Multi-source biomarker data analysis method and system

Abstract

The invention discloses a multi-source biomarker data analysis method and system, and relates to the technical field of data analysis, the method comprises the following steps: obtaining multi-source biomarker data corresponding to a biomarker to be analyzed, the multi-source biomarker data comprising local measurement data and reference measurement data of a biological sample; constructing a correlation expression model of the biomarker to be analyzed and the target disease according to the local measurement data and the reference measurement data; and solving the correlation expression model based on a Bayesian mode to obtain a correlation analysis result of the to-be-analyzed biomarker and the target disease. According to the method, the model for describing the association relationship of the local measurement data, the reference measurement data and the target disease is constructed, and the model is solved on the basis of the Bayesian mode, so that the analysis result of the association between the biomarker and the disease can be accurately estimated without depending on complex hypotheses, and the accuracy of the biomarker and the disease is improved. Compared with an existing mode, the method is more suitable for integrating and analyzing the multi-source biomarkers in a clinical test scene.

Description

Technical Field

[0001] The present application relates to the field of data analysis technology, and in particular to a method and system for analyzing multi-source biomarker data. Background Art

[0002] Biomarkers are defined as measurable indicators of health status, disease presence, or intervention effects and play a key role in clinical research. The study of biomarkers can assist clinical trials, and the correlation between biomarkers and clinical outcomes is crucial in drug development and regulatory decision-making.

[0003] In the existing methods, biomarker research is generally conducted by quantifying the association between biomarkers and diseases. In order to improve the statistical power and accuracy of the research, biomarker data from different sources can be integrated for association analysis. However, due to differences in measurement methods and laboratory settings, biomarker data from different sources vary greatly. The existing method of integrating multi-source biomarker data relies on complex assumptions, which limits its applicability and makes it difficult to meet the analysis needs in different scenarios.

[0004] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0005] The main purpose of this application is to provide an analysis method for multi-source biomarker data, aiming to solve the technical problem that the existing method of integrating multi-source biomarker data relies on complex assumptions and has limited applicable scenarios.

[0006] To achieve the above objectives, the present application proposes a method for analyzing multi-source biomarker data, the method comprising: Acquiring multi-source biomarker data corresponding to the biomarker to be analyzed, wherein the multi-source biomarker data includes local measurement data and reference measurement data of the biological sample; constructing an associated expression model of the biomarker to be analyzed and the target disease according to the local measurement data and the reference measurement data; The association expression model is solved based on a Bayesian approach to obtain an association analysis result between the biomarker to be analyzed and the target disease.

[0007] In one embodiment, the multi-source biomarker data also includes the disease output results corresponding to each of the biological samples, the local measurement data includes the local measurement values ​​of all biological samples, the reference measurement data includes the reference measurement values ​​of several biological samples, and the step of constructing the association expression model between the biomarker to be analyzed and the target disease according to the local measurement data and the reference measurement data includes: Constructing a conditionally independent association expression based on the local measurement value, the reference measurement value, and the disease output result in combination with a covariate vector; Among them, in the association expression, the joint probability of the disease output result and the reference measurement value satisfies the conditional independence principle.

[0008] In one embodiment, the step of solving the association expression model based on a Bayesian approach to obtain the association analysis result between the biomarker to be analyzed and the target disease includes: constructing a likelihood function representation of the biomarker to be analyzed based on all the biological samples; Determine the prior distribution function of the target disease association based on the normal distribution assumption; Determine a posterior distribution function of the target disease association degree according to the likelihood function representation and the prior distribution function; The posterior distribution function is sampled, and the association analysis result between the biomarker to be analyzed and the target disease is obtained according to the sampling result.

[0009] In one embodiment, before the step of constructing the likelihood function representation of the biomarkers to be analyzed based on the entire biological sample, the step includes: Among all biological samples, a biological sample corresponding to both a local measurement value and a reference measurement value is determined as a first biological sample, and other biological samples are determined as a second biological sample; Accordingly, the step of constructing a likelihood function representation of the biomarkers to be analyzed based on all biological samples includes: Initializing a first likelihood function according to the first biological sample and the second biological sample; Decomposing the first likelihood function representation according to the conditionally independent association expression to obtain a second likelihood function representation, wherein the second likelihood function representation includes: a reference-disease association representation, a reference-local association representation, and a reference prior distribution representation; The reference-disease association representation, the reference-local association representation and the reference prior distribution representation are solved respectively, and the likelihood function representation of the biomarker to be analyzed is determined according to the solution results.

