Construction method of prostatic cancer patient prognosis risk prediction model

Abstract

The invention discloses a construction method of a prostatic cancer patient prognosis risk prediction model, and relates to the field of model construction, and the method comprises the following steps: obtaining each data source corresponding to the construction of the prostatic cancer patient prognosis risk prediction model; performing statistics on the basic information of each specified data source for analysis to obtain a data packet selection proportion of each specified data source; performing statistics on the current storage data packet combination information of each specified data source for analysis, and performing screening based on the data packet selection proportion of each specified data source to obtain an effective combination data packet of each specified data source; constructing a prostatic cancer patient prognosis risk prediction model on the basis of the effective combined data packets of the specified data sources, and recording the constructed model as a model to be evaluated; according to the method, the model needing to be evaluated is verified, the verification result of the model needing to be evaluated is obtained and correspondingly prompted, the problem that basic information of an analysis data source is lacked at present is solved, and the model building efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of model construction, and in particular to a method for constructing a prostate cancer patient prognosis risk prediction model. Background Art

[0002] The risk of biochemical recurrence of prostate cancer is affected by multiple factors. Due to the diversity and complexity of these factors, existing clinical strategies are difficult to effectively assess and predict patients' risk of biochemical recurrence. In order to better predict the prognosis risk of prostate cancer, more accurate tools are needed for assessment.

[0003] The prior art, such as the invention patent with announcement number: CN116092664B, is a method for establishing a pancreatic cancer prognosis prediction model, which specifically includes: step S1, a first calculation unit performs statistical analysis of pancreatic cancer prognosis-related factors on case information stored in a database to obtain factors related to the long-term survival of pancreatic cancer patients; step S2, a second calculation unit divides the cases into a modeling group and a verification group, performs a multifactor regression analysis on factors related to the long-term survival of pancreatic cancer patients in the modeling group cases, and screens independent risk factors that affect the prognosis of pancreatic cancer patients; step S3, a model building unit converts the screened independent risk factors into scores in proportion to build a pancreatic cancer patient prognosis prediction model; step S4, a central control unit performs repeated sampling correction on the constructed pancreatic cancer patient prognosis prediction model, and verifies the pancreatic cancer patient prognosis prediction model based on the case data of the verification group, so as to more accurately obtain the pancreatic cancer prognosis survival life expectancy.

[0004] Prior art, such as the invention patent with announcement number: CN117079801B, is a prognosis risk prediction system for colorectal cancer, specifically including: an image acquisition module and a prediction module; the image acquisition module is used to acquire the CT image of the patient to be predicted, and input the CT image into the prediction module; the prediction module is used to input the CT image into a preset CRC prognosis prediction model when the genomic data paired with the CT image is not detected, so that the CRC prognosis prediction model infers the genomic characteristics of the patient to be predicted based on the CT image characteristics, generates the predicted genomic characteristics of the patient to be predicted, and then predicts the prognosis risk of the patient to be predicted based on the CT image characteristics and the predicted genomic characteristics, and obtains the prognosis risk prediction result of the patient to be predicted. Through the present invention, the prognosis risk of the patient to be predicted can be predicted based on a single CT image.

[0005] Based on the above scheme, it can be seen that for the construction of prognostic risk prediction models, usually only the linear construction of the model is emphasized, and usually only the collected data are uniformly invested in the construction of the model. However, in actual applications, the quality of data varies greatly. If all collected data are uniformly invested in the construction of the model, the model accuracy will be reduced, the generalization ability will be weakened, and the efficiency of model construction will be reduced. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention provides a method for constructing a prostate cancer patient prognosis risk prediction model. To achieve the above purpose, the present invention is implemented through the following technical scheme: A method for constructing a prostate cancer patient prognosis risk prediction model, comprising:

[0007] Obtaining data sources corresponding to the construction of a prostate cancer patient prognosis risk prediction model, recorded as designated data sources;

[0008] Analyze the basic information of each designated data source and obtain the data packet selection ratio of each designated data source;

[0009] Counting and analyzing the combination information of currently stored data packets of each designated data source, and filtering based on the data packet selection ratio of each designated data source to obtain the valid combination data packets of each designated data source;

[0010] Based on the valid combined data packages of each designated data source, a prognostic risk prediction model for prostate cancer patients is constructed, and the constructed model is recorded as a model to be evaluated;

[0011] Verify the model to be evaluated and give corresponding prompts based on the verification results of the model to be evaluated.

