WO2022048071A1

WO2022048071A1 - Tumor risk grading method and system, terminal, and storage medium

Info

Publication number: WO2022048071A1
Application number: PCT/CN2020/139298
Authority: WO
Inventors: 李霞; 蔡云鹏
Original assignee: 中国科学院深圳先进技术研究院
Priority date: 2020-09-03
Filing date: 2020-12-25
Publication date: 2022-03-10
Also published as: CN112117003A

Abstract

A tumor risk grading method and system, a terminal, and a storage medium. The method comprises: acquiring transcriptome data and clinical survival state information of patients; acquiring prognosis-associated genes of the patients according to the transcriptome data and the clinical survival state information; inputting the clinical survival state information and the prognosis-associated genes into a statistical model for molecular marker screening, and acquiring gene expression values of screened molecular markers in the patients; and scoring the patients according to the molecular markers and the gene expression values, and conducting tumor risk grading on the patients according to the score distribution. All prognosis-associated genes can be maximally retained, such that the obtained prognosis-associated genes have better accuracy, the acquisition of the molecular marker has a lower error rate, and risk grading results are more reasonable.

Description

A tumor risk grade classification method, system, terminal and storage medium

technical field

The present application belongs to the field of biomedical technology, and in particular relates to a method, system, terminal and storage medium for classifying tumor risk levels.

Background technique

Although there are certain treatment strategies for tumors, the prognosis of tumors has always been poor. Due to the high possibility of express development and metastasis of tumors, dividing tumors into different risk groups is helpful for early detection of high-risk disease groups , plays an important role in tumor diagnosis and providing appropriate regulatory measures. Traditionally, clinical phenotypic characteristics such as the clinical stage and metastasis of tumor patients are often used to assess the risk of patients, but this method does not fully consider the inherent molecular level differences of patients, which makes the risk judgment lack of accuracy. In practice, accurate prognostic judgment cannot be provided for tumor patients.

In recent years, molecular markers based on gene expression data acquisition and patient prognosis have been widely used. The current general method for classifying different risk groups based on gene expression levels mainly includes three steps:

1) Population grouping, such as dividing the population into tumor group and normal group, metastasis group and non-metastasis group, and then obtaining differentially expressed genes between groups according to gene expression levels;

2) Combined with the survival information of patients, use survival analysis to obtain genes associated with prognosis, and further use statistical models to screen molecular markers associated with prognosis;

3) Individuals are scored according to the gene expression levels of molecular markers combined with regression coefficients in survival analysis, and then ranked according to the scores, and the median is used as the cut-off point to divide the population into different risk groups.

technical problem

To sum up, the shortcomings of the existing risk group classification methods include:

1. Existing methods rely solely on simple population division to obtain differentially expressed genes, ignoring the real differentially expressed genes between populations, that is, the molecular composition of each individual patient is different, and genes with expression differences between groups may be within the group. There are also huge differences in expression; in addition, some genes are not differentially expressed between different groups obtained by dividing the population, but they are still associated with prognosis, and existing methods may miss other potential prognosis-associated genes.

2. Existing methods use lasso penalized Cox proportional hazards model in the process of screening prognostic-related genes. This model requires cross-validation to select the Lambda value, and based on this value, molecular markers are re-determined, and the results of existing methods often have Randomness, that is, the Lambda value obtained from the result of each operation is different.

3. When the existing method divides the risk of the population, the median of the scoring value is used as the dividing point, and this value may not be the best point for dividing all the scores, which will lead to certain errors in the final risk division result. sex.

technical solutions

The present application provides a method, system, terminal, and storage medium for classifying tumor risk levels, aiming to solve one of the above-mentioned technical problems in the prior art at least to a certain extent.

In order to solve the above problems, the application provides the following technical solutions:

A tumor risk grade classification method, comprising the following steps:

Obtaining the patient's transcriptome data and clinical survival status information; wherein, the transcriptome data at least includes the gene expression value of each gene in the patient;

Obtain the prognosis-related genes of the patient according to the transcriptome data and the clinical survival status information;

Inputting the clinical survival status information and prognosis-related genes into a statistical model for molecular marker screening, and obtaining the gene expression values of the screened molecular markers in the patient;

The patient is scored according to the molecular marker and the gene expression value, and the patient is divided into tumor risk levels according to the score distribution.