[0010] In one embodiment, the step of respectively solving the reference-disease association representation, the reference-local association representation, and the reference prior distribution representation, and determining the likelihood function representation of the biomarker to be analyzed according to the solution results, comprises: Determining the reference-disease association representation according to the reference measurement value of the first biological sample and the corresponding disease output result based on a Logistic regression model; determining the reference-local association representation based on the local measurement value of the first biological sample and the reference measurement value based on a normal distribution assumption; A reference measurement value of the first biological sample is determined as the reference prior distribution representation.

[0011] In one embodiment, before the step of determining the prior distribution function of the target disease association degree based on the normal distribution assumption, the following steps are included: Determine several components of the target disease association degree according to the reference-disease association representation, the reference-local association representation, and the reference prior distribution representation; Divide each of the constituent elements into conjugate prior elements, information prior elements and sample parameter elements; Accordingly, the step of determining the prior distribution function of the target disease association degree based on the normal distribution assumption includes: Based on the distribution consistency principle of the reference measurement value, determining the corresponding prior distribution representation according to the conjugate prior class element; Based on the principle of standard normal distribution, the corresponding prior distribution representation is determined according to the information prior class elements; Based on a preset distribution principle, determining a corresponding priori representation according to the sample parameter class element; The prior distribution function of the target disease association degree is determined according to the prior representations corresponding to the conjugate prior class elements, the information prior class elements and the sample parameter class elements respectively.

[0012] In one embodiment, the sample parameter class element includes a first sample parameter class element and a second sample parameter class element; the step of determining the corresponding prior representation according to the sample parameter class element based on a preset distribution principle includes: Based on the principle of overall normal distribution, determining the corresponding prior representation according to the first sample parameter class element; Based on the inverse gamma distribution principle, the corresponding prior representation is determined according to the second sample parameter class element.

[0013] In addition, to achieve the above objectives, the present application also proposes an analysis system for multi-source biomarker data, the system comprising: A data acquisition module, used to acquire multi-source biomarker data corresponding to the biomarker to be analyzed, wherein the multi-source biomarker data includes local measurement data and reference measurement data of the biological sample; An associated expression module, used to construct an associated expression model between the biomarker to be analyzed and the target disease according to the local measurement data and the reference measurement data; The association analysis module is used to solve the association expression model based on the Bayesian method to obtain the association analysis results between the biomarker to be analyzed and the target disease.

[0014] In one embodiment, the multi-source biomarker data further includes a disease output result corresponding to each of the biological samples, the local measurement data includes local measurement values ​​of all biological samples, and the reference measurement data includes reference measurement values ​​of several biological samples; The association expression module is further used to construct a conditionally independent association expression based on the local measurement value, the reference measurement value and the disease output result, and in combination with a covariate vector; Among them, in the association expression, the joint probability of the disease output result and the reference measurement value satisfies the conditional independence principle.

[0015] In one embodiment, the association analysis module is further used to construct a likelihood function representation of the biomarker to be analyzed based on all biological samples; Determine the prior distribution function of the target disease association based on the normal distribution assumption; Determine a posterior distribution function of the target disease association degree according to the likelihood function representation and the prior distribution function; The posterior distribution function is sampled, and the association analysis result between the biomarker to be analyzed and the target disease is obtained according to the sampling result.

[0016] The present application provides an analysis method for multi-source biomarker data, by obtaining multi-source biomarker data corresponding to the biomarker to be analyzed, the multi-source biomarker data including local measurement data and reference measurement data of the biological sample; constructing an association expression model between the biomarker to be analyzed and the target disease based on the local measurement data and the reference measurement data; solving the association expression model based on the Bayesian method to obtain the association analysis result between the biomarker to be analyzed and the target disease. Since a model describing the mutual correlation between local measurement data, reference measurement data and target disease is constructed in the present application, and the model is solved based on the Bayesian method, the analysis result of the association between biomarkers and diseases can be accurately estimated without relying on complex assumptions. Compared with the existing methods, it is more suitable for the integrated analysis of multi-source biomarkers in clinical trial scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0019] Figure 1 This is a flow chart of the first embodiment of the method for analyzing multi-source biomarker data of the present application; Figure 2 This is a flow chart of a second embodiment of a method for analyzing multi-source biomarker data of the present application; Figure 3 This is a flow chart of a third embodiment of a method for analyzing multi-source biomarker data of the present application; Figure 4 This is a schematic diagram of the module structure of the first embodiment of the multi-source biomarker data analysis system of the present application.