[0012] As a preferred technical solution, the statistical analysis of basic information of each designated data source specifically includes:

[0013] Based on the basic information of each designated data source, the basic information of each designated data source specifically includes the average daily page views, the average daily visit duration and the average daily download volume of each designated data source, and the basic information evaluation value of each designated data source is obtained through analysis. The specific execution conditions are as follows:

[0014]

[0015] Among them, S i is the basic information evaluation value of the i-th specified data source, V i is the average daily page views of the ith specified data source, T i is the average daily visit time of the personnel of the i-th designated data source, D iis the average daily download volume of the i-th designated data source, ω1 is the average daily browsing weight factor, ω2 is the average daily access time weight factor, ω3 is the average daily download weight factor, i is the number of each designated data source, i = 1, 2, 3, ..., n, n is the number of designated data sources.

[0016] As a preferred technical solution, the specific process of obtaining the data packet selection ratio of each designated data source is as follows:

[0017] Based on the basic information evaluation value of each designated data source, the data packet selection ratio corresponding to each basic information evaluation value interval stored in the database is matched to obtain the data packet selection ratio of each designated data source.

[0018] As a preferred technical solution, the statistical analysis of the current stored data packet combination information of each designated data source specifically includes:

[0019] Counting the currently stored data packet combination information of each designated data source, wherein the currently stored data packet combination information includes first combination information and second combination information;

[0020] The first combined information includes an initial storage time node and a neighboring update time node of each stored data packet, and based on the current time node, obtains an initial storage interval time difference and a neighboring update interval time difference of each stored data packet, thereby calculating a timeliness score of the current stored data packet of each designated data source;

[0021] The specific processing conditions for calculating the timeliness score of each stored data packet of each designated data source are as follows:

[0022]

[0023] Among them, S 1,i,u Score the timeliness of the u-th stored data packet of the i-th specified data source, A i,u is the initial storage interval time difference of the u-th storage data packet of the i-th specified data source, U i,u is the time difference between the adjacent update intervals of the u-th stored data packet of the i-th specified data source, A max The time difference of defining the initial storage interval pre-stored in the database, U max The time difference between adjacent update intervals stored in the database, is the weight factor of the initial storage interval time difference, is the weight factor of the time difference between adjacent update intervals;

[0024] The second combination information includes the sequencing data read length, the number of historical visitors, and the historical average stay time of each stored data packet, and calculates the data quality score of each stored data packet of each designated data source. The specific processing conditions include:

[0025]

[0026] Among them, S 2,i,u The data quality score of the u-th stored data packet of the i-th specified data source, Q 1,i,u is the sequencing data read length of the u-th storage data package of the i-th specified data source, Q 2,i,u is the number of historical visitors to the u-th storage data packet of the i-th specified data source, Q 3,i,u is the historical average length of stay of the u-th stored data packet of the i-th specified data source, Q 1,0 represents the reference sequencing data read length in the database, Q 2,0 represents the number of reference historical visitors in the database, Q 3,0 represents the reference historical average length of stay in the database, α1 is the weight factor of sequencing data read length, α2 is the weight factor of the number of historical visitors, and α3 is the weight factor of the historical average length of stay.

[0027] As a preferred technical solution, the statistical analysis of the currently stored data packet combination information of each designated data source also includes:

[0028] Based on the timeliness score of each storage data package of each designated data source and the data quality score of each storage data package of each designated data source, a comprehensive score of each storage data package of each designated data source is obtained. The specific processing conditions are:

[0029] S i,u =β1*S 1,i,u +β2*S 2,i,u ;

[0030] Among them, S i,u is the comprehensive score of the u-th stored data packet of the i-th specified data source, S 1,i,u S is the timeliness score of the u-th stored data packet of the i-th specified data source, 2,i,u is the data quality score of the u-th stored data packet of the i-th specified data source, β1 is the timeliness weight factor, and β2 is the data quality weight factor.

[0031] As a preferred technical solution, the selection ratio of the data packets of each designated data source is used to screen and obtain the valid combined data packets of each designated data source, specifically including:

[0032] Counting the total number of currently stored data packets of each designated data source, and based on the data packet selection ratio of each designated data source, multiplying the total number of currently stored data packets of each designated data source by the corresponding data packet selection ratio to obtain the number of data packets selected by each designated data source;

[0033] Arrange the comprehensive scores of the storage data packets of the designated data sources in descending order to obtain the arrangement order of the comprehensive scores of the storage data packets of the designated data sources, and extract a number of corresponding storage data packets according to the number of data packets selected from the designated data sources, and record them as valid data packets;

[0034] Valid data packets of each designated data source are counted and jointly marked as valid combined data packets of each designated data source.

