A nonlinear proteomics causal association analysis method based on genetic scoring
By combining the genetic scoring model with the restricted cubic spline model, the problem of linear assumption between protein levels and disease risk in existing technologies was solved, nonlinear proteomics causal association analysis was realized, and the accuracy of analysis and the precision of drug dosage guidance were improved.
Patent Information
- Application Number
- CN202411373000.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-29
- Publication Date
- 2025-10-14
- Estimated Expiration
- 2044-09-29
AI Technical Summary
Existing proteomic causal association analysis models mainly assume a linear correlation between protein levels and disease risk, and are unable to implement nonlinear association analysis, resulting in the inability to reveal the complex regulatory mechanisms and progression patterns of the disease, and difficulty in providing accurate drug dosage guidance.
A genetic scoring model was used to predict protein expression levels. The correlation coefficient was calculated by matching the actual measured values in the database. A restricted cubic spline model was used to fit nonlinear causal associations. A generalized linear model was combined to identify significantly associated proteins, and a dose-response curve was established to provide the optimal drug dose.
It realizes the nonlinear correlation analysis between protein levels and disease risk, improves the accuracy and stability of causal association analysis, and can identify new associated proteins and provide guidance on the optimal dosage of targeted drugs.
Smart Images

Figure CN119360972B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of proteomics causal association analysis, and more specifically, relates to a nonlinear proteomics causal association analysis method based on genetic scoring. Background Art
[0002] With the continuous development of genome sequencing technology, many genetic loci associated with disease risk have been identified through genome-wide association studies (GWAS). In addition, with the continuous development of protein sequencing technology (including antibody-based Olink technology and aptamer-based SomaScan technology), more and more publicly available proteomic data are being used to further study the regulatory mechanisms of disease pathogenesis and key therapeutic targets. How to integrate these proteomic data to analyze the role of GWAS susceptibility loci in the disease pathogenesis and locate key drug targets is of great significance. Therefore, the proteome-wide association study (PWAS) model was developed. The core idea of this type of method is to conduct association studies based on predicted protein levels and diseases to obtain conclusions of causal relationships.
[0003] However, most existing PWAS models assume a linear correlation between protein levels and disease risk, making it impossible to perform nonlinear association analysis. The identification of nonlinear proteomic causal relationships has important biological significance: first, it helps reveal the underlying complex regulatory mechanisms and progression patterns of diseases; second, it provides information for the development of protein-targeted drugs and helps determine the optimal drug dose in subsequent clinical trials; finally, known nonlinear causal relationships can also help researchers establish more accurate disease prediction models, providing guidance for disease prevention and control. Summary of the Invention
[0004] In response to the above-mentioned deficiencies or improvement needs of the prior art, the present invention provides a nonlinear proteomics causal association analysis method based on genetic scoring, which aims to realize nonlinear association analysis between protein levels and disease risk and improve the accuracy of causal association analysis.
[0005] To achieve the above objectives, according to a first aspect of the present invention, a nonlinear proteomics causal association analysis method based on genetic scoring is provided, comprising:
[0006] S1. Use the genetic scoring model to predict the expression levels of different proteins in each sample;
[0007] S2. Matching each predicted protein expression level with the actual measured protein expression level in the database, wherein the matching includes: consistency matching between the sample and the protein name corresponding to the protein expression level; and screening out the predicted protein expression levels that pass the matching;
[0008] S3. Calculating the correlation coefficient between the expression level of the screened matched protein and its corresponding actual measured value;
[0009] S4. Use a restricted cubic spline model to fit the nonlinear causal relationship between the predicted protein expression level with a correlation coefficient greater than a preset coefficient threshold and the disease risk, obtain proteins with significant nonlinear causal relationship with the disease risk, and realize causal association analysis.
[0010] Furthermore, S4 also includes:
[0011] A generalized linear model is used to perform linear identification on the proteins that have significant nonlinear causal association with disease risk, so as to screen out proteins that cannot be identified by the generalized linear model.