[0020] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0021] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0022] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0023] The present application embodiment provides a method for analyzing multi-source biomarker data, referring to Figure 1 , Figure 1 This is a flow chart of a first embodiment of a method for analyzing multi-source biomarker data of the present application. In this embodiment, the method includes: Steps S10 to S30: Step S10: Acquire multi-source biomarker data corresponding to the biomarker to be analyzed, wherein the multi-source biomarker data includes local measurement data and reference measurement data of the biological sample.

[0024] It should be noted that the method of this embodiment can be applied in the scenario of standardizing biomarker data from different sources. The execution subject of the method of this embodiment can be a computing service device with data processing, model calling, data storage and program running functions, such as a mobile phone, a personal computer, a biological data analyzer, etc., and can also be other electronic devices that can achieve the same or similar functions. This embodiment does not limit this. Here, an analysis device for multi-source biomarker data (hereinafter referred to as the analysis device) is used as an example to illustrate this embodiment and the following embodiments.

[0025] It should be understood that the biomarker to be analyzed is a biomarker determined based on the current research purpose and needs to be analyzed for its association with a target disease. The corresponding data of the biomarker to be analyzed can be obtained from multiple contributing clinical research sources.

[0026] It is understandable that due to differences in laboratory settings at different clinical research sources, and differences in the way the same biomarker is measured in different laboratories, in order to reduce the impact of the inherent differences between clinical research sources on subsequent analysis data, after obtaining local measurement data of biological samples from different sources, the biological samples can also be re-measured in a designated reference laboratory to obtain reference measurement data of the biological samples.

[0027] It should be understood that the practicality of reanalyzing all biological samples from different sources from a reference laboratory is hampered by logistical constraints (time, manpower, and cost), and therefore usually only a portion of the biological samples are selected for reanalysis, that is, the local measurement data includes the local measurement values ​​of all biological samples, and the reference measurement data includes the reference measurement values ​​of several biological samples.

[0028] Step S20: constructing an association expression model between the biomarker to be analyzed and the target disease according to the local measurement data and the reference measurement data.

[0029] It is understandable that the target disease can be a disease determined based on the current research purpose that needs to be analyzed for its association with the above-mentioned biological markers to be analyzed, or other clinical symptoms, etc. The multi-source biomarker data also includes the disease output results corresponding to each of the biological samples, and then a correlation expression model describing the relationship between the reference measurement, the local measurement and the result can be constructed.

[0030] It should be understood that the analysis device can construct a conditionally independent association expression based on the local measurement value, the reference measurement value and the disease output result, in combination with the covariate vector; wherein, in the association expression, the joint probability of the disease output result and the reference measurement value satisfies the conditional independence principle.

[0031] Specifically, s=1,2,...,S represents clinical studies from different sources. For individual i in study s, represents the binary disease output, represents the reference measurement value of the biological sample, represents the local measurement value of the biological sample, A vector of other covariates that affect the results of the association analysis.

[0032] For all biological samples, local measurements of biomarkers Available, but only a subset of biological samples have reference measurements Therefore, for individual i in study s, we can indicates that the reference measurement value is available, otherwise it is .

[0033] Assuming reference measurement value The distribution of is consistent across clinical studies, while local measurements The distribution of is heterogeneous across clinical studies. In addition, it can be assumed that In the case of disease output and local measurements are conditionally independent; assuming that the covariates There is no effect on the measurement bias of a particular study, i.e. The conditionally independent association expression is:

[0034] It is understandable that, since there are biological samples without reference measurement values ​​(not reanalyzed), the reference measurement values ​​corresponding to the biological samples that have not been reanalyzed can be regarded as unknown potential variables. Further, for ease of representation, it can be defined as:

[0035]

[0036]

[0037]

[0038]

[0039] in, is the total number of biological samples in the study, where is an unknown potential variable.

[0040] Therefore, the above conditionally independent association expression can be converted into the following association expression model:

[0041] Understandably, Here, it can be regarded as a parameter representation of the degree of association between the biomarker to be analyzed and the target disease, that is, the target disease association degree.

[0042] Step S30: solving the association expression model based on a Bayesian approach to obtain an association analysis result between the biomarker to be analyzed and the target disease.