[0035] As a preferred technical solution, the effective combination of data packages based on each designated data source and the construction of a prostate cancer patient prognosis risk prediction model specifically include:

[0036] Based on the effective combined data packets of each designated data source, and processed by Cox regression model and LASSO regression model, a prostate cancer prognosis risk prediction model was constructed.

[0037] As a preferred technical solution, the verification of the model to be evaluated specifically includes:

[0038] Extracting prostate cancer prognosis sample data from a database and inputting it into the model to be evaluated to obtain prediction result data of the model to be evaluated, wherein the prediction result data includes the number of true positive samples, the number of true negative samples, the number of false positive samples, and the number of false negative samples, thereby obtaining a verification evaluation index of the model to be evaluated;

[0039] The verification evaluation index of the model to be evaluated is used to quantify the verification result data of the model to be evaluated and to evaluate the performance of the model to be evaluated.

[0040] As a preferred technical solution, the verification result of the model to be evaluated is given a corresponding prompt, specifically including:

[0041] The verification evaluation index of the model to be evaluated is compared with the verification evaluation index threshold stored in the database. When the verification evaluation index of the model to be evaluated is lower than or equal to the verification evaluation index threshold, the verification result of the model to be evaluated is defined as a model that needs to be retrained. Conversely, when the verification evaluation index of the model to be evaluated is higher than the verification evaluation index threshold, the verification result of the model to be evaluated is defined as a valid training model. Therefore, the verification result of the model to be evaluated is prompted accordingly.

[0042] As a preferred technical solution, the verification evaluation index of the model to be evaluated has the following specific processing conditions:

[0043]

[0044] τ represents the validation evaluation index of the model to be evaluated, Z represents the total number of samples in the prostate cancer prognosis sample data, B1 represents the reference proportion of true positive samples defined in the database, B2 represents the reference proportion of true negative samples defined in the database, B3 represents the reference proportion of false positive samples defined in the database, B4 represents the reference proportion of false negative samples defined in the database, TP is the number of true positive samples, TN is the number of true negative samples, FP is the number of false positive samples, and FN is the number of false negative samples.

[0045] Compared with the prior art, the embodiments of the present invention have at least the following beneficial effects:

[0046] (1) The present invention analyzes the basic information of each designated data source by statistics to obtain the data packet selection ratio of each designated data source, which can help optimize the integration process of different data sources. By analyzing the selection ratio of data packets, the overall quality of each data source can be evaluated, thereby providing reliable guarantee for the selection of subsequent data packets.

[0047] (2) The present invention can identify and screen out high-quality data by analyzing the data packets under the data source instead of directly putting all the collected data into the construction of the model, effectively reducing the impact of bad data on the model, and can screen out the most valuable parts for model construction, avoiding processing a large amount of redundant or invalid data. This can not only reduce the model training time and resource consumption, but also make the model construction process more efficient.

[0048] (3) The present invention uses Cox-LASSO to construct a model, which can work effectively in a high-dimensional environment and solve the overfitting problem of traditional Cox regression in high-dimensional conditions. Cox-LASSO can handle the multicollinearity problem between variables through regularization, thereby improving the stability and prediction performance of the model.

[0049] Of course, any product implementing the present invention does not necessarily need to achieve all of the above advantages at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] In the description of the present invention, it should be understood that the terms "opening", "upper", "lower", "thickness", "top", "middle", "length", "inside", "all around" and the like indicating orientation or positional relationship are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the components or elements referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0053] See also Figure 1 As shown, the embodiment of the present invention provides a method for constructing a prostate cancer patient prognosis risk prediction model, which specifically includes:

[0054] Obtaining data sources corresponding to the construction of a prostate cancer patient prognosis risk prediction model, recorded as designated data sources;

[0055] Analyze the basic information of each designated data source and obtain the data packet selection ratio of each designated data source;

[0056] The statistical analysis of basic information of each designated data source specifically includes: based on the basic information of each designated data source, the basic information of each designated data source specifically includes the average daily browsing volume, the average daily access time and the average daily download volume of each designated data source, and obtaining the basic information evaluation value of each designated data source through analysis, and the specific execution conditions are as follows:

[0057]

[0058] Among them, S i is the basic information evaluation value of the i-th specified data source, V i is the average daily page views of the ith specified data source, T i is the average daily visit time of the personnel of the i-th designated data source, D i is the average daily download volume of the i-th designated data source, ω1 is the average daily browsing weight factor, ω2 is the average daily access time weight factor, ω3 is the average daily download weight factor, i is the number of each designated data source, i = 1, 2, 3, ..., n, n is the number of designated data sources.

