Construction of a novel 9-gene RISK acute myeloid leukemia prognostic model

Abstract

The present invention discloses a method for constructing a novel 9-gene RISK acute myeloid leukemia prognosis model and a kit and a nomogram model based on the prognosis model. The present invention uses the acute myeloid leukemia transcriptome to screen the gene module related to disease prognosis through weighted gene co-expression network analysis, uses univariate Cox regression analysis combined with WGCNA module analysis to determine the AML prognosis core-related genes, and uses Lasso regression to construct a model for the above-screened AML prognosis core genes to obtain a novel 9-gene RISK acute myeloid leukemia prognosis model. After various verifications, the new prognosis model has a better evaluation effect.

Description

Technical Field

[0001] The invention relates to the field of medical biotechnology, and in particular to a method for constructing a novel 9-gene RISK acute myeloid leukemia prognosis model. Background Art

[0002] Acute myeloid leukemia (AML) is a complex and highly heterogeneous disease that presents different genetic backgrounds and responses to treatment. Risk stratification and prognostic assessment are of great significance for the treatment of AML patients. The current evaluation system mainly depends on patient-related factors, such as increasing age, as well as prognostic-related cytogenetics and gene mutations. Studies have shown that single chromosome karyotypes and RUNX1, ASXL1, and TP53 mutations are considered to be features of poor prognosis in AML due to their independent association with AML prognosis. In addition, according to the 2017 European Leukemia (ELN) recommendations, about 30% of AML patients are still classified as intermediate-risk patients because they do not carry karyotype abnormalities or gene mutations with prognostic value. These patients are particularly lacking in effective risk stratification to guide clinical treatment. Therefore, the current risk stratification and prognostic assessment of AML patients needs to be further improved.

[0003] Evaluation systems that mainly depend on patient-related factors and prognosis-related cytogenetics and gene mutations are very inadequate, and to some extent, abnormalities based on gene expression profiles may be able to make up for the aforementioned shortcomings. Changes in the expression profiles of AML patients based on high-throughput sequencing and gene microarray technology have important prognostic indicative value. Studies have shown that scoring systems derived from gene expression identification have clinical significance for the prognosis of AML, such as 86 probe set gene expression signatures, 24-gene prognostic models, and 11-gene risk scoring systems. However, due to issues such as scope of application and technology, these scoring systems are not widely used in clinical practice. Although some studies have shown that single genes have prognostic value in AML, these results are still insufficient. Developing new AML prognostic models with clinical promotion value is a clinical problem that urgently needs results. Summary of the invention

[0004] To address the above technical deficiencies, the present invention provides a method for constructing a novel 9-gene RISK acute myeloid leukemia prognosis model. The constructed AML prognosis model has good evaluation efficiency and is easy to promote clinically.

[0005] The present invention is achieved through the following technical solutions:

[0006] The present invention uses the whole genome gene expression data of 151 AML patients to perform WGCNA (weighted correlation network analysis) weighted gene co-expression network analysis to extract gene modules related to the survival of AML patients and analyze the core genes that affect the prognosis of AML. At the same time, the applicant conducted a univariate Cox regression analysis of the whole genome to explore AML prognosis-related genes from another aspect. The two groups of genes screened out were intersected to obtain 10 AML prognosis core genes. The gene expression matrix of the screened prognosis-related genes was used as the input file, and the Lasso regression, one of the classic algorithms in machine learning, was used to construct the model, and a new AML prognosis evaluation model was obtained:

[0007] RISK=0.0145*CALCRL+0.00147*FAM30A+0.0280*MRPL16+0.0495*PLA2G4A+0.00794*SEL1L3+0.000641*SLC2A5+0.0962*SOCS1+0.0349*TBCC+0.0301*TREML2.

[0008] Through survival analysis, it was found that a high RISKscore score (Riskscore) predicted worse overall survival, event-free survival, and relapse-free survival in AML patients (including patients with non-acute promyelocytic leukemia and patients with normal karyotype AML).

[0009] The prognostic evaluation ability of the RISK model for AML patients was also validated in three other independent cohorts.

[0010] In addition, the applicant found that the RISK model can be used as a predictor of overall survival, event-free survival, and relapse-free survival in AML independent of patient age, WBC, cytogenetic abnormalities, and FLT3, DNMT3A, TP53, and RUNX1 mutations.