[0043] It is understandable that the Bayesian approach can be used to infer unknown parameters based on prior knowledge, as shown by the Bayesian formula:

[0044] in, Expressed as a proportion, is the unnormalized posterior probability, which means that given the observed data (i.e., known disease output y, local measurement value W, and some reference measurement values and the covariate vector ), the unknown parameter (the degree of association between the biomarker to be analyzed and the target disease and the corresponding reference measurements of the biological samples that were not reanalyzed ) probability; is the likelihood probability, which indicates the probability of observing the above observation data assuming that the unknown parameter is true; is the prior probability, which represents the belief about the unknown parameters before observing the data.

[0045] It should be understood that the process of solving the above-mentioned association expression model can be the process of solving a specific function expression. Therefore, step S30 specifically includes: steps S301 to S304: Step S301: constructing a likelihood function representation of the biomarker to be analyzed based on all the biological samples.

[0046] It can be understood that among all biological samples, the biological sample corresponding to both the local measurement value and the reference measurement value is determined as the first biological sample (i.e. The corresponding biological sample, and the other biological sample is identified as the second biological sample (i.e. corresponding biological samples).

[0047] It should be understood that the likelihood function can be expressed as:

[0048] Step S302: Determine the prior distribution function of the target disease association degree based on the normal distribution assumption.

[0049] It should be understood that since the above likelihood function involves unknown , it can therefore be considered a latent variable that can be transformed into an estimable quantity by constructing an appropriate posterior distribution. The distribution of is consistent in various clinical studies, namely and It can be the same distribution function, so the prior distribution function of the target disease association can be determined based on the assumption of normal distribution .

[0050] Step S303: Determine the posterior distribution function of the target disease association degree according to the likelihood function representation and the prior distribution function.

[0051] Step S304: sampling the posterior distribution function, and obtaining the association analysis result between the biomarker to be analyzed and the target disease according to the sampling result.

[0052] It should be understood that since the above-mentioned posterior distribution function is proportional to the above-mentioned likelihood function representation and the prior distribution function, the posterior distribution function is difficult to calculate directly. Therefore, a Markov Chain Monte Carlo (MCMC) method, such as NUTS (No-U-Turn Sampler), can be used to approximate the posterior distribution function.

[0053] Specifically, based on the MCMC sampler, a large number of corresponding samples can be extracted from the above posterior distribution function, and these samples can reflect the shape of the above posterior distribution function; the mean of the MCMC samples is further calculated as a point estimate of the parameter, and the 95% highest posterior density interval is calculated as an interval estimate; and then the posterior distribution function of the target disease association is interpreted according to the estimated value, including the point estimate, interval estimate and the correlation between parameters.

[0054] It can be understood that by using the MCMC method, sampling estimation can be performed from a complex posterior distribution, and then the degree of association between the biomarker to be analyzed and the target disease can be obtained based on the interval estimation value, that is, the final association analysis result can be obtained.

[0055] This embodiment obtains multi-source biomarker data corresponding to the biomarker to be analyzed, and the multi-source biomarker data includes local measurement data and reference measurement data of the biological sample; constructs an association expression model between the biomarker to be analyzed and the target disease based on the local measurement data and the reference measurement data; solves the association expression model based on the Bayesian method to obtain the association analysis result between the biomarker to be analyzed and the target disease. Since a model describing the mutual correlation between local measurement data, reference measurement data and target disease is constructed in this embodiment, and the model is solved based on the Bayesian method, the analysis result of the association between biomarkers and diseases can be accurately estimated without relying on complex assumptions. Compared with the existing method, it is more suitable for the integrated analysis of multi-source biomarkers in clinical trial scenarios.

[0056] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction, and will not be repeated in the following. Figure 2 , Figure 2 This is a flow chart of the second embodiment of the method for analyzing multi-source biomarker data of the present application.

[0057] In this embodiment, the step S301 specifically includes steps S3011 to S3013: Step S3011: Initialize a first likelihood function according to the first biological sample and the second biological sample.

[0058] It can be understood that the first likelihood function can be initialized as:

[0059] Step S3012: Decompose the first likelihood function representation according to the conditionally independent association expression to obtain a second likelihood function representation, wherein the second likelihood function representation includes: a reference-disease association representation, a reference-local association representation, and a reference prior distribution representation.

[0060] It can be understood that the above conditional independence assumption is further invoked, that is, based on the above first likelihood function expression, the above conditional independence association expression is combined , we can get the following second likelihood function expression:

[0061] It can be understood that the above second likelihood function representation consists of three components, namely the reference-disease association representation , Reference-Local Association Representation And the reference prior distribution is expressed as .