[0059] It is worth noting that the average daily page views weight factor, the average daily access time weight factor and the average daily downloads weight factor all take values ​​between 0 and 1 and satisfy ω1+ω2+ω3=1. In an embodiment, the preset value can be directly extracted from the database. In the application, through the relationship between the basic information evaluation value of each designated data source in the historical data and the average daily page views, the average daily access time and the average daily downloads of each designated data source, a mapping set of the basic information evaluation value of each designated data source and the average daily page views, the average daily access time and the average daily downloads of each designated data source and their corresponding average daily page views weight factor, average daily access time weight factor and average daily downloads weight factor of each designated data source is established. The real-time average daily page views, the average daily access time and the average daily downloads of each designated data source are input, and the corresponding average daily page views weight factor, average daily access time weight factor and average daily downloads weight factor of each designated data source are obtained from the mapping set.

[0060] It should be noted that, generally speaking, a higher number of views means that more users visit the data source. If the content of the data source is rich, it will usually lead to an increase in the access time. Therefore, an increase in views is often accompanied by an increase in the access time. A higher number of views means that more users are exposed to the data source. If the data source provides high-quality downloadable content, this will lead to an increase in downloads. Therefore, an increase in views may be accompanied by an increase in downloads. If users stay on the data source for a long time, this usually indicates that they are carefully browsing or studying the content of the data source. This in-depth use usually leads to a higher number of downloads because users may download the content they are interested in for further use. A comprehensive analysis of views, access time and downloads can help evaluate the quality of the data source. High views, high access time and high downloads usually indicate that the data source is rich in content, high in quality and has strong practicality.

[0061] The specific process of obtaining the data packet selection ratio of each designated data source is as follows: based on the basic information evaluation value of each designated data source, the data packet selection ratio corresponding to each basic information evaluation value interval stored in the database is matched to obtain the data packet selection ratio of each designated data source.

[0062] The present invention analyzes the basic information of each designated data source by statistics to obtain the data packet selection ratio of each designated data source, which can help optimize the integration process of different data sources. By analyzing the selection ratio of data packets, the overall quality of each data source can be evaluated, thereby providing reliable guarantee for the selection of subsequent data packets.

[0063] The statistical analysis of the currently stored data packet combination information of each designated data source specifically includes:

[0064] Counting and analyzing the combination information of currently stored data packets of each designated data source, and filtering based on the data packet selection ratio of each designated data source to obtain the valid combination data packets of each designated data source;

[0065] Counting the currently stored data packet combination information of each designated data source, wherein the currently stored data packet combination information includes first combination information and second combination information;

[0066] The first combined information includes an initial storage time node and a neighboring update time node of each stored data packet, and based on the current time node, obtains an initial storage interval time difference and a neighboring update interval time difference of each stored data packet, thereby calculating a timeliness score of the current stored data packet of each designated data source;

[0067] The specific processing conditions for calculating the timeliness score of each stored data packet of each designated data source are as follows:

[0068]

[0069] Among them, S 1,i,u Score the timeliness of the u-th stored data packet of the i-th specified data source, A i,u is the initial storage interval time difference of the u-th storage data packet of the i-th specified data source, U i,u is the time difference between the adjacent update intervals of the u-th stored data packet of the i-th specified data source, A max The time difference of defining the initial storage interval pre-stored in the database, U max The time difference between adjacent update intervals stored in the database, is the weight factor of the initial storage interval time difference, is the weight factor of the time difference between adjacent update intervals;

[0070] It is worth noting that the weight factor of the initial storage interval time difference and the weight factor of the adjacent update interval time difference are both between 0 and 1 and satisfy In an embodiment, a preset value can be directly extracted from a database. In an application, a mapping set is established between the timeliness score of each storage data packet of each designated data source in historical data and the initial storage interval time difference and adjacent update interval time difference of each storage data packet, and the corresponding weight factors of the initial storage interval time difference and the weight factors of the adjacent update interval time difference. The real-time initial storage interval time difference and adjacent update interval time difference of each storage data packet are input, and the corresponding weight factors of the initial storage interval time difference and the weight factors of the adjacent update interval time difference are obtained from the mapping set.

[0071] The second combination information includes the sequencing data read length, the number of historical visitors, and the historical average stay time of each stored data packet, and calculates the data quality score of each stored data packet of each designated data source. The specific processing conditions include:

[0072]

[0073] Among them, S 2,i,u The data quality score of the u-th stored data packet of the i-th specified data source, Q 1,i,u is the sequencing data read length of the u-th storage data package of the i-th specified data source, Q 2,i,u is the number of historical visitors to the u-th storage data packet of the i-th specified data source, Q 3,i,u is the historical average length of stay of the u-th stored data packet of the i-th specified data source, Q 1,0 represents the reference sequencing data read length in the database, Q 2,0 represents the number of reference historical visitors in the database, Q 3,0 represents the reference historical average length of stay in the database, α1 is the weight factor of sequencing data read length, α2 is the weight factor of the number of historical visitors, and α3 is the weight factor of the historical average length of stay.

