A method and system for constructing a prognostic model for acute myeloid leukemia

Abstract

The present invention belongs to the technical field of bioinformatics, and specifically relates to a method and system for constructing a prognostic model for acute myeloid leukemia. After performing differential expression analysis to obtain common differentially expressed genes with smaller batch effects, weighted gene co-expression network analysis is performed to identify modules with high correlation with clinical traits as key modules. Single factor analysis, regression algorithm analysis, and stepwise regression analysis are performed in sequence to analyze the effects of gene expression levels in the key modules on survival time, and 7 target prognostic markers, namely CD2, LRG1, ANGPT1, CSTB, ST6GALNAC4, LSP1, and AC135279.2, are obtained to construct a prognostic model for acute myeloid leukemia, which can more accurately predict the survival of patients with acute myeloid leukemia.

Description

Technical Field

[0001] The invention belongs to the technical field of bioinformatics, and in particular relates to a method and system for constructing a prognosis model for acute myeloid leukemia. Background Art

[0002] To date, studies have used weighted gene co-expression network analysis (WGCNA) to identify disease-related biological modules in acute myeloid leukemia (AML) and discovered many potential markers. These markers may be involved in a variety of biological processes such as AML stem cell proliferation and differentiation, apoptosis, immune response, and drug sensitivity. The process of screening prognostic markers aims to identify more new markers in order to use molecular information to improve the prognosis of acute myeloid leukemia.

[0003] However, despite the great potential of WGCNA in AML research, there are still some challenges. For example, traditional prognostic marker identification methods may have problems such as insufficient accuracy, low sensitivity and specificity, resulting in unreliable prognostic prediction. In addition, the small sample size and insufficient statistical power in many studies have led to poor applicability and stability of the identified markers in a wider range of populations. In addition, the limited leukemia transcriptome datasets in the database have led to many people repeatedly using the same dataset, and the results produced by the experiments lack innovation. Summary of the invention

[0004] In order to solve the problems raised by the background technology, the present invention provides a method and system for constructing a prognosis model for acute myeloid leukemia.

[0005] The technical solution of the present invention is as follows:

[0006] The present invention provides a method for constructing a prognostic model for acute myeloid leukemia, comprising the following steps:

[0007] S1: Collect bone marrow samples from patients with acute myeloid leukemia, calculate transcript expression, and construct the first data set;

[0008] Obtain transcriptome data and clinical data of acute myeloid leukemia patients in the database to construct the second data set;

[0009] Obtain transcriptome data of healthy people in the database to construct the third data set;

[0010] S2: After differential expression analysis, the common differentially expressed gene set was obtained;

[0011] S3: The common differentially expressed gene sets were analyzed through weighted gene co-expression network analysis to obtain several key modules, specifically:

[0012] Based on the shared differentially expressed gene set, the Pearson correlation coefficient between the shared differentially expressed genes is calculated. According to the Pearson correlation coefficient between the shared differentially expressed genes, a weighted adjacency matrix is ​​constructed. After the weighted adjacency matrix is ​​converted into a topological overlap matrix, the dissimilarity of the topological overlap matrix is ​​calculated. According to the dissimilarity, the shared differentially expressed genes are clustered, and a gene-module vector clustering diagram is constructed to obtain several modules and characteristic genes in each module.

[0013] The Pearson correlation coefficient between the characteristic genes in each module and the clinical data was calculated, and the modules where the characteristic genes with Pearson correlation coefficients greater than the Pearson correlation coefficient threshold and P values ​​less than the first P value threshold were located were recorded as key modules. After traversing all modules, several key modules were obtained;

[0014] S4: After univariate analysis of genes in several key modules and calculation of the risk ratio and preset confidence interval of each gene, the gene with a P value less than the second P value threshold is recorded as the first prognostic marker;

[0015] S5: The first prognostic marker is analyzed by regression algorithm, and the gene whose likelihood deviation is less than the likelihood deviation threshold is recorded as the second prognostic marker;

[0016] The second prognostic marker is subjected to stepwise regression analysis to obtain the target prognostic marker and the weight coefficient of the target prognostic marker;

[0017] S6: Construct a prognostic model for acute myeloid leukemia based on the target prognostic marker and the weight coefficient of the target prognostic marker.

[0018] In step S2, a common differentially expressed gene set is obtained through differential expression analysis, specifically:

[0019] After the first data set was merged with the third data set, differential expression analysis was performed using edgeR, limma, and DESeq2, respectively, to obtain the first up-regulated gene set and the first down-regulated gene set, the second up-regulated gene set and the second down-regulated gene set, and the third up-regulated gene set and the third down-regulated gene set. The first up-regulated differentially expressed gene set was obtained by taking the intersection of the first up-regulated gene set, the second up-regulated gene set, and the third up-regulated gene set. The first down-regulated differentially expressed gene set was obtained by taking the intersection of the first down-regulated gene set, the second down-regulated gene set, and the third down-regulated gene set.