[0011] On this basis, the applicant designed an AML prognosis assessment kit based on the RISK model, which contains specific primers for the 9 genes included in the RISK model and the required SYBRGreen and DNA-free water. This kit uses real-time quantitative PCR to detect gene expression, standardizes the gene test results and brings them into the RISK model calculation formula, and calculates the patient's 1-year, 3-year and 5-year overall survival probability based on the RISK model-based nomogram. The calibration curve of the nomogram shows that this model has good efficacy and is easy to promote clinically.

[0012] The beneficial effects of the present invention are: 1. The AML prognosis model constructed by the method provided by the present invention has good effects and good evaluation efficiency, and is easy to promote clinically. The RISK model provided by the present invention shows better evaluation efficiency compared with other reported 3-gene, 7-gene and 24-gene models, and is also superior to clinical pathological characteristics such as age, WBC count and cytogenetic abnormalities, and has good prospects for clinical promotion. 2. The AML prognosis model constructed by the method provided by the present invention has multiple uses. It was found that the RISK model can be used as a predictive indicator for AML overall survival, event-free survival and recurrence-free survival independent of patient age, WBC, cytogenetic abnormalities and FLT3, DNMT3A, TP53, and RUNX1 mutations. 3. The AML prognosis model constructed by the method provided by the present invention is stable and objective, and has been experimentally verified. The prognostic evaluation ability of the RISK model for AML patients has also been verified in three other independent cohorts. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is a weighted gene co-expression network analysis (WGCNA) of transcriptomes from patients with acute myeloid leukemia (AML).

[0014] Among them: A. Hierarchical clustering diagram of AML sample transcriptome; its clustering is based on the similarity of gene expression profiles, and shows the similarity of gene expression patterns between samples. The overall survival (OS) time, OS status, event-free survival (EFS) time and EFS status of each patient are shown at the bottom. B. Scale-free fit index analysis and determination of soft thresholds. C. Hierarchical clustering dendrogram of genes in different modules. D. Heat map of the topological overlap matrix of 1000 selected genes in WGCNA. E. Correlation analysis between modules and survival features. Each row corresponds to a different module, and the column represents a different survival feature. The p-value of the correlation between each module and the corresponding survival condition is shown in brackets (Pearson correlation coefficient).

[0015] Figure 2 The Lasso model for the prognosis of AML patients was constructed.

[0016] A. Lasso regression screening variables, the black dots in the vertical line represent the mean square error and the upper and lower standard deviations, the smaller the mean square error, the better the model; the number above indicates the number of independent variables that still exist in the model (not necessarily monotonically decreasing), and the first dotted line indicates the minimum mean square error. B. Each curve in the figure represents the change trajectory of each independent variable coefficient, the ordinate is the value of the coefficient, the lower abscissa is log(λ), and the upper abscissa is the number of non-zero coefficients in the model at this time.

[0017] Figure 3 It indicates that a high Riskscore score indicates a poor survival of AML patients.

[0018] Among them, the overall survival (OS), event-free survival (EFS) and relapse-free survival (RFS) of AML patients with high AC.Riskscore scores were shorter than those of patients with low scores; the overall survival (OS), event-free survival (EFS) and relapse-free survival (RFS) of non-APL AML patients with high DF.Riskscore scores were shorter than those of patients with low scores; the overall survival (OS), event-free survival and relapse-free survival of AML patients with high GI.Riskscore scores and normal karyotype were shorter than those of patients with low scores.

[0019] Figure 4 It indicates that three independent validation cohorts all showed that AML patients with high Riskscore scores had a worse prognosis.

[0020] Figure 5 The results showed that multivariate Cox regression indicated that the RISK model could be used as a prognostic factor for overall survival, event-free survival, and relapse-free survival in AML, independent of patient age, white blood cell count (WBC), cytogenetic stratification, FLT3 mutation, DNMT3A mutation, RUNX1 mutation, and TP53 mutation.

[0021] Figure 6 The ROC curve shows that the RISK model constructed by the applicant is significantly better than the patient's age, cytogenetic risk stratification and white blood cell count in evaluating the overall survival of AML patients.

[0022] Figure 7 The ROC curve shows that the RISK model constructed by the applicant is significantly better than the patient's age, cytogenetic risk stratification and white blood cell count in evaluating event-free survival of AML patients.