[0012] Furthermore, it also includes: based on the protein that has a significant nonlinear causal relationship with the disease risk, obtaining a dose-response curve of the protein and the disease risk; wherein the dose-response curve is used to provide an optimal drug dosage.
[0013] Furthermore, the dose-response curve includes J-shaped, U-shaped and L-shaped curves; the inflection points of the J-shaped, U-shaped and L-shaped curves correspond to the optimal drug dose.
[0014] Furthermore, the genetic scoring model is:
[0015]
[0016] Where, PGS i is the predicted protein expression level of the i-th sample, i is 1, 2, ..., n, and n is the total number of samples; X ij is the number of effect alleles of the jth SNP in the i-th sample, J is the total number of SNPs corresponding to the current protein; w j is the marginal effect value of the jth SNP;
[0017] The restricted cubic spline model is:
[0018]
[0019] Where Y i is the trait of the disease, i takes 1, 2, ..., N, N is the total number of samples that have passed the verification, and is a subset of the total number of samples n; g() is the connection function; C t () is the cubic spline function corresponding to the t-th window after each predicted protein expression level is truncated into different windows, k is the preset number of truncation nodes; β0, β1 and β t are the regression coefficients of the model.
[0020] Furthermore, the marginal effect value w of the j-th SNP is j Calculated using Olink technology and SomaScan technology.
[0021] Furthermore, in S3, the correlation coefficient is the Pearson correlation coefficient r, and the calculation formula is:
[0022]
[0023] in, is the mean of all predicted protein expression levels that were screened and matched; M i is the actual measured protein expression level of the i-th sample; It's M i The mean of .
[0024] According to a second aspect of the present invention, a nonlinear proteomics causal association analysis system based on genetic scoring is provided, which is used to perform the nonlinear proteomics causal association analysis method based on genetic scoring according to any one of the first aspects, comprising:
[0025] Protein expression level prediction module, used to predict the expression levels of different proteins in each sample using a genetic scoring model;
[0026] A matching module is used to match each predicted protein expression level with the actual measured protein expression level in the database, wherein the matching includes: consistency matching between the sample and the protein name corresponding to the protein expression level; and screening out the predicted protein expression levels that pass the matching;
[0027] A correlation coefficient calculation module is used to calculate the correlation coefficient between the predicted protein expression level screened out through matching and the corresponding actual measurement value;
[0028] The causal association analysis module is used to use a restricted cubic spline model to fit the nonlinear causal association between the predicted protein expression level with a correlation coefficient greater than a preset coefficient threshold and the disease risk, obtain proteins with significant nonlinear causal association with the disease risk, and realize causal association analysis.
[0029] According to a third aspect of the present invention, there is provided an electronic device comprising a computer-readable storage medium and a processor;
[0030] The computer-readable storage medium is used to store executable instructions;
[0031] The processor is configured to read the executable instructions stored in the computer-readable storage medium to execute the nonlinear proteomics causal association analysis method based on genetic scoring according to any one of the first aspects.
[0032] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the nonlinear proteomics causal association analysis method based on genetic scoring as described in any one of the first aspects is implemented.
[0033] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects:
[0034] (1) The nonlinear proteomics causal association analysis method based on genetic scoring in the present invention performs nonlinear proteomics causal association analysis based on the protein expression level predicted by the genetic scoring (PGS) model. The essence of PGS is the weighted sum of the single nucleotide polymorphism sites (SNPs) of the whole genome and the estimated genetic effect size. The present invention takes into account the high cost of measuring protein expression levels. In the prior art, only small samples of protein expression levels are generally used for measurement, which reduces the accuracy of nonlinear proteomics causal association analysis. Based on this, the present invention first uses the PGS model to use the genotype of a larger sample to predict the corresponding protein expression level, and indirectly obtains the predicted protein expression level of the large sample, which can make the predicted protein expression level result obtained more stable, and the predicted protein level itself will not be affected by environmental factors during measurement. Then, the predicted protein expression level is matched with the actual measured value of the protein expression level in the database, and the correlation coefficient between the predicted protein expression level that passes the match and its corresponding actual measured value is calculated. The protein corresponding to the predicted protein expression level with a correlation coefficient greater than the preset coefficient threshold is subjected to subsequent causal association analysis. In this way, the accuracy of the predicted protein expression level used for subsequent causal association analysis can be guaranteed to be high. Finally, a restricted cubic spline model (RCS) is used to fit the nonlinear causal relationship between the predicted protein expression level and the disease risk with a correlation coefficient greater than a preset coefficient threshold. Proteins with significant nonlinear causal associations with the disease risk are obtained, thus realizing nonlinear association analysis. In the present invention, a restricted cubic spline model for fitting the association between macroscopic exposure factors and traits is used to identify the nonlinear causal relationship between microscopic protein expression levels and disease risk, establish a nonlinear causal relationship between microscopic proteins and disease risk, and improve the accuracy of causal association analysis.