[0074] It is worth noting that the sequencing data read length weight factor, the historical number of visiting people weight factor and the historical average stay time weight factor all take values ​​between 0 and 1 and satisfy α1+α2+α3=1. In an embodiment, the preset value can be directly extracted from the database. In the application, through the relationship between the data quality score of each storage data packet of each designated data source in the historical data and the sequencing data read length, the historical number of visiting people, and the historical average stay time of each storage data packet, a mapping set of the data quality score of each storage data packet of each designated data source and the sequencing data read length, the historical number of visiting people, the historical average stay time of each storage data packet and the corresponding sequencing data read length weight factor, the historical number of visiting people weight factor and the historical average stay time weight factor is established, the real-time sequencing data read length, the historical number of visiting people, and the historical average stay time of each storage data packet are input, and the corresponding sequencing data read length weight factor, the historical number of visiting people weight factor and the historical average stay time weight factor are obtained from the mapping set.

[0075] The statistical analysis of the current storage data package combination information of each designated data source also includes: obtaining a comprehensive score of each storage data package of each designated data source based on the timeliness score of each storage data package of each designated data source and the data quality score of each storage data package of each designated data source, and the specific processing conditions are:

[0076] S i,u=β1*S 1,i,u +β2*S 2,i,u ;

[0077] Among them, S i,u is the comprehensive score of the u-th stored data packet of the i-th specified data source, S 1,i,u S is the timeliness score of the u-th stored data packet of the i-th specified data source, 2,i,u is the data quality score of the u-th stored data packet of the i-th specified data source, β1 is the timeliness weight factor, and β2 is the data quality weight factor.

[0078] It is worth noting that the timeliness weight factor and the data quality weight factor are both between 0 and 1, and satisfy β1+β2=1. In an embodiment, the preset value can be directly extracted from the database. In the application, through the relationship between the comprehensive score of each storage data packet of each designated data source in the historical data and the timeliness score of each storage data packet of each designated data source and the data quality score of each storage data packet of each designated data source, a mapping set of the comprehensive score of each storage data packet of each designated data source and the timeliness score of each storage data packet of each designated data source and the data quality score of each storage data packet of each designated data source and their corresponding timeliness weight factor and data quality weight factor is established, the real-time timeliness score of each storage data packet of each designated data source and the data quality score of each storage data packet of each designated data source are input, and the corresponding timeliness weight factor and data quality weight factor are obtained from the mapping set.

[0079] It should be noted that the timeliness score directly affects the overall result of the data quality score. If the timeliness of the data cannot be guaranteed, even if the data performs well in other aspects (such as completeness and accuracy), its overall quality will be greatly reduced, because outdated data may not reflect the current real situation, which will lead to deviations in analysis and decision-making based on these data. The timeliness score may also indirectly affect the data quality score by affecting other data quality dimensions. For example, outdated data may lead to increased data inconsistency because newly generated data may conflict with old data. At the same time, outdated data may also reduce the accessibility of data because users may prefer to use the latest and most accurate data.

[0080] The present invention analyzes the data packets under the data source instead of directly putting all the collected data into the construction of the model. It can identify and screen out high-quality data, effectively reduce the impact of bad data on the model, and can screen out the most valuable parts for model construction, avoiding processing of large amounts of redundant or invalid data. This can not only reduce model training time and resource consumption, but also make the model construction process more efficient.

[0081] The method of screening based on the data packet selection ratio of each designated data source to obtain the valid combined data packets of each designated data source specifically includes: counting the total number of currently stored data packets of each designated data source, and based on the data packet selection ratio of each designated data source, multiplying the total number of currently stored data packets of each designated data source by the corresponding data packet selection ratio to obtain the number of data packets selected from each designated data source;

[0082] Arrange the comprehensive scores of the storage data packets of the designated data sources in descending order to obtain the arrangement order of the comprehensive scores of the storage data packets of the designated data sources, and extract a number of corresponding storage data packets according to the number of data packets selected from the designated data sources, and record them as valid data packets;

[0083] Valid data packets of each designated data source are counted and jointly marked as valid combined data packets of each designated data source.