[0023] Figure 8 The ROC curve shows that the RISK model constructed by the applicant is significantly better than the patient's age, cytogenetic risk stratification and white blood cell count in evaluating the relapse-free survival of AML patients.

[0024] Figure 9 It is the construction of a nomogram that is more suitable for clinical application based on the RISK model.

[0025] Among them, A. A comprehensive nomogram model based on RISK model scoring and including patient age and TP53 mutation information; B, C. The calibration curves of the nomogram model show that the comprehensive nomogram model has good evaluation efficiency. DETAILED DESCRIPTION

[0026] Example 1 Construction and verification of a novel 9-gene RISK acute myeloid leukemia prognosis model

[0027] 1. Research Methods

[0028] 1. Weighted Gene Co-expression Network (WGCNA) Analysis

[0029] WGCNA is a systems biology method used to describe gene correlation patterns between different samples, which can identify candidate biomarker genes or therapeutic targets through the association between genome and phenotype. In this study, the applicant constructed a transcriptome gene expression matrix of 151 AML patients and selected 25% of the genes with the largest variance as input data. WGCNA was performed using the "WGCNA" package in R software, and the adjacency matrix and topological overlap matrix (TOM) were constructed, and the corresponding differences (1-TOM) were calculated. Gene tree structure and module identification were performed by dynamic tree cutting, the correlation between modules and survival conditions was calculated, and modules closely related to the survival of AML patients were selected for further analysis.

[0030] 2. Univariate Cox regression analysis

[0031] In order to identify AML prognosis-related genes, we used the "survival" and "survminer" packages in R software to perform univariate Cox regression analysis on the whole genome expression data of the above 150 AML samples (samples with survival time of 0 were removed) to obtain the assumed hazard ratio of each gene to patient survival (HR, HR=1 has no effect; HR<1 has a better prognosis for patients with high gene expression; HR>1 has a worse prognosis for patients with high gene expression). The FDR-corrected P value <0.001 was used as the statistical boundary, and 50 genes related to AML prognosis were obtained.

[0032] 3. Construction and verification of Lasso model

[0033] Lasso regression, also known as lasso regression, is a new variable selection technique proposed by Robert Tibshirani in 1996. Lasso is a shrinkage estimation method. Its basic idea is to minimize the residual sum of squares under the constraint that the sum of the absolute values ​​of the regression coefficients is less than a constant, so that some regression coefficients that are strictly equal to 0 can be generated, and further an interpretable model can be obtained. The expression matrix of the 10 intersection genes of the module genes closely related to the survival of AML patients obtained by WGCNA and the prognosis-related genes with univariate Cox regression FDR<0.001 was used as the input file, and the "GLMNET" package of R software was used for Lasso regression to establish the RISK risk model.

[0034] 4. Survival Curve Analysis

[0035] Compared with the median Riskscore of the cohort to which the patient belonged, the patients were divided into high Riskscore group and low Riskscore group, and the Kaplan-Meier survival curve was used to verify the differences in overall survival (OS), event-free survival (EFS) and relapse-free survival among 150 AML patients, non-acute promyelocytic leukemia AML patients and AML patients with normal karyotype. Among them, EFS refers to the interval between the initial diagnosis and relapse or death of AML patients. The log-rank test was used to calculate whether there was a statistically significant difference in the effect of Riskscore on OS, EFS and RFS.

[0036] 5.ROC

[0037] Receiver operating curve (ROC curve) analysis is a means of evaluating the predictive ability of a factor and is an effective method for grouping continuous variables. In survival analysis, both the disease status and the factor values ​​change over time. In this case, the use of time-dependent ROC is undoubtedly a better choice. In this study, the time-dependent ROC analysis and the calculation of the area under the curve were performed by the "SURVCOMP" and "SURVIVAL" packages of the R software.

[0038] 6. Real-time fluorescence quantitative PCR

[0039] Real-time fluorescence quantitative PCR was performed using cDNA from AML patient samples using SYBR Green reaction (20 μL system) with the following formula:

[0040]

[0041] The real-time fluorescence quantitative PCR process was completed in a CFX96 fluorescence quantitative PCR instrument. Three duplicate wells were set for each sample, and the reaction procedure was as follows: the first stage: 95°C for 10 s; the second stage: 95°C for 5 s and 60°C for 31 s, repeated for 40 cycles; the third stage: melting curve collection at 65°C-95°C for a total of 5 s.