[0035] (2) Furthermore, the proteins identified as having significant nonlinear causal associations with disease risk can be identified using a generalized linear model, and proteins that cannot be identified by the generalized linear model can be regarded as newly discovered proteins that are significantly associated with disease risk and given special attention.
[0036] (3) Furthermore, the dose-response curve of protein and disease risk can be obtained based on the RCS model fitting results, and the optimal drug dosage guidance can be provided based on the dose-response curve.
[0037] (4) Furthermore, according to the dose-response curves of different shapes such as J-shaped, U-shaped and L-shaped, the inflection point of the curve corresponds to the optimal dosage, which can provide guidance for the optimal dosage of targeted drugs.
[0038] (5) As a preference, the marginal effect value is calculated by using two relatively mature protein sequencing technologies, Olink technology and SomaScan technology, which can further improve the robustness and accuracy of the prediction results.
[0039] (6) As a preference, the Pearson correlation coefficient is used to screen the predicted protein expression levels for subsequent causal association analysis, which can further ensure the accuracy of the predicted protein expression levels and further improve the accuracy of the causal association analysis.
[0040] In summary, the present invention performs nonlinear association analysis based on protein expression levels predicted by PGS, which not only increases the sample size of the analysis, but also upgrades the association evidence to causal evidence, providing a robust and widely applicable nonlinear proteomics causal association analysis method with high association analysis accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Figure 1 Schematic diagram of the nonlinear proteomics causal association analysis method based on genetic scoring in an embodiment of the present invention.
[0042] Figure 2 Schematic diagram of the process of predicting and verifying different protein expression levels in an embodiment of the present invention.
[0043] Figure 3 This is a causal relationship diagram of two proteins with significant linear and nonlinear trends in the examples of the present invention.
[0044] Figure 4 This is a causal relationship diagram of two proteins in the examples of the present invention, which have only significant nonlinear trends but no significant linear trends.
[0045] Figure 5 The figure shows the Manhattan plot of the linear and nonlinear associations between the predicted protein expression levels and disease risks fitted in the embodiments of the present invention, a schematic diagram of distinguishing the two modes through a Venn diagram, and a schematic diagram of subsequent analysis of nonlinear proteins. DETAILED DESCRIPTION
[0046] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only intended to illustrate the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
[0047] Example 1
[0048] like Figure 1 As shown, the nonlinear proteomics causal association analysis method based on genetic scoring provided in the embodiment of the present invention mainly includes:
[0049] S1. Use the genetic scoring model to predict the expression levels of different proteins in each sample; the genetic scoring model corresponding to each protein is:
[0050]
[0051] Among them, PGS i is the predicted protein expression level of the i-th sample, also known as the genetic score; i takes 1, 2, ..., n, where n is the total number of samples; j takes 1, 2, ..., J, where J is the total number of SNPs corresponding to the current protein; X ij is the number of effect alleles of the jth SNP in the i-th sample; w j is the marginal effect value of the jth SNP, also known as the regression coefficient. It is a fixed parameter of the genetic scoring model and may vary for different measurement techniques. Therefore, to improve the robustness and accuracy of the prediction results, in this embodiment of the present invention, the expression levels of 308 Olink proteins and 2384 SomaScan proteins were predicted. The corresponding protein expression levels were calculated for the marginal effect values obtained using the Olink and SomaScan measurement techniques.
