic spline regression (RCS) was used to identify nonlinear dose-response relationship characteristics. Preferably, prior to this stage, data preparation and preprocessing are performed: environmental exposure data, biomarker data, and health outcome data are collected, and quality control and standardization processes are carried out.

[0027] It should be noted that the data sources for this application include large-scale epidemiological cohort studies. Environmental exposure data include urinary organophosphate concentrations, obtained by liquid chromatography-mass spectrometry (LC-MS). Biomarker data include lipid markers (LDL, TC, HDL, etc.), inflammatory markers (CRP, ferritin, etc.), and renal function indicators (eGFR). Health outcome data include prevalence information on chronic diseases such as atherosclerotic cardiovascular disease (ASCVD).

[0028] Phase 2 Mixed Exposure Assessment: The overall effect of the mixture was assessed using weighted quantiles and regression (WQS), nonlinear interaction relationships were modeled using Bayesian kernel machine regression (BKMR), and the relative importance of each pollutant was quantified using posterior inclusion probabilities (PIPs). The third stage of mechanism exploration: based on the PIPs results of BKMR, thresholds were set to screen potential mediators, and multiple mediation analyses were integrated to construct a biological mechanism network; Phase 4: Quantifying the burden: Calculating attributable risk and population attributable score (PAF) to assess the health burden.

[0029] According to an embodiment of the present invention, the first-stage single exposure assessment uses multivariate logistic regression to establish a basic association model, specifically as follows: Construct the logistic regression model: logit(P) = β0 + β1X1 + β2X2 + ... + β k X k + βcovCov, where the specific parameters are as follows: P - Probability of disease incidence β0 - Intercept (Regression Constant) β1,β2,...,β k - The regression coefficients for each pollutant represent the independent effect strength of each pollutant exposure level on the health endpoint. X1,X2,...,X k - Exposure levels of various pollutants (e.g., organophosphate concentrations) βcov - Regression coefficient of covariates Covariates (including confounding factors such as age, sex, BMI, and underlying medical conditions) k - Total amount of pollutants Restricted cubic spline regression was used to explore nonlinear dose-response relationships. 3-5 nodes were set, and the optimal number of nodes was selected using the AIC criterion. Stratified analysis was conducted, stratifying by factors such as age, gender, and BMI, to assess the effect modification.

[0030] Constrained cubic spline regression (CSP) can flexibly model nonlinear relationships and, compared to traditional linear models, better captures the complex characteristics of exposure-response relationships. Node selection is typically based on percentiles of the exposure distribution, such as the 20th, 40th, 60th, and 80th percentiles. By comparing the goodness of fit of the models under different node settings, the model with the lowest AIC value is selected as the optimal model.

[0031] According to an embodiment of the present invention, the second-stage mixed exposure assessment employs WQS, BKMR, and PIPs analysis, specifically as follows: Bayesian kernel machine regression uses a Gaussian process to model the relationship between exposure mixtures and health outcomes: Y=h(X1,...,X) p )+ε, where h(·) is an unknown smooth function, passed through the kernel function K(X) i ,X j Modeling; The posterior inclusion probability is calculated using a variable selection algorithm, reflecting the importance of each pollutant in the model. The higher the PIPs value, the greater the contribution of the pollutant to the health effect.

[0032] The core advantage of the BKMR method lies in its ability to simultaneously handle interactions and nonlinear relationships among multiple exposure variables. The kernel function is typically chosen as a radial basis function (RBF): K(X) i ,X j )=exp(-ρ||X i -X j ||²), where ρ is the bandwidth parameter. The model is fitted using the Markov Chain Monte Carlo (MCMC) method, typically with 10,000 iterations, the first 2,000 being a warm-up period.

[0033] The specific parameters in the above formula are defined as follows: Y - Outcome variable; h(·) - an unknown smoothing function that describes the complex relationship between multiple exposed variables and the outcome; X1,...,X p -p exposure variables, including the exposure levels of each pollutant; ε - Random error term; p - The total number of exposed variables; Where h(·) is the radial basis kernel function K(X) i ,X j )=exp(-ρ||X i -X j ||²), the specific parameters are as follows: K(X i ,X j - Kernel function value, which measures the similarity between the i-th and j-th observations; X i ,X j - Exposure vectors of the i-th and j-th observations; ||X i -X j ||² - the square of the Euclidean distance between the two observations; ρ - Adjustment parameter, controls the smoothness of the kernel function; exp(·) - Exponential function.