[0062] Step S3013: respectively solving the reference-disease association representation, the reference-local association representation and the reference prior distribution representation, and determining the likelihood function representation of the biomarker to be analyzed according to the solution results.

[0063] It can be understood that the reference-disease association representation can be determined based on the Logistic regression model according to the reference measurement value of the first biological sample and the corresponding disease output result; based on the normal distribution assumption, the reference-local association representation can be determined according to the local measurement value of the first biological sample and the reference measurement value; the reference measurement value of the first biological sample is determined as the reference prior distribution representation.

[0064] Specifically, we can first use a logistic regression model with a random intercept term to describe the reference-disease association representation, then we can Converted to the following formula:

[0065] in, To study a specific intercept, is the covariate parameter, is the inverse of the logit function. is the logarithm of the odds ratio (OR), which is used to describe the relationship between the biomarker and the target disease.

[0066] It should be understood that the reference-local correlation representation can be viewed as a calibration model, assuming that the reference measurement and local measurements There is a linear correlation between and Both are normally distributed and follow the following normal distribution:

[0067] as well as

[0068] It can be understood that, for the convenience of description, the embodiments adopt Represents the parameter elements obtained by the above conversion process. , , , , , , ,as well as can be regarded as constituent elements.

[0069] It should also be noted that the above reference prior distribution can be directly obtained from the reference measurement value of the first biological sample To express.

[0070] It should be understood that after respectively determining the above-mentioned reference-disease association representation, reference-local association representation and reference prior distribution representation, the likelihood function representation of the complete biomarker to be analyzed can be obtained.

[0071] In this embodiment, the first likelihood function is initialized according to the first biological sample and the second biological sample; the first likelihood function representation is decomposed according to the conditionally independent association expression to obtain a second likelihood function representation, which includes: a reference-disease association representation, a reference-local association representation, and a reference prior distribution representation; the reference-disease association representation, the reference-local association representation, and the reference prior distribution representation are solved respectively, and the likelihood function representation of the biomarker to be analyzed is determined according to the solution results. Since the likelihood function representation of the biomarker to be analyzed is determined based on the assumption of conditional independence in this embodiment, the maximum likelihood estimate of the parameter can be obtained by maximizing the likelihood function so that it can be combined with the prior distribution to calculate the posterior distribution and obtain the final analysis result, which is conducive to improving the accuracy of the analysis result.

[0072] Based on the first and second embodiments of the present application, in the third embodiment of the present application, the same or similar contents as those of the first and second embodiments can be referred to the above description, and will not be described in detail later. Figure 3 , Figure 3 This is a flowchart of the third embodiment of the method for analyzing multi-source biomarker data of the present application.

[0073] In this embodiment, before the step S302, the following steps are also included: step S001 to step S002: Step S001: Determine several components of the target disease association degree according to the reference-disease association representation, the reference-local association representation and the reference prior distribution representation.

[0074] Step S002: Divide the constituent elements into conjugate prior elements, information prior elements and sample parameter elements.

[0075] It should be understood that the constituent elements may be: , , , , , , ,as well as Therefore, based on the differences of the components, they can be divided into conjugate prior elements, information prior elements and sample parameter elements. Conjugate prior elements include: ,as well as ; Information prior elements include as well as ; Sample parameter class elements include: , , , .

[0076] It should also be noted that since the likelihood function representation of the above-determined biomarkers to be analyzed involves unknown , it can therefore be considered as a latent variable, which can be converted into an estimable quantity in a Bayesian way by constructing an appropriate posterior distribution. Similarly, the hierarchical structure can be reflected in the prior distribution function of the target disease association, for example: Etc., thereby improving the efficiency of obtaining subsequent posterior distribution estimates.

[0077] Accordingly, the step S302 specifically includes: step S3021 to step S3024: Step S3021: Based on the distribution consistency principle of the reference measurement value, determine the corresponding prior distribution representation according to the conjugate prior class element.

[0078] Understandably, assuming and The joint prior distribution of is decomposed and independent of , that is, the prior distribution of the target disease association can be expressed as:

[0079] It is understandable that since the reference measurement value is assumed The distribution of is consistent in various clinical studies, namely and can be the same distribution function. Based on the assumption of normal distribution, It is expressed as:

[0080] In the formula, and It can be set according to different reference measurement values ​​in actual scenarios, or conjugate priors (normal distribution and inverse gamma distribution) can be used without loss of generality. In this embodiment, in order to avoid excessive use of prior information to affect the results, a non-informative prior can be set:

[0081] Step S3022: Based on the standard normal distribution principle, determine the corresponding prior distribution representation according to the information prior class elements.