[0052] In the embodiment of the present invention, the sample is the target population, X ij The data comes from the target population database UK Biobank, j From the INTERVAL database. In actual use, the population can be flexibly set according to the actual situation. Different protein expression levels refer to the protein expression levels of different protein names.
[0053] S2, the protein expression level of each prediction is matched with the actual measurement value in the known protein expression level database, including: the consistency matching of the sample corresponding to the protein expression level and the protein name, if the sample and protein name of the protein expression level of the current prediction are consistent with the actual measurement value of the protein expression level in the database, then it is considered that the protein expression level of the current prediction has completed matching, and only the predicted protein expression level that is consistent with the screening sample and the protein name participates in subsequent calculations, and the predicted protein expression level that does not pass the matching is discarded. In the embodiment of the present invention, the verification data as the actual measurement value (i.e., known protein expression level database) are from UK Biobank Pharma Proteomics Project (UKB-PPP) data.
[0054] S3, calculate the correlation coefficient between each predicted protein expression level after matching and its corresponding actual measurement value; only the protein corresponding to the predicted protein expression level whose correlation coefficient is greater than the preset coefficient threshold is subjected to subsequent causal association analysis. Wherein, the preset coefficient threshold is selected according to actual needs, and the smaller the preset coefficient threshold, the closer the predicted protein expression level through matching is to the protein expression level in the database. In an embodiment of the present invention, after matching the target population with the verification data, 15341 overlapping samples, 295 overlapping Olink proteins and 944 overlapping SomaScan proteins are obtained, wherein the overlapping samples are samples consistent with the known protein expression level database in n samples, and the overlapping proteins, i.e., the proteins whose predicted protein expression levels are consistent with the protein names of the actual measurement values of the protein expression levels in the database. Calculate the Pearson correlation coefficient between the protein expression level of each overlapping protein and its corresponding actual measurement value, and the calculation formula is:
[0055]
[0056] Where r is the Pearson correlation coefficient, N is the total number of overlapping samples, and in the embodiment of the present invention, N=15341; is the mean of the protein expression levels predicted by all overlapping samples; M i is the actual measured protein expression level of the i-th sample. In the embodiment of the present invention, M i From UKB-PPP; It's M i Only select r>0.1, that is, the coefficient of determination R 2237 Olink proteins and 409 SomaScan proteins with a correlation coefficient greater than 0.01 were subjected to subsequent correlation analysis. In other embodiments, other correlation analysis methods can also be selected to calculate the correlation coefficient between each predicted protein expression level after verification and its corresponding actual measured value, such as Spearman correlation analysis method.
[0057] S4. Use a restricted cubic spline model (RCS) to fit the nonlinear causal relationship between the predicted protein expression level and disease risk, with a correlation coefficient greater than a preset coefficient threshold, to obtain proteins with significant nonlinear causal association with disease risk, thereby achieving causal association analysis. The calculation formula of the RCS model is as follows:
[0058]
[0059] Where Y i is the trait of the disease, which can be continuous data, discrete data or survival data, i takes 1, 2, ..., N; g() is the link function used to convert the disease trait Y i The mapping is the form in which the right side of the equation is fitted with the predicted protein expression level; C t () is the cubic spline function corresponding to the t-th window after each predicted protein expression level is truncated into different windows, k is the preset number of truncation nodes, and the total number of windows is k+1; β0, β1 and β t are the regression coefficients of the model, which are used to reflect the Y i The nonlinear relationship between the protein expression level and the predicted protein expression level was calculated by the model.
[0060] In the embodiment of the present invention, β0, β1 and β t The solution (β t Including β2, β3, ..., β k-1 ) can be obtained by the following formula:
[0061]
[0062] in,
[0063] Where Q t ,t=1,2,3,…,k are artificially set window nodes. When solving, only the information of the middle 2 to k-1 nodes is needed. In the embodiment of the present invention, 4 nodes are set (i.e., k=4), which are the 5% quantile, 35% quantile, 65% quantile and 95% quantile of the genetic score respectively.