[0084] Based on the valid combined data packages of each designated data source, a prognostic risk prediction model for prostate cancer patients is constructed, and the constructed model is recorded as a model to be evaluated;

[0085] The method of constructing a prostate cancer patient prognosis risk prediction model based on a valid combination data package of each designated data source specifically includes: constructing a prostate cancer prognosis risk prediction model based on a valid combination data package of each designated data source and processing it through a Cox regression model and a LASSO regression model.

[0086] The effective combined data packets based on each designated data source are processed by the Cox regression model and the LASSO regression model to construct a prostate cancer prognosis risk prediction model. The specific process is as follows:

[0087] The data in the valid combination data packages of each specified data source after standardization were subjected to differential expression analysis, the gene expression differences between prostate cancer samples and normal samples were compared, the Pearson correlation coefficient between the gene expression data was calculated, and a correlation matrix was constructed. The Pearson correlation coefficient is a statistical indicator for measuring the linear relationship between two variables, and the value range is -1 to 1, 1 indicates a perfect positive correlation, -1 indicates a perfect negative correlation, and 0 indicates no linear relationship.

[0088] Using statistical methods to compare the gene expression differences between prostate cancer samples and normal samples, setting the threshold of differential expression, screening out genes with significant differential expression, calculating the Pearson correlation coefficients between different samples for the screened out genes with significant differential expression, the Pearson correlation coefficients are used to characterize the linear relationship between gene expression amounts, organizing the Pearson correlation coefficients between all gene pairs to construct a correlation matrix, and based on the correlation matrix, converting the correlation matrix into an adjacency matrix using a soft threshold method;

[0089] Based on the converted adjacency matrix, the topological overlap between gene pairs was calculated to construct a weighted gene co-expression network.

[0090] Genes were clustered using hierarchical clustering. During the hierarchical clustering process, genes were gradually merged into larger clusters until a predetermined number of clusters was reached. Each cluster represented a group of highly related genes. Gene sets with highly similar expression patterns were recorded as modules, and module scores were calculated for each module. The module scores were compared with the module score threshold.

[0091] The module score is used to characterize the comprehensive measure of all gene expression patterns within the module;

[0092] The module score is calculated for each module, and the specific conditions are:

[0093]

[0094] Among them, MS M is the module score of module M, W gh is a weight matrix constructed based on the topological overlap matrix, (∑ h∈M,h≠ g W gh ) represents the sum of the connection strengths between gene g and all other genes in module M, which is used as the weight of g, e r g is the expression level of gene g in module M.

[0095] Based on the module score, genes with module scores higher than the module score threshold were selected as key genes. The Cox regression model was used to analyze the relationship between the expression data of key genes and the prognosis time, and the regression coefficient of each gene was obtained. LASSO regression was used to select variables for key genes, and the most important genes for prognostic risk prediction were screened by adjusting the regularization parameters.

[0096] Combining the results of Cox regression and LASSO regression, a prostate cancer prognostic risk prediction model was constructed.

[0097] The prostate cancer prognosis risk prediction model is constructed by combining the results of Cox regression and LASSO regression, which specifically includes:

[0098]

[0099] Among them, x tg is the expression level of gene g in sample t, δ t is the event indicator of sample t, indicating whether an event with a worse prognosis has occurred, β0 is the baseline value of the logarithmic hazard function, β gis the regression coefficient of gene expression, which indicates the change of logarithmic risk function for each unit change of gene expression. λ is the regularization parameter used to control the intensity of penalty. The larger the value of λ, the more coefficients will be estimated to be zero. t is the sample number, t=1,2,3,...,m, m is the number of samples, g is the gene number, g=1,2,3,...,p, p is the number of genes, R t is the risk set of sample t, and k is the sample index when summing the logarithmic partial likelihood functions in the regression analysis;

[0100] By minimizing the objective function, a set of optimal regression coefficients was obtained to construct a prostate cancer prognosis risk prediction model.

[0101] The present invention uses Cox-LASSO to build a model, which can work effectively in a high-dimensional environment and solve the overfitting problem of traditional Cox regression in high-dimensional conditions. Cox-LASSO can handle the multicollinearity problem between variables through regularization, thereby improving the stability and prediction performance of the model.

[0102] Verify the model to be evaluated and give corresponding prompts based on the verification results of the model to be evaluated.

[0103] The verification of the model to be evaluated specifically includes: extracting prostate cancer prognosis sample data from a database and inputting it into the model to be evaluated to obtain prediction result data of the model to be evaluated, wherein the prediction result data includes the number of true positive samples, the number of true negative samples, the number of false positive samples and the number of false negative samples, thereby obtaining the verification evaluation index of the model to be evaluated;

[0104] The verification evaluation index of the model to be evaluated is used to quantify the verification result data of the model to be evaluated and to evaluate the performance of the model to be evaluated.