[0082] It should be understood that and , informative or weakly informative priors can be used to maintain element weights and avoid severe instability. For example, using the standard normal distribution:

[0083] in, ,and is the number of covariates.

[0084] Step S3023: Based on a preset distribution principle, determine a corresponding priori representation according to the sample parameter class element.

[0085] It should be noted that the sample parameter class element ( , , , ) are parameters specific to different clinical studies. A prior representation of the sample parameter class elements can be determined using the simple assumption that the parameters for each study are independent samples from a hyper-population distribution governed by some unknown hyper-parameters.

[0086] It can be understood that the sample parameter type element includes a first sample parameter type element and a second sample parameter type element, and the first sample parameter type element includes: , and ; The second sample parameter class elements include: .

[0087] It should be understood that the corresponding prior representation can be determined according to the first sample parameter class element based on the principle of overall normal distribution; and the corresponding prior representation can be determined according to the second sample parameter class element based on the principle of inverse gamma distribution.

[0088] Specifically, for , and , the overall distribution is normal distribution, expressed as follows:

[0089]

[0090]

[0091] And for , the inverse gamma distribution can be used as follows:

[0092] in , , , , , , ,and is the corresponding hyperparameter, which can also be regarded as an unknown parameter part of. express It follows the inverse gamma distribution, as follows:

[0093] and and Similar to the prior distribution of , an appropriate weak information prior can be assigned to the above hyperparameters. , and , the standard normal distribution can be used, as shown below:

[0094] for , , , ,and , we can use the semi-Cauchy distribution as follows:

[0095] Step S3024: Determine the prior distribution function of the target disease association degree according to the prior representations corresponding to the conjugate prior class elements, the information prior class elements and the sample parameter class elements respectively.

[0096] It should be understood that after obtaining the prior distribution representations corresponding to different components respectively, the prior distribution function of the complete target disease correlation can be obtained.

[0097] In this embodiment, based on the distribution consistency principle of the reference measurement value, the corresponding prior distribution representation is determined according to the conjugate prior class element; based on the standard normal distribution principle, the corresponding prior distribution representation is determined according to the information prior class element; based on the preset distribution principle, the corresponding prior representation is determined according to the sample parameter class element; the prior distribution function of the target disease association is determined according to the prior representations corresponding to the conjugate prior class element, the information prior class element and the sample parameter class element. In this embodiment, by constructing an appropriate prior distribution, the reference measurement value corresponding to the biological sample that has not been re-analyzed can be converted into an estimable quantity based on the Bayesian method when the posterior distribution is obtained, so that the posterior distribution of the association between the biomarker and the target disease can be accurately estimated. The method of this embodiment can provide a more flexible way to use known multi-source biomarker data to obtain integrated analysis results when it is impractical to re-analyze biological specimens.

[0098] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the analysis method of multi-source biomarker data of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0099] The present application also provides a multi-source biomarker data analysis system, please refer to Figure 4 , Figure 4 This is a schematic diagram of the module structure of the first embodiment of the analysis system for multi-source biomarker data of the present application, the system comprising: The data acquisition module 401 is used to acquire multi-source biomarker data corresponding to the biomarker to be analyzed, wherein the multi-source biomarker data includes local measurement data and reference measurement data of the biological sample.

[0100] The association expression module 402 is used to construct an association expression model between the biomarker to be analyzed and the target disease according to the local measurement data and the reference measurement data.

[0101] The association analysis module 403 is used to solve the association expression model based on the Bayesian method to obtain the association analysis result between the biomarker to be analyzed and the target disease.

[0102] Further, the multi-source biomarker data also includes the disease output results corresponding to each of the biological samples, the local measurement data includes the local measurement values ​​of all biological samples, and the reference measurement data includes the reference measurement values ​​of several biological samples; The association expression module 402 is further used to construct a conditionally independent association expression based on the local measurement value, the reference measurement value and the disease output result, and in combination with a covariate vector; Among them, in the association expression, the joint probability of the disease output result and the reference measurement value satisfies the conditional independence principle.

[0103] Furthermore, the association analysis module 403 is also used to construct a likelihood function representation of the biomarker to be analyzed based on all biological samples; Determine the prior distribution function of the target disease association based on the normal distribution assumption; Determine a posterior distribution function of the target disease association degree according to the likelihood function representation and the prior distribution function; The posterior distribution function is sampled, and the association analysis result between the biomarker to be analyzed and the target disease is obtained according to the sampling result.