[0064] As a further design of the present invention, S4 further includes:
[0065] The proteins identified to have a significant nonlinear causal association with disease risk are identified using a generalized linear model, and the proteins identified to have a significant nonlinear causal association with disease risk are divided into two modes according to whether they can be identified by the generalized linear model. Mode 1: The proteins identified to have a significant nonlinear causal association with disease risk can be identified by the generalized linear model; Mode 2: The proteins identified to have a significant nonlinear causal association with disease risk cannot be identified by the generalized linear model. Proteins that have a significant nonlinear causal association with disease risk that cannot be identified by the generalized linear model are newly discovered associated proteins and require special attention. In the embodiment of the present invention, being able to be identified by the generalized linear model means satisfying the linear causal association between protein expression level and disease risk in the generalized linear model.
[0066] According to the RCS model fitting results, proteins with significant nonlinear causal association with disease risk and regression coefficients β0, β1 and β t , we can obtain the dose-response curve of protein and disease risk. According to the shape of the curve, proteins with significant nonlinear causal association with disease risk can be divided into J-type, U-type and L-type. The inflection points of different types of proteins correspond to the optimal drug dose. For example, the protein corresponding to the U-shaped curve shows that too low or too high protein expression levels will increase the risk of disease, which can provide guidance for the optimal dosage of targeted drugs. At the same time, prospective cohort verification, biological pathway analysis and drug target analysis can also be performed to further explore the function and drug value of nonlinear proteins.
[0067] The following combination Figure 2-Figure 5 The method of the present invention is further described.
[0068] like Figure 2 The figure shows the process of predicting and validating the expression levels of different proteins. The validated proteins are identified through the restricted cubic spline model (RCS) to obtain proteins with significant nonlinear causal association with disease risk.
[0069] The proteins with significant nonlinear causal association with disease risk were then identified using a generalized linear model. The causal relationship between the two proteins with significant nonlinear causal association with disease risk corresponding to pattern 1 is as follows: Figure 3 As shown in Figure 2, these two proteins with significant nonlinear causal associations with disease risk can also be identified by the generalized linear model, that is, the linear and nonlinear trends between these two proteins and disease risk are significant. The causal relationship between the two proteins with significant nonlinear causal associations with disease risk corresponding to model 2 is shown in Figure 2. Figure 4As shown in the figure, these two proteins with significant nonlinear causal association with disease risk cannot be identified by the generalized linear model, that is, the nonlinear trend between these two proteins and disease risk is significant, but the linear trend is not significant. Figure 3 and Figure 4 The blue curve in the figure represents a nonlinear trend, which is U-shaped or L-shaped; the red curve represents a linear trend. The purple triangle represents the reference point.
[0070] like Figure 5 As shown, it is a Manhattan diagram of the linear and nonlinear associations between the predicted protein expression level and disease risk fitted in an embodiment of the present invention, and a schematic diagram of distinguishing the two modes through a Venn diagram.
[0071] Figure 5 The left side of the figure shows the fitted linear Manhattan plot, and the right side shows the fitted nonlinear Manhattan plot. The horizontal axis represents the position of the sample chromosome, and the vertical axis represents the significance level. Figure 5 The dotted line in the figure represents the set significance level threshold. Figure 5 The points on the left side that are higher than the significance level threshold represent proteins with linear significance, and the purple points on the right side that are higher than the significance level threshold represent proteins with nonlinear significance. Figure 5 In the middle is a Venn diagram distinguishing the two patterns. Figure 5 The figure also shows the subsequent analysis of the identified significant nonlinear proteins, including prospective cohort replication, pathway analysis, and drug target analysis.