[0105] The verification result of the model to be evaluated is given a corresponding prompt, specifically including: based on the verification evaluation index of the model to be evaluated, the verification evaluation index threshold stored in the database is compared; when the verification evaluation index of the model to be evaluated is lower than or equal to the verification evaluation index threshold, the verification result of the model to be evaluated is defined as a model that needs to be retrained; conversely, when the verification evaluation index of the model to be evaluated is higher than the verification evaluation index threshold, the verification result of the model to be evaluated is defined as a valid training model, thereby giving a corresponding prompt for the verification result of the model to be evaluated.

[0106] The verification evaluation index of the model to be evaluated, the specific processing conditions are:

[0107]

[0108] τ represents the validation evaluation index of the model to be evaluated, Z represents the total number of samples in the prostate cancer prognosis sample data, B1 represents the reference proportion of true positive samples defined in the database, B2 represents the reference proportion of true negative samples defined in the database, B3 represents the reference proportion of false positive samples defined in the database, B4 represents the reference proportion of false negative samples defined in the database, TP is the number of true positive samples, TN is the number of true negative samples, FP is the number of false positive samples, and FN is the number of false negative samples.

[0109] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device 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 device.

[0110] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that technicians in the relevant technical field can understand and use the present invention well. As long as they do not deviate from the structure of the present invention or exceed the scope defined by the present invention, they should all belong to the protection scope of the present invention.

Claims

1. A method for constructing a prostate cancer patient prognosis risk prediction model, characterized in that: include: Obtaining data sources corresponding to the construction of a prostate cancer patient prognosis risk prediction model, recorded as designated data sources; Analyze the basic information of each designated data source and obtain the data packet selection ratio of each designated data source; Counting and analyzing the combination information of currently stored data packets of each designated data source, and filtering based on the data packet selection ratio of each designated data source to obtain the valid combination data packets of each designated data source; Based on the valid combined data packages of each designated data source, a prognostic risk prediction model for prostate cancer patients is constructed, and the constructed model is recorded as a model to be evaluated; Verify the model to be evaluated and give corresponding prompts based on the verification results of the model to be evaluated.

2. The method for constructing a prostate cancer patient prognosis risk prediction model according to claim 1, characterized in that: The statistical analysis of basic information of each designated data source specifically includes: Based on the basic information of each designated data source, the basic information of each designated data source specifically includes the average daily page views, the average daily visit duration and the average daily download volume of each designated data source, and the basic information evaluation value of each designated data source is obtained through analysis. The specific execution conditions are as follows: Among them, S i is the basic information evaluation value of the i-th specified data source, V i is the average daily page views of the ith specified data source, T i is the average daily visit time of the personnel of the i-th designated data source, D i is the average daily download volume of the i-th designated data source, ω1 is the average daily browsing weight factor, ω2 is the average daily access time weight factor, ω3 is the average daily download weight factor, i is the number of each designated data source, i = 1, 2, 3, ..., n, n is the number of designated data sources.

3. The method for constructing a prostate cancer patient prognosis risk prediction model according to claim 2, characterized in that: The specific process of obtaining the data packet selection ratio of each designated data source is as follows: Based on the basic information evaluation value of each designated data source, the data packet selection ratio corresponding to each basic information evaluation value interval stored in the database is matched to obtain the data packet selection ratio of each designated data source.

4. The method for constructing a prostate cancer patient prognosis risk prediction model according to claim 3, characterized in that: The statistical analysis of the currently stored data packet combination information of each designated data source specifically includes: Counting the currently stored data packet combination information of each designated data source, wherein the currently stored data packet combination information includes first combination information and second combination information; The first combined information includes an initial storage time node and a neighboring update time node of each stored data packet, and based on the current time node, obtains an initial storage interval time difference and a neighboring update interval time difference of each stored data packet, thereby calculating a timeliness score of the current stored data packet of each designated data source; The specific processing conditions for calculating the timeliness score of each stored data packet of each designated data source are as follows: Among them, S 1,i,u Score the timeliness of the u-th stored data packet of the i-th specified data source, A i,u is the initial storage interval time difference of the u-th storage data packet of the i-th specified data source, U i,u is the time difference between the adjacent update intervals of the u-th stored data packet of the i-th specified data source, A max The time difference of defining the initial storage interval pre-stored in the database, U max The time difference between adjacent update intervals stored in the database, is the weight factor of the initial storage interval time difference, is the weight factor of the time difference between adjacent update intervals; The second combination information includes the sequencing data read length, the number of historical visitors, and the historical average stay time of each stored data packet, and calculates the data quality score of each stored data packet of each designated data source. The specific processing conditions include: Among them, S 2,i,u The data quality score of the u-th stored data packet of the i-th specified data source, Q 1,i,u is the sequencing data read length of the u-th storage data package of the i-th specified data source, Q 2,i,u is the number of historical visitors to the u-th storage data packet of the i-th specified data source, Q 3,i,u is the historical average length of stay of the u-th stored data packet of the i-th specified data source, Q 1,0 represents the reference sequencing data read length in the database, Q 2,0 represents the number of reference historical visitors in the database, Q 3,0 represents the reference historical average length of stay in the database, α1 is the weight factor of sequencing data read length, α2 is the weight factor of the number of historical visitors, and α3 is the weight factor of the historical average length of stay.