[0104] Furthermore, the association analysis module 403 is further used to determine, among all biological samples, a biological sample that corresponds to both a local measurement value and a reference measurement value as a first biological sample, and to determine other biological samples as a second biological sample; Initializing a first likelihood function according to the first biological sample and the second biological sample; Decomposing the first likelihood function representation according to the conditionally independent association expression to obtain a second likelihood function representation, wherein the second likelihood function representation includes: a reference-disease association representation, a reference-local association representation, and a reference prior distribution representation; The reference-disease association representation, the reference-local association representation and the reference prior distribution representation are solved respectively, and the likelihood function representation of the biomarker to be analyzed is determined according to the solution results.

[0105] Furthermore, the association analysis module 403 is further configured to determine the reference-disease association representation according to the reference measurement value of the first biological sample and the corresponding disease output result based on a Logistic regression model; determining the reference-local association representation based on the local measurement value of the first biological sample and the reference measurement value based on a normal distribution assumption; A reference measurement value of the first biological sample is determined as the reference prior distribution representation.

[0106] Furthermore, the association analysis module 403 is further used to determine several components of the target disease association degree according to the reference-disease association representation, the reference-local association representation and the reference prior distribution representation; Divide each of the constituent elements into conjugate prior elements, information prior elements and sample parameter elements; Based on the distribution consistency principle of the reference measurement value, determining the corresponding prior distribution representation according to the conjugate prior class element; Based on the principle of standard normal distribution, the corresponding prior distribution representation is determined according to the information prior class elements; Based on a preset distribution principle, determining a corresponding priori representation according to the sample parameter class element; The prior distribution function of the target disease association degree is determined according to the prior representations corresponding to the conjugate prior class elements, the information prior class elements and the sample parameter class elements respectively.

[0107] Further, the sample parameter type element includes a first sample parameter type element and a second sample parameter type element; Furthermore, the association analysis module 403 is further used to determine the corresponding prior representation according to the first sample parameter class element based on the overall normal distribution principle; Based on the inverse gamma distribution principle, the corresponding prior representation is determined according to the second sample parameter class element.

[0108] This embodiment obtains multi-source biomarker data corresponding to the biomarker to be analyzed, and the multi-source biomarker data includes local measurement data and reference measurement data of the biological sample; constructs an association expression model between the biomarker to be analyzed and the target disease based on the local measurement data and the reference measurement data; solves the association expression model based on the Bayesian method to obtain the association analysis result between the biomarker to be analyzed and the target disease. Since a model describing the mutual correlation between local measurement data, reference measurement data and target disease is constructed in this embodiment, and the model is solved based on the Bayesian method, the analysis result of the association between biomarkers and diseases can be accurately estimated without relying on complex assumptions. Compared with the existing method, it is more suitable for the integrated analysis of multi-source biomarkers in clinical trial scenarios.

[0109] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0110] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. Through the description of the above implementation methods, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course, by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as a read-only memory / random access memory, a disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0111] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for analyzing multi-source biomarker data, characterized in that: The method comprises: Acquiring multi-source biomarker data corresponding to the biomarker to be analyzed, wherein the multi-source biomarker data includes local measurement data and reference measurement data of the biological sample; constructing an associated expression model of the biomarker to be analyzed and the target disease according to the local measurement data and the reference measurement data; The association expression model is solved based on a Bayesian approach to obtain an association analysis result between the biomarker to be analyzed and the target disease.

2. The method according to claim 1, characterized in that The multi-source biomarker data also includes the disease output results corresponding to each of the biological samples, the local measurement data includes the local measurement values ​​of all biological samples, and the reference measurement data includes the reference measurement values ​​of several biological samples. The step of constructing the correlation expression model between the biomarker to be analyzed and the target disease according to the local measurement data and the reference measurement data includes: Constructing a conditionally independent association expression based on the local measurement value, the reference measurement value, and the disease output result in combination with a covariate vector; Among them, in the association expression, the joint probability of the disease output result and the reference measurement value satisfies the conditional independence principle.