[0020] After the second data set was merged with the third data set, differential expression analysis was performed using edgeR, limma, and DESeq2, respectively, to obtain the fourth up-regulated gene set and the fourth down-regulated gene set, the fifth up-regulated gene set and the fifth down-regulated gene set, and the sixth up-regulated gene set and the sixth down-regulated gene set. The intersection of the fourth up-regulated gene set, the fifth up-regulated gene set, and the sixth up-regulated gene set was used to obtain the second up-regulated differentially expressed gene set. The intersection of the fourth down-regulated gene set, the fifth down-regulated gene set, and the sixth down-regulated gene set was used to obtain the second down-regulated differentially expressed gene set.

[0021] The first up-regulated differentially expressed gene set, the second up-regulated differentially expressed gene set, the first down-regulated differentially expressed gene set, and the second down-regulated differentially expressed gene set are merged to obtain a common differentially expressed gene set.

[0022] The target prognostic markers in step S5 are CD2, LRG1, ANGPT1, CSTB, ST6GALNAC4, LSP1, and AC135279.2.

[0023] The second prognostic marker in step S5 is CD2, AL359715.1, ANGPT1, ST6GALNAC4, LSP1, PLEKHA5, AC135279.2, SELPLG, GZMB, ISG20, LRG1, RTF2, ZBP1, CSTB and DNASE1L1.

[0024] In step S3, a weighted adjacency matrix is ​​constructed according to the Pearson correlation coefficients between the shared differentially expressed genes, specifically:

[0025] According to the formula: , construct a weighted adjacency matrix;

[0026] In the formula, , They are two different genes in the common differentially expressed set; is the weighted adjacency matrix; For gene and genes The Pearson correlation coefficient matrix between is the soft threshold.

[0027] In step S3, the differentially expressed genes are clustered to construct a gene-module vector clustering diagram, specifically:

[0028] The nodes in the vector clustering diagram represent genes, different colors represent different modules, and the connections between nodes represent the co-expression relationship between genes.

[0029] In step S1, a bone marrow sample of an acute myeloid leukemia patient is collected and the transcript expression level is calculated, specifically:

[0030] Bone marrow samples were collected from patients with acute myeloid leukemia, and after pretreatment, they were sequenced to obtain sequencing data. The sequencing data were compared with the human genome to obtain comparison data. After deduplication processing, the transcript expression level was calculated.

[0031] The clinical data in step S1 include age, race, gender, tumor history, treatment history, and treatment methods.

[0032] In step S6, based on the target prognostic marker and the weight coefficient of the target prognostic marker, an acute myeloid leukemia prognostic model is constructed, specifically:

[0033] The expression level of each target prognostic marker and the corresponding weight coefficient of the target prognostic marker were multiplied to obtain the score of each target prognostic marker. The scores of all target prognostic markers were added together to obtain the risk score. The survival rate was predicted based on the risk score, and a prognostic model for acute myeloid leukemia was constructed.

[0034] The present invention also provides a system for constructing a prognostic model for acute myeloid leukemia, comprising:

[0035] Dataset construction module: used to collect bone marrow samples from patients with acute myeloid leukemia, calculate transcript expression, and construct the first dataset;

[0036] Obtain transcriptome data and clinical data of acute myeloid leukemia patients in the database to construct the second data set;

[0037] Obtain transcriptome data of healthy people in the database to construct the third data set;

[0038] Differential expression analysis module: used to obtain the common differentially expressed gene set through differential expression analysis;

[0039] Key module division module: used to analyze the weighted gene co-expression network of the common differentially expressed gene sets to obtain several key modules, specifically:

[0040] Based on the shared differentially expressed gene set, the Pearson correlation coefficient between the shared differentially expressed genes is calculated. According to the Pearson correlation coefficient between the shared differentially expressed genes, a weighted adjacency matrix is ​​constructed. After the weighted adjacency matrix is ​​converted into a topological overlap matrix, the dissimilarity of the topological overlap matrix is ​​calculated. According to the dissimilarity, the shared differentially expressed genes are clustered, and a gene-module vector clustering diagram is constructed to obtain several modules and characteristic genes in each module.

[0041] The Pearson correlation coefficient between the characteristic genes in each module and the clinical data was calculated, and the modules where the characteristic genes with Pearson correlation coefficients greater than the Pearson correlation coefficient threshold and P values ​​less than the first P value threshold were located were recorded as key modules. After traversing all modules, several key modules were obtained;

[0042] The first prognostic marker screening module: used for the genes in several key modules to be subjected to univariate analysis, calculate the risk ratio and preset confidence interval of each gene, and record the gene with a P value less than the second P value threshold as the first prognostic marker;

[0043] Target prognostic marker screening module: used for analyzing the first prognostic marker through regression algorithm, and recording the gene with likelihood deviation less than the likelihood deviation threshold as the second prognostic marker;

[0044] The second prognostic marker is subjected to stepwise regression analysis to obtain the target prognostic marker and the weight coefficient of the target prognostic marker;

[0045] Acute myeloid leukemia prognostic model construction module: construct an acute myeloid leukemia prognostic model based on the target prognostic marker and the weight coefficient of the target prognostic marker.