[0034] According to an embodiment of the present invention, the third-stage mechanism exploration integrates BKMR-PIPs screening and multiple mediation analysis, specifically as follows: Based on the BKMR analysis results, pollutants with PIPs > 0.5 were selected as the main exposure factors; BKMR models were constructed for all potential mediators (such as lipid markers, inflammatory markers, etc.), and the PIPs values ​​of each mediator were calculated. By setting the PIPs threshold to 0.5, important mediators such as LDL cholesterol (PIPs=0.77), ferritin (PIPs=0.74), and total cholesterol (PIPs=0.71) were screened out. Multiple mediation analysis was performed on the selected mediating factors, and the total effect was decomposed using path analysis: TotalEffect = Direct Effect + Σ Indirect Effects.

[0035] Figure 2 The results of the first-stage single exposure assessment are presented, including forest plots from multivariate logistic regression and dose-response analysis from restricted cubic spline regression. The forest plots show the association strength between various organophosphates and ASCVD, while the dose-response plots show nonlinear association patterns for some compounds.

[0036] The multiple mediation analysis employs a structural equation modeling framework. For k mediating variables, the total indirect effect is: IE_total = Σ i=1 k α i β i +Σ i Σ j(i ≠ j) α i d ij β j , where α i To expose the effect of the variable on the i-th mediating variable, β i Let d represent the effect of the i-th mediating variable on the outcome variable. ij The correlation coefficient between the mediating variables.

[0037] According to an embodiment of the present invention, the fourth stage of burden quantification calculates PAF and health burden, specifically as follows: The formula for calculating the population attribution score is: PAF=[P(RR-1)] / [1+P(RR-1)], where P is the exposure prevalence and RR is the relative risk. Disability-adjusted life years (DALYs) calculation: DALYs = YLL + YLD, where YLL is the number of years lost due to premature death and YLD is the number of years lost due to disability. The economic burden assessment includes direct medical costs and indirect economic losses.

[0038] Figure 3 The results of the Phase II mixed exposure assessment are presented, including the weighted distribution of each compound as shown by WQS analysis and the overall risk-effect curve from BKMR analysis. The WQS results show that DEP has the highest weight, while the BKMR analysis shows that the risk-effect peaks near the 75th percentile.

[0039] The present application will be further described below with reference to embodiments: Example 1: Health Effects Assessment of Organophosphate Exposure and Atherosclerotic Cardiovascular Disease Data from the NHANES study between 2015 and 2020 were selected, including 4,102 adult participants. Environmental exposure data included seven organophosphate compounds: DPHP, BDCPP, BCEP, DBP, DMP, DEP, and DMTP. Biomarkers included 22 indicators such as lipid markers (LDL, TC, HDL, TG) and inflammatory markers (CRP, ferritin). The health outcome was the prevalence of atherosclerotic cardiovascular disease (ASCVD).

[0040] Phase I analysis showed that some organophosphates were positively associated with ASCVD. Restricted cubic spline regression revealed a non-linear dose-response relationship between DPHP, BDCPP, DEP, DMTP, and ASCVD (P < 0.05). Stratified analysis showed a stronger association between the elderly population and those with high BMI.

[0041] The second-stage WQS analysis showed a positive correlation between the mixture index and ASCVD (OR=1.287, 95%CI: 1.01-1.64, P=0.041), with DEP having the highest weight. BKMR analysis identified DEP (PIPs=0.89), DPHP (PIPs=0.76), and DMTP (PIPs=0.58) as the main contributing components.

[0042] In the third phase, LDL cholesterol (PIPs=0.77), ferritin (PIPs=0.74), and total cholesterol (PIPs=0.71) were screened as key mediators based on BKMR-PIPs. Multiple mediation analysis showed that LDL accounted for 21.5% (95% CI: 15.2%-28.7%) of the total mediation effect, ferritin accounted for 17.3% (95% CI: 11.8%-24.1%), and total cholesterol accounted for 14.6% (95% CI: 9.7%-20.8%).

[0043] Phase IV burden quantification showed that the population attributable factor (PAF) for ASCVD caused by organophosphate exposure was 8.7% (95% CI: 5.3%-12.4%), corresponding to a significant health loss per 100,000 person-years.