[0072] The nonlinear proteomics causal association analysis method based on genetic score in the present invention performs nonlinear proteomics causal association analysis based on the protein expression level predicted by genetic score (PGS) model. The essence of PGS is the weighted sum of whole genome single nucleotide polymorphism sites (SNP) and estimated genetic effect size. The present invention takes into account the high cost of measuring protein expression level. In the prior art, the protein expression level measurement of only a small sample is generally adopted, which reduces the accuracy of nonlinear proteomics causal association analysis. Based on this, the present invention first utilizes the PGS model, adopts the genotype of a larger sample to predict the corresponding protein expression level, indirectly obtains the predicted protein expression level of the large sample, can make the predicted protein expression level result obtained more stable, and the predicted protein level itself will not be affected by environmental factors during measurement. Then the predicted protein expression level is matched with the actual measured value in the known protein expression level database, and the correlation coefficient between the predicted protein expression level and its corresponding actual measured value is calculated, and the protein corresponding to the predicted protein expression level with the correlation coefficient greater than the preset coefficient threshold is subjected to subsequent causal association analysis. In this way, the accuracy of the predicted protein expression level for subsequent causal association analysis can be guaranteed to be higher. Finally, a restricted cubic spline model (RCS) is used to fit the nonlinear causal relationship between the predicted protein expression level and the disease risk with a correlation coefficient greater than a preset coefficient threshold. Proteins with significant nonlinear causal associations with the disease risk are obtained, thus realizing nonlinear association analysis. In the present invention, a restricted cubic spline model for fitting the association between macroscopic exposure factors and traits is used to identify the nonlinear causal relationship between microscopic protein expression levels and disease risk, establish a nonlinear causal relationship between microscopic proteins and disease risk, and improve the accuracy of causal association analysis.
[0073] It should be noted that the above methods are all implemented by computers.
[0074] Example 2
[0075] The embodiment of the present invention provides a nonlinear proteomics causal association system based on genetic scoring, comprising:
[0076] Protein expression level prediction module, used to predict the expression levels of different proteins in each sample using a genetic scoring model;
[0077] A matching module is used to match each predicted protein expression level with the actual measured protein expression level in the database, wherein the matching includes: consistency matching between the sample and protein name corresponding to the protein expression level; and screening out the predicted protein expression levels that pass the matching;
[0078] A correlation coefficient calculation module, used to calculate the correlation coefficient between the predicted protein expression level that has passed the matching and its corresponding actual measurement value;
[0079] The causal association analysis module is used to use a restricted cubic spline model to fit the nonlinear causal association between the predicted protein expression level and disease risk with a correlation coefficient greater than a preset coefficient threshold, obtain proteins with significant nonlinear causal association with disease risk, and realize causal association analysis.
[0080] Each of the above modules is used to execute the corresponding steps of the nonlinear proteomics causal association analysis method based on genetic scoring in Example 1. For details, please refer to the corresponding description in Example 1 and will not be repeated here.
[0081] Example 3
[0082] An embodiment of the present invention provides an electronic device, including a computer-readable storage medium and a processor;
[0083] Computer-readable storage media for storing executable instructions;
[0084] The processor is configured to read the executable instructions stored in the computer-readable storage medium to execute the nonlinear proteomics causal association analysis method based on genetic scoring in Example 1. For details, please refer to the corresponding description in Example 1 and will not be repeated here.
[0085] Example 4
[0086] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the nonlinear proteomics causal association analysis method based on genetic scoring as described in Example 1 is implemented. For details, please refer to the corresponding description in Example 1 and will not be repeated here.
[0087] The present invention uses a genetic scoring model to predict the expression levels of different proteins in each sample; each predicted protein expression level is verified with the actual measured value in the database, including consistency matching of the sample and protein name corresponding to the protein expression level and calculation of the correlation coefficient between the matched predicted protein expression level and the actual measured value corresponding to it; a restricted cubic spline model is used to fit the nonlinear causal association between the predicted protein expression level with a correlation coefficient greater than a preset coefficient threshold and the disease risk, thereby obtaining proteins with a significant nonlinear causal association with the disease risk, and realizing causal association analysis. The present invention performs nonlinear association analysis based on the protein expression level predicted by the genetic score, which not only increases the sample size of the analysis, but also upgrades the association evidence to causal evidence. It is a robust and widely applicable nonlinear proteomics causal association analysis method.