[0046] Beneficial effect: The present invention increases the complexity of the acquired data, performs differential expression analysis on the complex data, obtains the common differentially expressed genes with smaller batch effects, and then performs weighted gene co-expression network analysis to identify the modules with high correlation with clinical traits as key modules. The effects of gene expression levels in key modules on survival time are analyzed in sequence through univariate analysis, regression algorithm analysis, and stepwise regression analysis. Seven target prognostic markers, CD2, LRG1, ANGPT1, CSTB, ST6GALNAC4, LSP1, and AC135279.2, are obtained to construct a prognostic model for acute myeloid leukemia, which can more accurately predict the survival of patients with acute myeloid leukemia, identify high-risk patients, adjust treatment strategies in time, and improve the overall survival rate of patients.

[0047] The present invention introduces and processes bone marrow samples from acute myeloid leukemia patients in the hospital to increase the sample size, and uses complex clinical data to avoid the statistical power deficiency caused by a small number of samples, thereby enhancing the applicability and stability of the target prognostic markers in a wider population and improving the accuracy, sensitivity and specificity of acute myeloid leukemia prognosis.

[0048] The present invention obtains two categories of differentially expressed genes by performing differential expression analysis on the third data set (healthy population sample data) and the first data set and the second data set (transcriptome sample data of acute myeloid leukemia patients), and then takes the intersection of the two categories of differentially expressed genes. The obtained common differentially expressed genes are differential genes with smaller batch effects, thereby reducing experimental errors and batch effects.

[0049] The present invention divides modules according to the correlation between genes, calculates the Pearson correlation coefficient between the characteristic genes in each module and the clinical data, identifies the modules with high correlation with clinical traits as key modules, and then performs subsequent analysis on the genes in the key modules, which can more comprehensively reflect the correlation between the modules and various clinical traits, and further enhance the reliability of prognosis prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 The figure is a flow chart of a method for constructing a prognostic model for acute myeloid leukemia in an embodiment. DETAILED DESCRIPTION

[0051] The following examples are intended to illustrate the present invention rather than to further limit the present invention.

[0052] The present invention provides a method for constructing a prognostic model for acute myeloid leukemia, such as Figure 1 As shown, the following steps are included:

[0053] S1: Collect bone marrow samples from patients with acute myeloid leukemia, calculate transcript expression, and construct the first data set.

[0054] Bone marrow samples were collected from 41 patients with acute myeloid leukemia diagnosed by the hospital's hematology and oncology department, and the bone marrow samples were pretreated first. Specifically, first, a work area was set up in a biosafety cabinet to ensure sterile conditions. Then, cells were extracted from the bone marrow samples of patients with acute myeloid leukemia, and cell pellets were obtained by centrifugation. Subsequently, the cell pellets were centrifuged in a high-speed low-temperature centrifuge to separate the supernatant and the cell pellet. The supernatant was accurately collected using a pipette containing a cell lysis solution in which RNA had been dissolved. The supernatant was mixed with a specialized RNA extraction reagent and reacted in a thermostatic mixer to ensure that the appropriate temperature was maintained to promote RNA dissolution. The mixed sample was then shaken for a short time with a vortex shaker to ensure that the reagent was fully mixed with the cellular components. The sample was briefly centrifuged with a handheld centrifuge to remove cell debris and other impurities for subsequent testing.

[0055] Preferably, in step S1, a bone marrow sample is collected from an acute myeloid leukemia patient, and the transcript expression level is calculated, specifically:

[0056] Bone marrow samples were collected from patients with acute myeloid leukemia, and after pretreatment, they were sequenced to obtain sequencing data. The sequencing data were compared with the human genome to obtain comparison data. After deduplication processing, the transcript expression level was calculated.

[0057] For example, after preprocessing, 96RXN was used to prepare cDNA libraries, and sequencing was performed on the Illumina NovaSeq 6000Dx-CN sequencing platform to obtain sequencing data RNA-seq data; STAR was used to align RNA-seq data with the hg19 version of the human genome to generate BAM alignment file data; Picard tools were used to deduplicate BAM alignment file data; and the featureCount component of the subread toolkit was used to calculate transcript expression. In addition, the AnnoProbe package in R language was used to annotate gene names in a collection format.

[0058] The present invention introduces bone marrow samples from acute myeloid leukemia patients in the hospital, performs transcriptome sequencing, alignment, deduplication and calculates its expression level, organizes it into a new data set, and then combines it with other public data sets to ensure that the target prognostic markers obtained by subsequent screening can more accurately prognose acute myeloid leukemia and enhance the reliability of prognosis prediction.

[0059] The transcriptome data and clinical data of acute myeloid leukemia patients in the database were obtained to construct the second data set.

[0060] For example, we downloaded transcriptome data and clinical data of patients with acute myeloid leukemia from the TCGA database. We merged the transcriptome data with the clinical data, removed samples lacking clinical data, and finally obtained transcriptome data samples of 129 patients with acute myeloid leukemia. Among them, there were 58 female patients aged between 18 and 88 years old.

[0061] Preferably, the clinical data in step S1 include age, race, gender, tumor history, treatment history, and treatment methods.