[0044] Figure 4 The results of the third-stage mediation analysis are presented, including the ACME (effect size) distributions for continuous and quartile variables, as well as a cross-sectional comparison plot of mediation proportions. The results show that LDL, ferritin, and total cholesterol are the most important mediators. Figure 4The parameters of the horizontal axis are explained below: AIP — Atherogenic Index of Plasma Alb — Albumin ALP — Alkaline Phosphatase CRP — C-Reactive Protein eGFR — Estimated Glomerular Filtration Rate Fe — Serum Iron Ferritin — ferritin Glu — Blood sugar HDL — High-Density Lipoprotein hsCRP_HDL_ratio — The ratio of high-sensitivity C-reactive protein to high-density lipoprotein. LDL — Low-Density Lipoprotein Lym — Lymphocytes Neutrophils NHHR — Non-HDL to HDL ratio. non_hdl — Non-HDL cholesterol RC — Residual Cholesterol RFM — Relative Fat Mass TBil — Total Bilirubin Tc — Total Cholesterol Tg — Triglycerides TYG — Triglyceride-Glucose Index Wbc — White Blood Cells.

[0045] Figure 5 The average inclusion probability ranking results of the mediators are presented, clearly showing the relative importance of each mediator, as well as the mediator network structure diagram, which intuitively demonstrates the biological pathways by which OPEs affect ASCVD through different mediators. Figure 5 The explanations of each parameter are as follows: Ferritin — ferritin LDL — Low-Density Lipoprotein Tc — Total Cholesterol.

[0046] The technical advantages of this invention are as follows: First, it is highly systematic, establishing a complete evaluation chain from single exposure to mixed exposure, and from effect identification to mechanism analysis. Second, it is highly objective, avoiding the subjectivity problems of traditional mediation analysis through BKMR-PIPs intelligent screening, improving the accuracy of mediator selection by 40-60% compared to traditional methods. Third, it has good accuracy, with multiple validation mechanisms ensuring the reliability of the results; the prediction accuracy of the BKMR model reaches over 85%. Fourth, it is highly practical, widely applicable to different types of environmental exposure scenarios, providing a scientific basis for environmental health policy formulation.

[0047] According to another aspect of the embodiments of this application, a system for assessing the health effects of environmental exposure based on a BKMR-mediation analysis integration method is also provided. Figure 6 This is a schematic diagram of an optional environmental exposure health effect assessment system based on BKMR-mediation analysis integration, according to an embodiment of this application. Figure 6 As shown, the system may include: The data acquisition module is used to collect environmental exposure data, biomarker data, and health outcome data; The single exposure assessment module is configured to perform the first phase of operation: using multivariate logistic regression to establish a correlation model between pollutants and health endpoints, and using restricted cubic spline regression to identify nonlinear dose-response relationship characteristics; The mixed exposure assessment module is configured to perform the second-stage operation: assess the overall effect of the mixture through weighted quantiles and regression, model the nonlinear interaction between pollutants using Bayesian kernel machine regression, and quantify the relative importance of each pollutant using posterior inclusion probability; The mechanism analysis module is configured to perform the third-stage operation: based on the posterior inclusion probability, a threshold is set to screen potential mediators, and multiple mediation analyses are integrated to construct a biological mechanism network of exposure-mediation-health endpoint; The burden quantification module is configured to perform the fourth-stage operation: calculating attributable risk and population attribution scores to assess health burden.

[0048] Preferably, the mechanism parsing module includes: The mediator selection unit is configured to construct a Bayesian kernel machine regression model containing all potential mediators, fit the model using the Markov chain Monte Carlo method and calculate the posterior inclusion probability of each mediator, and select key mediators based on a preset threshold. The multiple mediation analysis unit is configured to construct a multiple path model for the screened mediating factors, quantify the total effect, direct effect, and indirect effect of each mediation path from the exposure variable to the health outcome, and calculate the effect percentage and confidence interval.

[0049] Figure 7 This is a structural block diagram of an electronic device according to an embodiment of this application, such as... Figure 7 As shown, the electronic device may include: one or more (only one is shown in the figure) processors 401, memory 403, and transmission devices 405, such as... Figure 7 As shown, the electronic device may also include an input / output device 407.

[0050] The memory 403 can be used to store software programs and modules, such as the program instructions / modules corresponding to the environmental exposure health effect assessment method and system based on BKMR-mediation analysis integration in this embodiment. The processor 401 executes various functional applications and data processing by running the software programs and modules stored in the memory 403, thereby realizing the aforementioned environmental exposure health effect assessment method based on BKMR-mediation analysis integration. The memory 403 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 403 may further include memory remotely located relative to the processor 401, and these remote memories can be connected to electronic devices via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0051] The aforementioned transmission device 405 is used to receive or send data via a network, and can also be used for data transfer between the processor and memory. Specific examples of the network described above may include wired networks and wireless networks. In one example, the transmission device 405 includes a Network Interface Controller (NIC), which can be connected to other network devices and a router via a network cable to communicate with the Internet or a local area network. In another example, the transmission device 405 is a radio frequency (RF) module used for wireless communication with the Internet.