The technical solution adopted in the embodiment of the present application further includes: the obtaining of the patient's transcriptome data and clinical survival status information includes:

The gene expression value of each gene in the transcriptome data was logarithmically transformed by statistics, the genes with low expression levels were eliminated, and the genes that were expressed in more than a set proportion of patients were retained; Obtain the clinical survival status information of the patient from the data;

The clinical survival status information includes patient ID number, survival status and survival time.

The technical solution adopted in the embodiment of the present application further includes: the obtaining of the patient's prognosis-related genes according to the transcriptome data and the clinical survival status information includes:

Sort the gene expression values of each gene in the patient, and obtain the high and low expression populations corresponding to each gene respectively;

Calculate the correlation between the expression level of each gene and prognosis in combination with the clinical survival status information and the high and low expression population, output the regression coefficient and P value of each gene, and filter all the genes of the patient according to the size of the P value , obtain the prognosis-related genes of the patient, and extract the gene expression value of each prognosis-related gene in the patient to generate a prognosis-related gene expression matrix.

The technical solution adopted in the embodiment of the present application further includes: the sorting of the gene expression values of each gene in the patient, and obtaining the high and low expression populations corresponding to each gene respectively include:

The high and low expression populations corresponding to each gene were obtained by the method of dichotomy, trichotomy, or quartile, respectively.

Based on the high and low expression populations obtained by different algorithms, the relationship between the expression level and prognosis is calculated respectively. If a gene is associated with prognosis in the high and low expression populations obtained by at least two algorithms, the gene is determined as a prognosis. associated genes.

The technical solutions adopted in the embodiments of the present application further include: the inputting the clinical survival status information and prognosis-related genes into a statistical model for molecular marker screening includes:

Described statistical model is lasso punishment Cox proportional hazards model, described model adopts cross-validation to carry out the Lambda value operation of the set number of times, and determines the final Lambda value according to the minimum mean standard error value of all Lambda values, according to the final Lambda value Value screening of molecular markers of the prognosis-related genes.

The technical solution adopted in the embodiment of the present application further includes: the scoring of the patient according to the molecular marker and the gene expression level includes:

The regression coefficient of each molecular marker is multiplied by the gene expression value of the molecular marker in the patient to obtain the score value of each molecular marker in the patient, and then all the molecular markers of the patient are calculated. The scores of the subjects are added to obtain the final score of the patient.

The technical solution adopted in the embodiment of the present application further includes: the classification of tumor risk levels for the patient according to the score distribution includes:

Analyze the relationship between the distribution of the patient's score and the clinical survival status information, and use a statistical method to perform a maximum selection test, calculate the maximum selection log-rank statistic value, and use this value as a cut-off point to carry out the risk level of the patient. Divide.

Another technical solution adopted in the embodiment of the present application is: a tumor risk grade classification system, comprising:

Data acquisition module: used to acquire the patient's transcriptome data and clinical survival status information; wherein, the transcriptome data at least includes the gene expression value of each gene in the patient;

Prognosis-related gene acquisition module: used to obtain the prognosis-related genes of the patient according to the transcriptome data and clinical survival status information;

Molecular marker screening module: used to input the clinical survival status information and prognosis-related genes into a statistical model for molecular marker screening, and obtain the gene expression values of the screened molecular markers in the patient;

Risk grading module: used to score the patient according to the molecular marker and the gene expression value, and divide the patient according to the score distribution to tumor risk grading.

Another technical solution adopted by the embodiments of the present application is: a terminal, the terminal includes a processor and a memory coupled to the processor, wherein,

The memory stores program instructions for implementing the method for classifying tumor risk levels;

The processor is configured to execute the program instructions stored in the memory to control tumor risk grading.