[0062] The present invention uses complex clinical data and is not limited to a certain clinical trait, so as to facilitate subsequent analysis, more comprehensively reflect the correlation between the module and multiple clinical traits, and further enhance the reliability of prognosis prediction.

[0063] The transcriptome data of healthy people in the database were obtained to construct the third data set.

[0064] For example, as a control sample, transcriptome data of 200 healthy people were taken from the GTEx database.

[0065] In terms of data, the present invention introduces bone marrow samples from acute myeloid leukemia patients in the hospital to increase the sample size, and uses complex clinical data to avoid the statistical power deficiency caused by a small number of samples, thereby enhancing the applicability and stability of the target prognostic markers in a wider population, and improving the accuracy, sensitivity and specificity of acute myeloid leukemia prognosis.

[0066] S2: After differential expression analysis, the common differentially expressed gene set was obtained.

[0067] Differential expression analysis uses three commonly used differential expression analysis tools, namely edgeR, limma, and DESeq2, which are used for the analysis of high-throughput gene expression data such as RNA-Seq. Among them, edgeR processes RNA-Seq count data based on the negative binomial distribution model, which is suitable for differential expression analysis with low sample sizes. limma (microarray analysis of linear models) is combined with the voom function to post-process RNA-Seq data. The voom function converts RNA-Seq count data into weighted normal distribution data suitable for linear models, so as to perform differential expression analysis using linear models. DESeq2 is also based on the negative binomial distribution and uses a local regression method to estimate the dispersion of gene expression. It is particularly suitable for processing small sample RNA-Seq data. By performing a variance stabilization transformation on the count data, the accuracy of differential expression analysis is improved.

[0068] Specifically:

[0069] After the first data set was merged with the third data set, differential expression analysis was performed using edgeR, limma, and DESeq2, respectively, to obtain the first up-regulated gene set and the first down-regulated gene set, the second up-regulated gene set and the second down-regulated gene set, and the third up-regulated gene set and the third down-regulated gene set. The first up-regulated differentially expressed gene set was obtained by taking the intersection of the first up-regulated gene set, the second up-regulated gene set, and the third up-regulated gene set. The first down-regulated differentially expressed gene set was obtained by taking the intersection of the first down-regulated gene set, the second down-regulated gene set, and the third down-regulated gene set.

[0070] After the second data set was merged with the third data set, differential expression analysis was performed using edgeR, limma, and DESeq2, respectively, to obtain the fourth up-regulated gene set and the fourth down-regulated gene set, the fifth up-regulated gene set and the fifth down-regulated gene set, and the sixth up-regulated gene set and the sixth down-regulated gene set. The intersection of the fourth up-regulated gene set, the fifth up-regulated gene set, and the sixth up-regulated gene set was used to obtain the second up-regulated differentially expressed gene set. The intersection of the fourth down-regulated gene set, the fifth down-regulated gene set, and the sixth down-regulated gene set was used to obtain the second down-regulated differentially expressed gene set.

[0071] The first up-regulated differentially expressed gene set, the second up-regulated differentially expressed gene set, the first down-regulated differentially expressed gene set, and the second down-regulated differentially expressed gene set are merged to obtain a common differentially expressed gene set.

[0072] For example, after the first data set was merged with the third data set, a total of 35,672 genes were obtained. After differential expression analysis by edgeR, 5,797 first up-regulated genes and 3,536 first down-regulated genes were obtained; after differential expression analysis by limma, 4,021 second up-regulated genes and 3,986 second down-regulated genes were obtained; after differential expression analysis by DESeq2, 6,694 third up-regulated genes and 3,085 third down-regulated genes were obtained; after the intersection of 5,797 first up-regulated genes, 4,021 second up-regulated genes and 6,694 third up-regulated genes, the first up-regulated differentially expressed gene set was obtained; after the intersection of 3,536 first down-regulated genes, 3,986 second down-regulated genes and 3,085 third down-regulated genes, the first down-regulated differentially expressed gene set was obtained.

[0073] After the second data set was merged with the third data set, 7607 fourth up-regulated genes and 6125 fourth down-regulated genes were obtained after differential expression analysis by edgeR; and 6021 fifth up-regulated genes and 6374 fifth down-regulated genes were obtained after differential expression analysis by limma; and 8661 sixth up-regulated genes and 5290 sixth down-regulated genes were obtained after differential expression analysis by DESeq2; the second up-regulated differentially expressed gene set was obtained by taking the intersection of 7607 fourth up-regulated genes, 6021 fifth up-regulated genes and 8661 sixth up-regulated genes; the second down-regulated differentially expressed gene set was obtained by taking the intersection of 6125 fourth down-regulated genes, 6374 fifth down-regulated genes and 5290 sixth down-regulated genes.

[0074] The first up-regulated differentially expressed gene set and the second up-regulated differentially expressed gene set were merged to obtain 926 commonly up-regulated differentially expressed genes; the first down-regulated differentially expressed gene set and the second down-regulated differentially expressed gene set were merged to obtain 1088 commonly down-regulated differentially expressed genes. The commonly up-regulated differentially expressed genes and the commonly down-regulated differentially expressed genes constitute the common differentially expressed gene set.