3. The method according to claim 2, characterized in that The step of solving the association expression model based on a Bayesian approach to obtain the association analysis result between the biomarker to be analyzed and the target disease comprises: constructing a likelihood function representation of the biomarker to be analyzed based on all the biological samples; Determine the prior distribution function of the target disease association based on the normal distribution assumption; Determine a posterior distribution function of the target disease association degree according to the likelihood function representation and the prior distribution function; The posterior distribution function is sampled, and the association analysis result between the biomarker to be analyzed and the target disease is obtained according to the sampling result.

4. The method according to claim 3, characterized in that Before the step of constructing the likelihood function representation of the biomarkers to be analyzed based on all the biological samples, the method includes: Among all biological samples, a biological sample corresponding to both a local measurement value and a reference measurement value is determined as a first biological sample, and other biological samples are determined as a second biological sample; Accordingly, the step of constructing a likelihood function representation of the biomarkers to be analyzed based on all biological samples includes: Initializing a first likelihood function according to the first biological sample and the second biological sample; Decomposing the first likelihood function representation according to the conditionally independent association expression to obtain a second likelihood function representation, wherein the second likelihood function representation includes: a reference-disease association representation, a reference-local association representation, and a reference prior distribution representation; The reference-disease association representation, the reference-local association representation and the reference prior distribution representation are solved respectively, and the likelihood function representation of the biomarker to be analyzed is determined according to the solution results.

5. The method according to claim 4, characterized in that The step of respectively solving the reference-disease association representation, the reference-local association representation and the reference prior distribution representation, and determining the likelihood function representation of the biomarker to be analyzed according to the solution results, comprises: Determining the reference-disease association representation according to the reference measurement value of the first biological sample and the corresponding disease output result based on a Logistic regression model; determining the reference-local association representation based on the local measurement value of the first biological sample and the reference measurement value based on a normal distribution assumption; A reference measurement value of the first biological sample is determined as the reference prior distribution representation.

6. The method according to claim 5, characterized in that Before the step of determining the prior distribution function of the target disease association degree based on the normal distribution assumption, the method includes: Determine several components of the target disease association degree according to the reference-disease association representation, the reference-local association representation, and the reference prior distribution representation; Divide each of the constituent elements into conjugate prior elements, information prior elements and sample parameter elements; Accordingly, the step of determining the prior distribution function of the target disease association degree based on the normal distribution assumption includes: Based on the distribution consistency principle of the reference measurement value, determining the corresponding prior distribution representation according to the conjugate prior class element; Based on the principle of standard normal distribution, the corresponding prior distribution representation is determined according to the information prior class elements; Based on a preset distribution principle, determining a corresponding priori representation according to the sample parameter class element; The prior distribution function of the target disease association degree is determined according to the prior representations corresponding to the conjugate prior class elements, the information prior class elements and the sample parameter class elements respectively.

7. The method according to claim 6, characterized in that The sample parameter type elements include a first sample parameter type element and a second sample parameter type element; The step of determining the corresponding prior representation according to the sample parameter class element based on the preset distribution principle includes: Based on the principle of overall normal distribution, determining the corresponding prior representation according to the first sample parameter class element; Based on the inverse gamma distribution principle, the corresponding prior representation is determined according to the second sample parameter class element.

8. A system for analyzing multi-source biomarker data, characterized in that: The system comprises: A data acquisition module, used to acquire multi-source biomarker data corresponding to the biomarker to be analyzed, wherein the multi-source biomarker data includes local measurement data and reference measurement data of the biological sample; An associated expression module, used to construct an associated expression model between the biomarker to be analyzed and the target disease according to the local measurement data and the reference measurement data; The association analysis module is used to solve the association expression model based on the Bayesian method to obtain the association analysis results between the biomarker to be analyzed and the target disease.

9. The system according to claim 8, characterized in that The multi-source biomarker data also includes the disease output results corresponding to each of the biological samples, the local measurement data includes the local measurement values ​​of all biological samples, and the reference measurement data includes the reference measurement values ​​of several biological samples; The association expression module is further used to construct a conditionally independent association expression based on the local measurement value, the reference measurement value and the disease output result, and in combination with a covariate vector; Among them, the associated expression, the joint probability of the disease output result and the reference measurement value satisfies the conditional independence principle.

10. The system according to claim 9, characterized in that The association analysis module is further used to construct a likelihood function representation of the biomarker to be analyzed based on all the biological samples; Determine the prior distribution function of the target disease association based on the normal distribution assumption; Determine a posterior distribution function of the target disease association degree according to the likelihood function representation and the prior distribution function; The posterior distribution function is sampled, and the association analysis result between the biomarker to be analyzed and the target disease is obtained according to the sampling result.

Images