Another technical solution adopted by the embodiments of the present application is: a storage medium storing program instructions executable by a processor, where the program instructions are used to execute the tumor risk level classification method.

beneficial effect

Compared with the prior art, the beneficial effects of the embodiments of the present application are: the tumor risk level classification method of the embodiments of the present application obtains the prognosis-related genes of the patients based on the transcriptome data of the patients, and on this basis, uses a statistical model to perform multiple Operation is performed to screen molecular markers of prognosis-related genes, and score each patient in combination with the screened molecular markers and gene expression levels, and divide the patients into risk levels according to the score distribution. Compared with the prior art, the embodiment of the present application can maximize the retention of all prognosis-related genes, so that the obtained prognosis-related genes have better accuracy, and the acquisition of molecular markers has a lower error rate, and the risk level classification results have better rationality.

Description of drawings

FIG. 1 is a flowchart of a method for classifying tumor risk levels according to the first embodiment of the present application;

FIG. 2 is a flowchart of a method for classifying tumor risk levels according to a second embodiment of the present application;

3 is a schematic structural diagram of a tumor risk grade classification system according to an embodiment of the present application;

FIG. 4 is a schematic structural diagram of a terminal according to an embodiment of the present application;

FIG. 5 is a schematic structural diagram of a storage medium according to an embodiment of the present application.

Embodiments of the present invention

In order to make the purpose, technical solutions and advantages of the present application more clearly understood, the present application will be described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application.

In order to solve the deficiencies of the prior art, the tumor risk grade classification method of the embodiment of the present application is based on the transcriptome data of all the study patients, combined with their clinical survival information, to obtain the prognosis-related genes of the patients, and on this basis, a statistical model is used to conduct multiple The second operation is performed to screen the molecular markers of prognosis-related genes, and score each patient in combination with the screened molecular markers and gene expression levels, and classify the patients according to the score distribution.

Specifically, please refer to FIG. 1 , which is a flowchart of the method for classifying tumor risk levels according to the first embodiment of the present application. The tumor risk grade classification method according to the first embodiment of the present application includes the following steps:

S100: Obtain transcriptome data of all patients of the same tumor type;

In this step, the transcriptome data is input in the form of a matrix, each row is a single patient individual, each column is a gene name, and the value of each cell in the matrix represents the gene expression level of each gene in each individual patient (that is, the gene expression level). expression value).

S110: Preprocess the transcriptome data to obtain the clinical survival status information of all patients;

In this step, the preprocessing operation specifically includes: first, using statistics to perform logarithmic transformation on the gene expression value of each gene in the transcriptome data, remove genes with low expression levels, and only keep genes that exceed the set ratio (this application The ratio is set to 90%, which can be set according to the actual operation), and there are expressed genes in the patients; then the clinical survival status information such as the ID number, survival status, and survival time of all patients is obtained from the transcriptome data; The clinical survival status information also exists in the form of a matrix. Each row of the matrix is the ID number of a single patient, and each column corresponds to the survival status and survival time of each patient.

S120: respectively obtain the high and low expression populations corresponding to each gene according to the gene expression level;

In this step, the algorithms for obtaining high and low gene expression populations include, but are not limited to, the method of dichotomy, the method of thirds, and the method of quartering, etc. Each obtaining algorithm is specifically:

1) Dichotomy: sort the gene expression values of each gene in all patients, and take the mean or median of all gene expression values as the dividing point to obtain the high and low expression population of each gene;

2) Trichotomy: Rank the gene expression value of each gene in all patients, and intercept the top 1/3 population and the bottom 1/3 population as the high and low expression population of the corresponding gene. ;

3) Quartile method: Rank the gene expression values of each gene in all patients, take the quartile as the cutoff point, and intercept the highest quartile and the lowest quartile as the high and low expression of the corresponding gene. crowd.