[0075] The present invention obtains two categories of differentially expressed genes by performing differential expression analysis on the third data set (healthy population sample data) and the first data set and the second data set (transcriptome sample data of acute myeloid leukemia patients), and then takes the intersection of the two categories of differentially expressed genes. The obtained common differentially expressed genes are differential genes with smaller batch effects, thereby reducing experimental errors and batch effects.

[0076] S3: The common differentially expressed gene sets were analyzed through weighted gene co-expression network analysis to obtain several key modules, specifically:

[0077] Based on the shared differentially expressed gene set, the Pearson correlation coefficient between the shared differentially expressed genes was calculated. According to the Pearson correlation coefficient between the shared differentially expressed genes, a weighted adjacency matrix was constructed. After the weighted adjacency matrix was converted into a topological overlap matrix, the dissimilarity of the topological overlap matrix was calculated. According to the dissimilarity, the shared differentially expressed genes were clustered, and a gene-module vector clustering diagram was constructed to obtain several modules and the characteristic genes in each module. The characteristic genes in each module are the most representative genes in the module, indicating the highest correlation with other modules.

[0078] The Pearson correlation coefficient between the characteristic genes in each module and the clinical data was calculated. The modules where the characteristic genes with a Pearson correlation coefficient greater than the Pearson correlation coefficient threshold and a P value less than the first P value threshold were located were recorded as key modules. After traversing all modules, several key modules were obtained.

[0079] In order to better measure the correlation between genes and to facilitate the classification of genes with high correlation into the same module, preferably, in step S3, a weighted adjacency matrix is ​​constructed according to the Pearson correlation coefficient between the common differentially expressed genes, specifically:

[0080] According to the formula: , construct a weighted adjacency matrix;

[0081] In the formula, , They are two different genes in the common differentially expressed set; is the weighted adjacency matrix; For gene and genes The Pearson correlation coefficient matrix between them represents the adjacency relationship between genes; is a soft threshold, and different soft thresholds are used to reconstruct the weighted gene co-expression network (WGCNA) to determine the average connectivity of the WGCNA. =0.9 as the reconstruction threshold of the network.

[0082] Preferably, in order to visualize the results for subsequent analysis of the correlation between modules, the correlation between genes, and the correlation between modules and clinical traits, in step S3, the common differentially expressed genes are clustered to construct a gene-module vector clustering diagram, specifically:

[0083] The nodes in the vector clustering diagram represent genes, different colors represent different modules, and the connections between nodes represent the co-expression relationship between genes.

[0084] For example, 15 modules were obtained, labeled black (69), blue (224), brown (209), cyan (20), green (141), green-yellow (51), magenta (57), pink (68), purple (53), red (73), salmon (41), tan (44), turquoise (306), and yellow (156), while genes that did not fit into any specific module (488) were classified into the gray modules. The numbers provided in brackets indicate the gene counts within each module.

[0085] Furthermore, by calculating the Pearson correlation coefficient between the characteristic genes in each module and the clinical data, the modules with high correlation with clinical traits were identified. For example, the module containing the characteristic genes with a Pearson correlation coefficient greater than 0.25 and a p value less than 0.05 was recorded as a key module, indicating that the module was significantly correlated with the clinical trait.

[0086] The present invention divides modules according to the correlation between genes, calculates the Pearson correlation coefficient between the characteristic genes in each module and the clinical data, identifies the modules with high correlation with clinical traits as key modules, and then performs subsequent analysis on the genes in the key modules, which can more comprehensively reflect the correlation between the modules and various clinical traits, and further enhance the reliability of prognosis prediction.

[0087] S4: After the genes in several key modules are subjected to univariate analysis, such as univariate Cox analysis, the risk ratio and the preset confidence interval of each gene, which can be a 95% confidence interval, are calculated, and the genes with a P value less than the second P value threshold are recorded as the first prognostic marker.

[0088] To study the effect of gene expression levels in key modules on survival time, univariate Cox analysis was used to calculate the hazard ratio (HR) and 95% confidence interval (CI) of each gene in the key modules screened by WGCNA to quantify its effect on patient prognosis. For example, genes with a P value less than 0.01 are considered to have expression levels that may affect the survival time of patients with acute myeloid leukemia.

[0089] S5: The first prognostic marker is analyzed by a regression algorithm, such as a LASSO regression algorithm, and the gene whose likelihood deviation is less than the likelihood deviation threshold is recorded as the second prognostic marker.

[0090] Preferably, the second prognostic marker in step S5 is CD2, AL359715.1, ANGPT1, ST6GALNAC4, LSP1, PLEKHA5, AC135279.2, SELPLG, GZMB, ISG20, LRG1, RTF2, ZBP1, CSTB and DNASE1L1.

[0091] The second prognostic marker is subjected to stepwise regression analysis, such as Cox stepwise regression analysis, to obtain the target prognostic marker and the weight coefficient of the target prognostic marker.

[0092] Preferably, the target prognostic markers in step S5 are CD2, LRG1, ANGPT1, CSTB, ST6GALNAC4, LSP1, and AC135279.2.