[0088] It will be easily understood by those skilled in the art that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A nonlinear proteomics causal association analysis method based on genetic scoring, characterized in that: include: S1. Use the genetic scoring model to predict the expression levels of different proteins in each sample; S2. Matching each predicted protein expression level with the actual measured protein expression level in the database, wherein the matching includes: consistency matching between the sample and the protein name corresponding to the protein expression level; and screening out the predicted protein expression levels that pass the matching; S3, calculating the correlation coefficient between the predicted protein expression level screened out by matching and the corresponding actual measured value; S4. Using a restricted cubic spline model to fit the nonlinear causal relationship between the predicted protein expression level and the disease risk, whose correlation coefficient is greater than a preset coefficient threshold, to obtain proteins with significant nonlinear causal relationship with the disease risk, thereby achieving causal association analysis; The genetic scoring model is: Where, It is the predicted The protein expression levels of the samples Pick , is the total number of samples; It is Sample No. j The number of effect alleles for each SNP, J is the total number of SNPs corresponding to the current protein; It is j The marginal effect value of each SNP; The restricted cubic spline model is: Where, It is a symptom of the disease, Take 1, 2, ..., N, where N is the total number of samples that have passed the verification and N is the total number of samples A subset of is the link function; After each predicted protein expression level is truncated into different windows, The cubic spline function corresponding to the window, is the preset number of truncation nodes; 、 and are the regression coefficients of the model.
2. The nonlinear proteomics causal association analysis method based on genetic scoring according to claim 1, characterized in that: S4 also includes: A generalized linear model is used to perform linear identification on the proteins that have significant nonlinear causal association with disease risk, so as to screen out proteins that cannot be identified by the generalized linear model.
3. The nonlinear proteomics causal association analysis method based on genetic scoring according to claim 1 or 2, characterized in that: Also includes: Based on the protein having a significant nonlinear causal relationship with disease risk, a dose-response curve of the protein and disease risk is obtained; wherein the dose-response curve is used to provide an optimal drug dosage.
4. The nonlinear proteomics causal association analysis method based on genetic scoring according to claim 3, characterized in that: The dose-response curve includes J-shaped, U-shaped and L-shaped curves; the inflection points of the J-shaped, U-shaped and L-shaped curves correspond to the optimal drug dose.
5. The nonlinear proteomics causal association analysis method based on genetic scoring according to claim 1, characterized in that: The said j Marginal effect value of each SNP Calculated using Olink and SomaScan technologies.
6. The nonlinear proteomics causal association analysis method based on genetic scoring according to claim 1, characterized in that: In S3, the correlation coefficient is the Pearson correlation coefficient , the calculation formula is: in, is the mean of all predicted protein expression levels that were filtered and matched; It is The actual measured protein expression level of each sample; yes The mean of .
7. A nonlinear proteomics causal association analysis system based on genetic scoring, characterized in that: The method for performing the nonlinear proteomics causal association analysis based on genetic scoring according to any one of claims 1 to 6 comprises: Protein expression level prediction module, used to predict the expression levels of different proteins in each sample using a genetic scoring model; A matching module is used to match each predicted protein expression level with the actual measured protein expression level in the database, wherein the matching includes: consistency matching between the sample and the protein name corresponding to the protein expression level; and screening out the predicted protein expression levels that pass the matching; A correlation coefficient calculation module is used to calculate the correlation coefficient between the predicted protein expression level screened out through matching and the corresponding actual measurement value; The causal association analysis module is used to use a restricted cubic spline model to fit the nonlinear causal association between the predicted protein expression level with a correlation coefficient greater than a preset coefficient threshold and the disease risk, obtain proteins with significant nonlinear causal association with the disease risk, and realize causal association analysis.
8. An electronic device, characterized in that: comprising a computer-readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is configured to read the executable instructions stored in the computer-readable storage medium to execute the nonlinear proteomics causal association analysis method based on genetic scoring according to any one of claims 1 to 6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the nonlinear proteomics causal association analysis method based on genetic scoring is implemented.