S130: Combine the clinical survival status information and the high and low expression population of each gene to calculate the correlation between the expression level of each gene and prognosis, obtain the prognosis-related genes, and extract the gene expression values of each prognosis-related gene in all patients , generate a prognostic-associated gene expression matrix;

Among them, in the embodiment of the present application, the statistical computing platform R uses the survival tool to calculate the prognosis correlation of all genes, and outputs the regression coefficient and P value of each gene. The threshold is 0.01, which can be set according to the actual operation) as a significant correlation condition to filter all genes to obtain prognosis-related genes. Since the high and low expression populations of genes obtained by different algorithms are different, the obtained prognosis-related genes are also different. Therefore, in the actual operation, the embodiments of the present application respectively calculate the high and low expression populations obtained based on different algorithms. If the high and low expression populations of at least two algorithms are associated with prognosis, the gene is determined as a prognosis-associated gene.

S140: Based on the clinical survival status information and the prognosis-related gene expression matrix, molecular markers are screened for the prognosis-related genes through a statistical model, and the gene expression values of each molecular marker in each patient are obtained separately, and each molecular marker is generated. The molecular marker expression matrix of the marker;

In this step, the lasso penalty Cox proportional hazards statistical model is used to screen molecular markers. The model uses cross-validation to select a statistic Lambda value, and the corresponding molecular marker is determined according to the Lambda value. In the calculation process, the Lambda value obtained by each operation of the model is different, resulting in a great randomness of the molecular markers to be screened. Therefore, in the embodiment of the present application, the model cycle is used for a set number of times (for example, 100 times or 1000 times). , which can be set according to the actual operation), and determine the final Lambda value according to the minimum average standard error value of all Lambda values, screen the molecular markers of prognosis-related genes according to the Lambda value, and then extract all molecular markers The gene expression values of the markers in all patients were made into a molecular marker expression matrix.

S150: Calculate the score value of each molecular marker in each patient according to the regression coefficient of each molecular marker and the gene expression value in each patient, and then add all the score values of each patient to obtain final score for each patient;

In this step, first, the regression coefficient of each molecular marker is multiplied by the gene expression value in each patient to obtain the score value of each molecular marker in each patient; The scores for the markers were added together to obtain the final score for each patient.

S160: Combining the high and low score distribution of each patient and the clinical survival status information to classify tumor patients into high and low risk levels;

In this step, the survminer tool is used to classify the risk levels through the statistical computing platform R. First, the scores of each patient are sorted according to their high and low levels, and combined with the clinical survival status information, the high and low score distribution and clinical survival status information of each patient are checked. Then use the statistical method to carry out the maximum choice test, and calculate the maximum choice log-rank statistic value, and use this value as the cut-off point to classify patients with different risk levels; and use the survival tool to evaluate the risk level classification results. , according to the size of the P value to determine whether there is a significant difference in the survival probability of high and low risk levels, if there is a significant difference, it indicates that the risk level classification results are reasonable.

Based on the above, the examples of the present application use the transcriptome data of all patients of the same tumor type to screen for prognosis-related genes. Since the transcriptome data includes the expression levels of all genes in the patient, it can maximize the retention of all prognosis-related genes without missing Select other potential prognostic associated genes. In the calculation of prognosis-related genes, the influence of different gene high-low population acquisition methods on the final prognosis-related gene judgment is fully considered, and the prognosis-related gene results obtained by common high-low population group acquisition methods are comprehensively considered, so that the obtained prognosis-related genes have better accuracy. The lasso penalized Cox proportional hazards model was used for multiple operations, and the final Lambda value was determined by the minimum average standard error value of the calculation results, so as to realize the screening of prognosis-related genes, and to obtain molecular markers with a lower error rate. Finally, based on the distribution of patient scores, statistical methods are used to automatically classify patients into high and low risk levels, respecting the true distribution of the data and having better rationality.