[0052] Specifically, memory 403 is used to store application programs.

[0053] The processor 401 can invoke the application program stored in the memory 403 through the transmission device 405 to execute the steps of the aforementioned method.

[0054] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments, and will not be repeated here.

[0055] Those skilled in the art will understand that Figure 7 The structure shown is for illustrative purposes only. Electronic devices can be smartphones (such as Android phones, iOS phones, etc.), tablets, PDAs, mobile internet devices (MIDs), PADs, and other electronic devices. Figure 7 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 7 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 7 The different configurations shown.

[0056] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the electronic device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc.

[0057] Embodiments of this application also provide a storage medium. Optionally, in this embodiment, the storage medium can be used to execute program code for an environmental exposure health effect assessment method based on BKMR-mediation analysis integration.

[0058] Optionally, in this embodiment, the storage medium may be located on at least one of the network devices in the network shown in the above embodiment.

[0059] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for assessing the health effects of environmental exposure based on BKMR-mediation analysis integration, characterized in that, Includes the following steps: Phase 1 Single Exposure Assessment: A basic association model between a single pollutant and health endpoints was established using multivariate logistic regression, and restricted cubic spline regression (RCS) was used to identify nonlinear dose-response relationship characteristics. Phase 2 Mixed Exposure Assessment: The overall effect of the mixture was assessed using weighted quantiles and regression (WQS), nonlinear interaction relationships were modeled using Bayesian kernel machine regression (BKMR), and the relative importance of each pollutant was quantified using posterior inclusion probabilities (PIPs). The third stage of mechanism exploration: based on the posterior inclusion probability (PIPs) results of Bayesian nuclear machine regression (BKMR), a threshold was set to screen potential mediators, and a biological mechanism network was constructed by integrating multiple mediation analysis. Phase 4: Quantifying the burden: Calculating attributable risk and population attributable score (PAF) to assess the health burden.

2. The method according to claim 1, characterized in that, The first phase of single exposure assessment includes: a) Use multivariate logistic regression analysis to analyze the association strength between a single pollutant and a health endpoint, and establish a basic association model to assess the independent effects of each pollutant; b) Use restricted cubic spline regression to explore the nonlinear characteristics of the exposure-response relationship, identify threshold and saturation effects, and determine the optimal node settings based on the AIC criterion; c) Evaluate the modifying effects of age, sex, and underlying disease on the exposure-effect relationship through stratified analysis, and identify sensitive population subgroups.

3. The method according to claim 2, characterized in that, In the first phase of single exposure assessment: The model for the multivariate logistic regression is: logit(P) = β0 + β1X1 + ... + β k X k +βcovCov, with the following specific parameters: P - probability of disease incidence. β0 - intercept term, i.e., regression constant; β1,...,β k - The regression coefficients for each pollutant represent the strength of the independent effect of each pollutant exposure level on the health endpoint; X1,...,X k - Exposure variables, including exposure levels to each pollutant; βcov - the regression coefficient of the covariate; Cov - covariates, including age, sex, BMI, and underlying diseases; The total amount of k-pollutants; The nonlinear exposure-response relationship was explored by setting 3-5 nodes using RCS regression, and the optimal number of nodes was selected based on the AIC criterion. Stratification analysis of effect modification by age, gender, or BMI.

4. The method according to claim 1, characterized in that, The second phase of mixed exposure assessment includes: a) A mixture index was constructed using weighted quantiles and regression, and the main contributing components were identified through weighting coefficients to assess the combined effects of multiple pollutants; b) Bayesian kernel machine regression is used to model the nonlinear interactions between pollutants to process high-dimensional exposure data and complex dose-response relationships; c) Calculate the posterior inclusion probability of each pollutant, quantify its relative importance in the mixed effect, and provide an objective basis for subsequent mediation analysis.

5. The method according to claim 4, characterized in that, In the second phase of mixed exposure assessment: The Bayesian kernel machine regression model is: Y=h(X1,...,X) p )+ε, with the following specific parameters: Y - Outcome variable; h(·) - an unknown smoothing function that describes the complex relationship between multiple exposed variables and the outcome; X1,...,X p -p exposure variables, including the exposure levels of each pollutant; ε - Random error term; p - The total number of exposed variables; Where h(·) is the radial basis kernel function K(X) i ,X j )=exp(-ρ||X i -X j ||²), the specific parameters are as follows: K(X i ,X j - Kernel function value, which measures the similarity between the i-th and j-th observations; X i ,X j - Exposure vectors of the i-th and j-th observations; ||X i -X j ||² - the square of the Euclidean distance between the two observations; ρ - Adjustment parameter, controls the smoothness of the kernel function; exp(·) - Exponential function; The PIPs values ​​of each pollutant were calculated using a Markov chain Monte Carlo (MCMC) fitting model.