[0093] The present invention uses CD2, LRG1, ANGPT1, CSTB, ST6GALNAC4, LSP1, and AC135279.2 as target prognostic markers, which are significantly correlated with acute myeloid leukemia.

[0094] S6: Construct a prognostic model for acute myeloid leukemia based on the target prognostic marker and the weight coefficient of the target prognostic marker.

[0095] Preferably, in step S6, an acute myeloid leukemia prognosis model is constructed based on the target prognostic marker and the weight coefficient of the target prognostic marker, specifically:

[0096] The expression level of each target prognostic marker and the corresponding weight coefficient of the target prognostic marker were multiplied to obtain the score of each target prognostic marker. The scores of all target prognostic markers were added together to obtain the risk score. The survival rate was predicted based on the risk score, and a prognostic model for acute myeloid leukemia was constructed.

[0097] For example, the risk scoring formula is as follows:

[0098] ;

[0099] In the formula, is the risk score; 0.0074, 0.0221, 0.0102, 0.0203, 0.0186, 0.0043, -0.0396 are the weight coefficients of CD2, LRG1, ANGPT1, CSTB, ST6GALNAC4, LSP1, AC135279.2 target prognostic markers, respectively; , , , , , , They are the expression levels of target prognostic markers CD2, LRG1, ANGPT1, CSTB, ST6GALNAC4, LSP1, and AC135279.2, respectively.

[0100] In addition, in order to enhance the predictive ability of the acute myeloid leukemia prognostic model, the second data set was used for training, and the optimal regularization parameter λ was obtained as 0.09 through 10-fold cross validation.

[0101] Furthermore, in the process of building the model, for a set of data, the risk score of each acute myeloid leukemia patient is calculated according to the risk score formula, and the median of the risk score in the set of data is used as the critical value, for example, the critical value is 2.03. Patients with risk scores higher than the critical value are included in the high-risk group, and patients with risk scores not higher than the critical value are included in the low-risk group. Based on the risk score, the 1-year, 3-year, and 5-year survival rates of patients with acute myeloid leukemia are predicted.

[0102] The present invention increases the complexity of the acquired data, performs differential expression analysis on the complex data, obtains the common differentially expressed genes with smaller batch effects, performs weighted gene co-expression network analysis, identifies the modules with high correlation with clinical traits as key modules, analyzes the effects of gene expression levels in the key modules on survival time through univariate analysis, regression algorithm analysis, and stepwise regression analysis, and obtains 7 target prognostic markers, namely CD2, LRG1, ANGPT1, CSTB, ST6GALNAC4, LSP1, and AC135279.2, to construct an acute myeloid leukemia prognosis model, which can more accurately predict the survival of patients with acute myeloid leukemia, identify high-risk patients, adjust treatment strategies in time, and improve the overall survival rate of patients.

[0103] The present invention also provides a system for constructing a prognostic model for acute myeloid leukemia, comprising:

[0104] Dataset construction module: used to collect bone marrow samples from patients with acute myeloid leukemia, calculate transcript expression, and construct the first dataset;

[0105] Obtain transcriptome data and clinical data of acute myeloid leukemia patients in the database to construct the second data set;

[0106] Obtain transcriptome data of healthy people in the database to construct the third data set;

[0107] Differential expression analysis module: used to obtain the common differentially expressed gene set through differential expression analysis;

[0108] Key module division module: used to analyze the weighted gene co-expression network of the common differentially expressed gene sets to obtain several key modules, specifically:

[0109] Based on the shared differentially expressed gene set, the Pearson correlation coefficient between the shared differentially expressed genes is calculated. According to the Pearson correlation coefficient between the shared differentially expressed genes, a weighted adjacency matrix is ​​constructed. After the weighted adjacency matrix is ​​converted into a topological overlap matrix, the dissimilarity of the topological overlap matrix is ​​calculated. According to the dissimilarity, the shared differentially expressed genes are clustered, and a gene-module vector clustering diagram is constructed to obtain several modules and characteristic genes in each module.

[0110] The Pearson correlation coefficient between the characteristic genes in each module and the clinical data was calculated, and the modules where the characteristic genes with Pearson correlation coefficients greater than the Pearson correlation coefficient threshold and P values ​​less than the first P value threshold were located were recorded as key modules. After traversing all modules, several key modules were obtained;

[0111] The first prognostic marker screening module: used for the genes in several key modules to be subjected to univariate analysis, calculate the risk ratio and preset confidence interval of each gene, and record the gene with a P value less than the second P value threshold as the first prognostic marker;

[0112] Target prognostic marker screening module: used for analyzing the first prognostic marker through regression algorithm, and recording the gene with likelihood deviation less than the likelihood deviation threshold as the second prognostic marker;

[0113] The second prognostic marker is subjected to stepwise regression analysis to obtain the target prognostic marker and the weight coefficient of the target prognostic marker;

[0114] Acute myeloid leukemia prognostic model construction module: construct an acute myeloid leukemia prognostic model based on the target prognostic marker and the weight coefficient of the target prognostic marker.