Please refer to FIG. 2 , which is a flowchart of a method for classifying tumor risk levels according to the second embodiment of the present application. This embodiment is to apply the present application to the classification of risk levels of melanoma patients, and specifically includes the following steps:

S200: Download transcriptome data of melanoma patients from ICGC database;

S210: Preprocess the transcriptome data, perform log2 logarithmic transformation on the gene expression value of each gene in the transcriptome data, retain the genes that are expressed in more than 90% of the patients, and obtain the ID numbers, survival numbers of all patients Clinical survival status information such as status and survival time;

S220: Use the survival tool to calculate the prognostic association of all genes through the statistical computing platform R, output the regression coefficient and P value of each gene, and filter all genes according to the P value <0.01 threshold as a significant correlation condition, and get 1086 prognosis Associate genes, and extract the gene expression values of each prognosis-associated gene in all patients to generate a prognosis-associated gene expression matrix;

S230: Combine the clinical survival status information and the prognosis-related gene expression matrix, use the glmnet tool in the statistical computing platform R to run 100 times in a loop to calculate the minimum average standard error value to determine the final Lambda value, and screen out 21 based on the Lambda value. molecular markers;

S240: Extract the gene expression values of each molecular marker in all patients, make a molecular marker expression matrix, and calculate each molecule according to the regression coefficient of each molecular marker and the gene expression value in each patient. The score of the marker in each patient, according to the score of each patient, a statistical method is used to classify the melanoma patients into high and low risk grades;

Among them, the survival tool was used in the statistical computing platform R to evaluate the risk level classification results of melanoma patients, and the P value was less than 0.001, indicating that there were significant differences between different risk levels, that is, the risk level classification results were very reasonable. .

Please refer to FIG. 3 , which is a schematic structural diagram of a tumor risk grade classification system according to an embodiment of the present application. The tumor risk level classification system 40 according to the embodiment of the present application includes:

Data acquisition module 41: used to acquire transcriptome data and clinical survival status information of the patient; wherein, the transcriptome data at least includes the gene expression value of each gene in the patient;

Prognosis-related gene acquisition module 42: used to acquire the prognosis-related genes of the patient according to the transcriptome data and the clinical survival state information;

Molecular marker screening module 43: used to input the clinical survival status information and prognosis-related genes into a statistical model for molecular marker screening, and obtain the gene expression values of the screened molecular markers in the patient;

Risk level classification module 44: used to score the patient according to the molecular marker and the gene expression value, and classify the patient according to the score distribution to tumor risk level.

Please refer to FIG. 4 , which is a schematic structural diagram of a terminal according to an embodiment of the present application. The terminal 50 includes a processor 51 and a memory 52 coupled to the processor 51 .

The memory 52 stores program instructions for implementing the above-described tumor risk level classification method.

The processor 51 is configured to execute program instructions stored in the memory 52 to control the classification of tumor risk levels.

The processor 51 may also be referred to as a CPU (Central Processing Unit, central processing unit). The processor 51 may be an integrated circuit chip with signal processing capability. The processor 51 may also be a general purpose processor, digital signal processor (DSP), application specific integrated circuit (ASIC), off-the-shelf programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components . A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

Please refer to FIG. 5 , which is a schematic structural diagram of a storage medium according to an embodiment of the present application. The storage medium of this embodiment of the present application stores a program file 61 capable of implementing all the above methods, wherein the program file 61 may be stored in the above-mentioned storage medium in the form of a software product, and includes several instructions to enable a computer device (which can be It is a personal computer, a server, or a network device, etc.) or a processor (processor) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only). Memory), random access memory (RAM, Random Various media that can store program codes, such as Access Memory), magnetic disks or CD-ROMs, or terminal devices such as computers, servers, mobile phones, and tablets.

The above description of the disclosed embodiments enables any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined in this application may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

A method for classifying tumor risk levels, comprising the following steps:

Obtaining the patient's transcriptome data and clinical survival status information; wherein, the transcriptome data at least includes the gene expression value of each gene in the patient;

Obtain the prognosis-related genes of the patient according to the transcriptome data and the clinical survival status information;

Inputting the clinical survival status information and prognosis-related genes into a statistical model for molecular marker screening, and obtaining the gene expression values of the screened molecular markers in the patient;