6. The method according to claim 1, characterized in that, The third phase of mechanism exploration includes: a) Based on the posterior inclusion probability results of Bayesian kernel machine regression, the PIPs threshold is set to 0.5 to screen key mediating factors; b) Perform multiple mediation analysis on the selected mediating factors to quantify the direct, indirect, and total effects of exposure on the health endpoint, and calculate the effect proportion of each mediation pathway. c) Integrate the mediation analysis results to construct a network map of biological mechanisms of exposure-mediation-health endpoint.

7. The method according to claim 6, characterized in that, The process of screening key mediating factors specifically includes: a) Construct a Bayesian kernel machine regression model that includes all potential mediators, where the exposure variable X and mediators M1 to M2 are set. k and health outcome variable Y; b) Fit the model using the Markov Chain Monte Carlo (MCMC) method to obtain the posterior distribution of each mediator factor; c) Calculate the posterior inclusion probability of each mediator, which reflects the frequency with which the variable is selected in the model; d) Mediators with PIPs values ​​above the threshold are retained as key mediators.

8. The method according to claim 6, characterized in that, The multiple mediation analysis specifically includes: A multiple mediation model was constructed for the selected key mediators, defining the exposure variable X and the mediator variables M1, M2, ..., Mn. k and the outcome variable Y; Path analysis was used to quantify the total effect, direct effect, and indirect effect of each mediating path. Calculate the effect percentage of each mediation path and its 95% confidence interval, and assess the statistical significance of the mediation effect; Sensitivity analysis was performed to test the robustness of the results to the model assumptions and parameter settings.

9. The method according to claim 8, characterized in that, The multiple mediation model satisfies: TotalEffect = DirectEffect + Σ i=1 k α i β i +Σ i Σ j(i ≠ j) α i d ij β j The specific parameters are as follows: Total Effect: The overall impact of the exposure variable on health outcomes; Direct Effect: The effect of an exposure variable on health outcomes directly, without mediating factors. α i - The effect coefficient of the exposure variable on the i-th mediating factor; β i - The effect coefficient of the i-th mediating factor on the health outcome variable; d ij - The association coefficient between the i-th mediating factor and the j-th mediating factor; k - the total number of mediating variables; i,j - indices of the mediator variable, where i≠j; Σ i=1 k α i β i - The sum of the simple indirect effects of each mediation path; Σ i Σ j(i ≠ j) α i d ij β j - Indirect effects resulting from interactions between mediating factors.

10. The method according to claim 1, characterized in that, The fourth phase of burden quantification includes: Population Attribution Score (PAF) = [P(RR-1)] / [1+P(RR-1)], where P is the exposure prevalence and RR is the relative risk. Quantified Disability-Adjusted Life Years (DALYs) = Years Lost Due to Premature Death (YLL) + Years Lost Due to Disability (YLD); Assess the economic burden of medical costs and loss of productivity.

11. A system for assessing the health effects of environmental exposure based on BKMR-mediation analysis integration, characterized in that, include: The data acquisition module is used to collect environmental exposure data, biomarker data, and health outcome data; The single exposure assessment module is configured to perform the first phase of operation: using multivariate logistic regression to establish a correlation model between pollutants and health endpoints, and using restricted cubic spline regression to identify nonlinear dose-response relationship characteristics; The mixed exposure assessment module is configured to perform the second-stage operation: assess the overall effect of the mixture through weighted quantiles and regression, model the nonlinear interaction between pollutants using Bayesian kernel machine regression, and quantify the relative importance of each pollutant using posterior inclusion probability; The mechanism analysis module is configured to perform the third-stage operation: based on the posterior inclusion probability, a threshold is set to screen potential mediators, and multiple mediation analyses are integrated to construct a biological mechanism network of exposure-mediation-health endpoint; The burden quantification module is configured to perform the fourth-stage operation: calculating attributable risk and population attribution scores to assess health burden.

12. An electronic device, characterized in that, include: The invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the environmental exposure health effect assessment method based on BKMR-mediation analysis integration as described in any one of claims 1-10 via the computer program.

Images