Claims

1. A method for constructing a prognostic model for acute myeloid leukemia, characterized in that: The following steps are involved: S1: Collect bone marrow samples from patients with acute myeloid leukemia, calculate transcript expression, and construct the first data set; Obtain transcriptome data and clinical data of acute myeloid leukemia patients in the database to construct the second data set; Obtain the transcriptome data of healthy people in the database to construct the third data set; S2: After differential expression analysis, the common differentially expressed gene set was obtained, specifically: After the first data set was merged with the third data set, differential expression analysis was performed using edgeR, limma, and DESeq2, respectively, to obtain the first up-regulated gene set and the first down-regulated gene set, the second up-regulated gene set and the second down-regulated gene set, and the third up-regulated gene set and the third down-regulated gene set. The first up-regulated differentially expressed gene set was obtained by taking the intersection of the first up-regulated gene set, the second up-regulated gene set, and the third up-regulated gene set. The first down-regulated differentially expressed gene set was obtained by taking the intersection of the first down-regulated gene set, the second down-regulated gene set, and the third down-regulated gene set. After the second data set was merged with the third data set, differential expression analysis was performed using edgeR, limma, and DESeq2, respectively, to obtain the fourth up-regulated gene set and the fourth down-regulated gene set, the fifth up-regulated gene set and the fifth down-regulated gene set, and the sixth up-regulated gene set and the sixth down-regulated gene set. The intersection of the fourth up-regulated gene set, the fifth up-regulated gene set, and the sixth up-regulated gene set was used to obtain the second up-regulated differentially expressed gene set. The intersection of the fourth down-regulated gene set, the fifth down-regulated gene set, and the sixth down-regulated gene set was used to obtain the second down-regulated differentially expressed gene set. The first up-regulated differentially expressed gene set, the second up-regulated differentially expressed gene set, the first down-regulated differentially expressed gene set, and the second down-regulated differentially expressed gene set are merged to obtain a common differentially expressed gene set; S3: The common differentially expressed gene sets were analyzed through weighted gene co-expression network analysis to obtain several key modules, specifically: Based on the shared differentially expressed gene set, the Pearson correlation coefficient between the shared differentially expressed genes is calculated. According to the Pearson correlation coefficient between the shared differentially expressed genes, a weighted adjacency matrix is ​​constructed. After the weighted adjacency matrix is ​​converted into a topological overlap matrix, the dissimilarity of the topological overlap matrix is ​​calculated. According to the dissimilarity, the shared differentially expressed genes are clustered, and a gene-module vector clustering diagram is constructed to obtain several modules and characteristic genes in each module. The Pearson correlation coefficient between the characteristic genes in each module and the clinical data was calculated, and the modules where the characteristic genes with Pearson correlation coefficients greater than the Pearson correlation coefficient threshold and P values ​​less than the first P value threshold were located were recorded as key modules. After traversing all modules, several key modules were obtained; S4: After univariate analysis of genes in several key modules and calculation of the risk ratio and preset confidence interval of each gene, the gene with a P value less than the second P value threshold is recorded as the first prognostic marker; S5: The first prognostic marker is analyzed by regression algorithm, and the gene whose likelihood deviation is less than the likelihood deviation threshold is recorded as the second prognostic marker; The second prognostic marker is subjected to stepwise regression analysis to obtain the target prognostic marker and the weight coefficient of the target prognostic marker; The target prognostic markers are CD2, LRG1, ANGPT1, CSTB, ST6GALNAC4, LSP1, and AC135279.2; The weight coefficients of CD2, LRG1, ANGPT1, CSTB, ST6GALNAC4, LSP1, and AC135279.2 are 0.0074, 0.0221, 0.0102, 0.0203, 0.0186, 0.0043, and -0.0396, respectively; S6: Construct a prognostic model for acute myeloid leukemia based on the target prognostic marker and the weight coefficient of the target prognostic marker.

2. The method for constructing a prognostic model for acute myeloid leukemia according to claim 1, characterized in that: The second prognostic marker in step S5 is CD2, AL359715.1, ANGPT1, ST6GALNAC4, LSP1, PLEKHA5, AC135279.2, SELPLG, GZMB, ISG20, LRG1, RTF2, ZBP1, CSTB and DNASE1L1.

3. The method for constructing a prognostic model for acute myeloid leukemia according to claim 1, characterized in that: In step S3, a weighted adjacency matrix is ​​constructed according to the Pearson correlation coefficients between the shared differentially expressed genes, specifically: According to the formula: , construct a weighted adjacency matrix; In the formula, , They are two different genes in the common differentially expressed set; is the weighted adjacency matrix; For gene and genes The Pearson correlation coefficient matrix between is the soft threshold.

4. The method for constructing a prognostic model for acute myeloid leukemia according to claim 1, characterized in that: In step S3, the differentially expressed genes are clustered to construct a gene-module vector clustering diagram, specifically: The nodes in the vector clustering diagram represent genes, different colors represent different modules, and the connections between nodes represent the co-expression relationship between genes.