The patient is scored according to the molecular marker and the gene expression value, and the patient is divided into tumor risk levels according to the score distribution.
The method for classifying tumor risk levels according to claim 1, wherein the obtaining of the patient's transcriptome data and clinical survival status information comprises:

The gene expression value of each gene in the transcriptome data was logarithmically transformed by statistics, the genes with low expression levels were eliminated, and the genes that were expressed in more than a set proportion of patients were retained; Obtain the clinical survival status information of the patient from the data;

The clinical survival status information includes patient ID number, survival status and survival time.
The method for classifying tumor risk levels according to claim 2, wherein the obtaining of the patient's prognosis-related genes according to the transcriptome data and clinical survival status information comprises:

Sort the gene expression values of each gene in the patient, and obtain the high and low expression populations corresponding to each gene respectively;

Calculate the correlation between the expression level of each gene and prognosis in combination with the clinical survival status information and the high and low expression population, output the regression coefficient and P value of each gene, and filter all the genes of the patient according to the size of the P value , obtain the prognosis-related genes of the patient, and extract the gene expression value of each prognosis-related gene in the patient to generate a prognosis-related gene expression matrix.
The method for classifying tumor risk levels according to claim 3, wherein the sorting of the gene expression values of each gene in the patient, and obtaining the high and low expression populations corresponding to each gene respectively comprises:

The high and low expression populations corresponding to each gene were obtained by the method of dichotomy, trichotomy, or quartile, respectively.
The method for classifying tumor risk levels according to claim 4, wherein the obtaining of the patient's prognosis-related genes according to the transcriptome data and clinical survival status information comprises:

Based on the high and low expression populations obtained by different algorithms, the relationship between the expression level and prognosis is calculated respectively. If a gene is associated with prognosis in the high and low expression populations obtained by at least two algorithms, the gene is determined as a prognosis. associated genes.
The method for classifying tumor risk levels according to claim 3, wherein the inputting the clinical survival status information and prognosis-related genes into a statistical model for molecular marker screening comprises:

Described statistical model is lasso punishment Cox proportional hazards model, described model adopts cross-validation to carry out the Lambda value operation of the set number of times, and determines the final Lambda value according to the minimum mean standard error value of all Lambda values, according to the final Lambda value Value screening of molecular markers of the prognosis-related genes.
The method for classifying tumor risk levels according to claim 6, wherein the scoring of the patient according to the molecular marker and the gene expression level comprises:

The regression coefficient of each molecular marker is multiplied by the gene expression value of the molecular marker in the patient to obtain the score value of each molecular marker in the patient, and then all the molecular markers of the patient are calculated. The scores of the subjects are added to obtain the final score of the patient.
The method for classifying tumor risk levels according to any one of claims 1 to 7, wherein the performing tumor risk level classification on the patient according to the score distribution comprises:

Analyze the relationship between the distribution of the patient's score and the clinical survival status information, and use a statistical method to perform a maximum selection test, calculate the maximum selection log-rank statistic value, and use this value as a cut-off point to carry out the risk level of the patient. Divide.
A tumor risk grading system, characterized in that it includes:

Data acquisition module: used to acquire the patient's transcriptome data and clinical survival status information; wherein, the transcriptome data at least includes the gene expression value of each gene in the patient;

Prognosis-related gene acquisition module: used to obtain the prognosis-related genes of the patient according to the transcriptome data and clinical survival status information;

Molecular marker screening module: used to input the clinical survival status information and prognosis-related genes into a statistical model for molecular marker screening, and obtain the gene expression values of the screened molecular markers in the patient;

Risk grading module: used to score the patient according to the molecular marker and the gene expression value, and divide the patient according to the score distribution to tumor risk grading.
A terminal, characterized in that the terminal includes a processor and a memory coupled to the processor, wherein,

The memory stores program instructions for implementing the method for classifying tumor risk levels according to any one of claims 1-8;

The processor is configured to execute the program instructions stored in the memory to control tumor risk grading.
A storage medium, characterized in that it stores program instructions executable by a processor, and the program instructions are used to execute the method for classifying tumor risk levels according to any one of claims 1 to 8.