5. The method for constructing a prostate cancer patient prognosis risk prediction model according to claim 4, characterized in that: The statistical analysis of the currently stored data packet combination information of each designated data source also includes: Based on the timeliness score of each storage data package of each designated data source and the data quality score of each storage data package of each designated data source, a comprehensive score of each storage data package of each designated data source is obtained. The specific processing conditions are: S i,u =β1*S 1,i,u +β2*S 2,i,u ; Among them, S i,u is the comprehensive score of the u-th stored data packet of the i-th specified data source, S 1,u,u S is the timeliness score of the u-th stored data packet of the i-th specified data source, 2,i,u is the data quality score of the u-th stored data packet of the i-th specified data source, β1 is the timeliness weight factor, and β2 is the data quality weight factor.

6. The method for constructing a prostate cancer patient prognosis risk prediction model according to claim 5, characterized in that: The method of selecting data packets based on the ratio of each designated data source to obtain valid combined data packets of each designated data source specifically includes: Counting the total number of currently stored data packets of each designated data source, and based on the data packet selection ratio of each designated data source, multiplying the total number of currently stored data packets of each designated data source by the corresponding data packet selection ratio to obtain the number of data packets selected by each designated data source; Arrange the comprehensive scores of the storage data packets of the designated data sources in descending order to obtain the arrangement order of the comprehensive scores of the storage data packets of the designated data sources, and extract a number of corresponding storage data packets according to the number of data packets selected from the designated data sources, and record them as valid data packets; Valid data packets of each designated data source are counted and jointly marked as valid combined data packets of each designated data source.

7. The method for constructing a prostate cancer patient prognosis risk prediction model according to claim 6, characterized in that: The effective combination of data packages based on the designated data sources and the construction of a prostate cancer patient prognosis risk prediction model specifically include: Based on the valid combined data packets of each designated data source, and processed by Cox regression model and LASSO regression model, a prostate cancer prognosis risk prediction model was constructed.

8. The method for constructing a prostate cancer patient prognosis risk prediction model according to claim 1, characterized in that: The verification of the model to be evaluated specifically includes: Extracting prostate cancer prognosis sample data from a database and inputting it into the model to be evaluated to obtain prediction result data of the model to be evaluated, wherein the prediction result data includes the number of true positive samples, the number of true negative samples, the number of false positive samples, and the number of false negative samples, thereby obtaining a verification evaluation index of the model to be evaluated; The verification evaluation index of the model to be evaluated is used to quantify the verification result data of the model to be evaluated and to evaluate the performance of the model to be evaluated.

9. The method for constructing a prostate cancer patient prognosis risk prediction model according to claim 8, characterized in that: The verification result of the model to be evaluated is given a corresponding prompt, specifically including: The verification evaluation index of the model to be evaluated is compared with the verification evaluation index threshold stored in the database. When the verification evaluation index of the model to be evaluated is lower than or equal to the verification evaluation index threshold, the verification result of the model to be evaluated is defined as a model that needs to be retrained. Conversely, when the verification evaluation index of the model to be evaluated is higher than the verification evaluation index threshold, the verification result of the model to be evaluated is defined as a valid training model. Therefore, the verification result of the model to be evaluated is prompted accordingly.

10. The method for constructing a prostate cancer patient prognosis risk prediction model according to claim 9, characterized in that: The verification evaluation index of the model to be evaluated, the specific processing conditions are: τ represents the validation evaluation index of the model to be evaluated, Z represents the total number of samples in the prostate cancer prognosis sample data, B1 represents the reference proportion of true positive samples defined in the database, B2 represents the reference proportion of true negative samples defined in the database, B3 represents the reference proportion of false positive samples defined in the database, B4 represents the reference proportion of false negative samples defined in the database, TP is the number of true positive samples, TN is the number of true negative samples, FP is the number of false positive samples, and FN is the number of false negative samples.

Images