5. The method for constructing a prognostic model for acute myeloid leukemia according to claim 1, characterized in that: In step S1, a bone marrow sample of an acute myeloid leukemia patient is collected and the transcript expression level is calculated, specifically: Bone marrow samples were collected from patients with acute myeloid leukemia, and after pretreatment, they were sequenced to obtain sequencing data. The sequencing data were compared with the human genome to obtain comparison data. After deduplication processing, the transcript expression level was calculated.

6. The method for constructing a prognostic model for acute myeloid leukemia according to claim 1, characterized in that: The clinical data in step S1 include age, race, gender, tumor history, treatment history, and treatment methods.

7. The method for constructing a prognostic model for acute myeloid leukemia according to claim 1, characterized in that: In step S6, based on the target prognostic marker and the weight coefficient of the target prognostic marker, an acute myeloid leukemia prognostic model is constructed, specifically: The expression level of each target prognostic marker and the corresponding weight coefficient of the target prognostic marker were multiplied to obtain the score of each target prognostic marker. The scores of all target prognostic markers were added together to obtain the risk score. The survival rate was predicted based on the risk score, and a prognostic model for acute myeloid leukemia was constructed.

8. A system for constructing a prognostic model for acute myeloid leukemia, characterized in that: include: Dataset construction module: used to collect bone marrow samples from patients with acute myeloid leukemia, calculate transcript expression, and construct the first dataset; Obtain transcriptome data and clinical data of acute myeloid leukemia patients in the database to construct the second data set; Obtain the transcriptome data of healthy people in the database to construct the third data set; Differential expression analysis module: used to obtain the common differentially expressed gene set through differential expression analysis, specifically: After the first data set was merged with the third data set, differential expression analysis was performed using edgeR, limma, and DESeq2, respectively, to obtain the first up-regulated gene set and the first down-regulated gene set, the second up-regulated gene set and the second down-regulated gene set, and the third up-regulated gene set and the third down-regulated gene set. The first up-regulated differentially expressed gene set was obtained by taking the intersection of the first up-regulated gene set, the second up-regulated gene set, and the third up-regulated gene set. The first down-regulated differentially expressed gene set was obtained by taking the intersection of the first down-regulated gene set, the second down-regulated gene set, and the third down-regulated gene set. After the second data set was merged with the third data set, differential expression analysis was performed using edgeR, limma, and DESeq2, respectively, to obtain the fourth up-regulated gene set and the fourth down-regulated gene set, the fifth up-regulated gene set and the fifth down-regulated gene set, and the sixth up-regulated gene set and the sixth down-regulated gene set. The intersection of the fourth up-regulated gene set, the fifth up-regulated gene set, and the sixth up-regulated gene set was used to obtain the second up-regulated differentially expressed gene set. The intersection of the fourth down-regulated gene set, the fifth down-regulated gene set, and the sixth down-regulated gene set was used to obtain the second down-regulated differentially expressed gene set. The first up-regulated differentially expressed gene set, the second up-regulated differentially expressed gene set, the first down-regulated differentially expressed gene set, and the second down-regulated differentially expressed gene set are merged to obtain a common differentially expressed gene set; Key module division module: used to analyze the weighted gene co-expression network of the common differentially expressed gene sets to obtain several key modules, specifically: Based on the shared differentially expressed gene set, the Pearson correlation coefficient between the shared differentially expressed genes is calculated. According to the Pearson correlation coefficient between the shared differentially expressed genes, a weighted adjacency matrix is ​​constructed. After the weighted adjacency matrix is ​​converted into a topological overlap matrix, the dissimilarity of the topological overlap matrix is ​​calculated. According to the dissimilarity, the shared differentially expressed genes are clustered, and a gene-module vector clustering diagram is constructed to obtain several modules and characteristic genes in each module. The Pearson correlation coefficient between the characteristic genes in each module and the clinical data was calculated, and the modules where the characteristic genes with Pearson correlation coefficients greater than the Pearson correlation coefficient threshold and P values ​​less than the first P value threshold were located were recorded as key modules. After traversing all modules, several key modules were obtained; The first prognostic marker screening module: used for the genes in several key modules to be subjected to univariate analysis, calculate the risk ratio and preset confidence interval of each gene, and record the gene with a P value less than the second P value threshold as the first prognostic marker; Target prognostic marker screening module: used for analyzing the first prognostic marker through regression algorithm, and recording the gene with likelihood deviation less than the likelihood deviation threshold as the second prognostic marker; The second prognostic marker is subjected to stepwise regression analysis to obtain the target prognostic marker and the weight coefficient of the target prognostic marker; The target prognostic markers are CD2, LRG1, ANGPT1, CSTB, ST6GALNAC4, LSP1, and AC135279.2; The weight coefficients of CD2, LRG1, ANGPT1, CSTB, ST6GALNAC4, LSP1, and AC135279.2 are 0.0074, 0.0221, 0.0102, 0.0203, 0.0186, 0.0043, and -0.0396, respectively; Acute myeloid leukemia prognostic model construction module: construct an acute myeloid leukemia prognostic model based on the target prognostic marker and the weight coefficient of the target prognostic marker.

Images