[0042] The primer sequences are as follows:

[0043] Table 1 Real-time fluorescence quantitative PCR primer sequences (nucleotide sequences are shown in SEQ ID No. 1-18)

[0044] Forward primer Reverse primer FAM30A 5′-TTGAATAGAGTAGTTCCTTGCGCTG-3′ 5′-GGCTACTTCACCCAGCTGTCTAG-3′ CALCRL 5′-CAGCAAGCAACAGAACATGGA-3′ 5′-TGCCAAGCGAGATAAGCAGTG-3′ PLA2G4A 5′-TACCAGCACATTATAGTGGAGCA-3′ 5′-GCTGTCAGGGGTTGTAGAGAT-3′ TREML2 5′-CCTATAAGGGCTACAAAAACCGC-3′ 5′-CCCATCAAGGGGTACAGGA-3′ SOCS1 5′-CACGCACTTCCGCACATTC-3′ 5′-TAAGGGCGAAAAAGCAGTTCC-3′ SLC2A5 5′-ACGTTGCTGTGGTCTGTAACC-3′ 5′-CATTAAGATCGCAGGCACGATA-3′ MRPL16 5′-TGCTGGCGTAAAGACACTGC-3′ 5′-TCTTACTTTTGGCACAAGTGGTG-3′ SEL1L3 5′-TTCAGAAGTAGCATTCCCGTGT-3′ 5′-TTTCCGCCTTTCAACTTCCAG-3′ TBCC 5′-CAGGACCGGAGACATGGAGT-3′ 5′-TTTCCGCCTTTCAACTTCCAG-3′

[0045] 7. Construction and evaluation of nomograms

[0046] The alignment diagram, also known as the nomogram, is based on multi-factor regression analysis. It integrates multiple prediction indicators and then uses scaled segments to draw them on the same plane in a certain proportion to express the relationship between the variables in the prediction model. The applicant used the RMS R package to construct the alignment diagram and calibration diagram based on the RISK model, where the calibration diagram was used to evaluate the accuracy of the nomogram.

[0047] 2. Research Results:

[0048] 1. The acute myeloid leukemia (AML) transcriptome was screened for gene modules related to disease prognosis through weighted gene co-expression network analysis (WGCNA).

[0049] WGCNA is a systems biology method that divides genes into different clusters or modules based on the similarity of gene expression patterns and can be used to analyze the potential functions of genes. In order to identify modules related to AML prognosis, we performed WGCNA analysis based on the gene expression profiles and survival information of AML patients. Samples with large differences in gene expression patterns ( Figure 1 A). We use scale-free R2=0.8 as the soft threshold to construct a scale-free network ( Figure 1 B), and after dynamic tree cutting and merging, 32 gene co-expression modules were obtained ( Figure 1 C). Plotting the heatmap of the topological overlap matrix of 1000 selected genes in WGCNA, showing that each module was validated independently of each other ( Figure 1 D). We calculated the correlation between each module and the survival of AML patients (overall survival time, overall survival status, event-free survival time, and event-free survival status), and identified five modules with statistically significant associations with AML patient survival, namely: darkgrey module, green module, grey60 module, lightyellow module, and red module ( Figure 1 E).

[0050] 2. Univariate Cox regression analysis combined with WGCNA module analysis to determine the core genes related to AML prognosis

[0051] Univariate Cox regression analysis was performed on the whole genome expression data of AML genes, and the corresponding P values ​​were corrected for FDR, and 50 genes were obtained that were significantly closely related to the prognosis of AML (FDR<0.001) (Table 2). The genes screened by univariate Cox regression were intersected with the prognosis-related module genes obtained by the above WGCNA, and 10 core genes related to the prognosis of AML patients were obtained, namely FAM30A, CALCRL, PLA2G4A, TREML2, CCDC6, SOCS1, SLC2A5, MRPL16, SEL1L3 and TBCC (Table 3).

[0052] Table 2. Univariate Cox regression analysis of transcriptomes in AML patients (FDR-corrected P value < 0.001)

[0053]

[0054]

[0055] Table 3. Intersection genes of univariate Cox regression and WGCNA prognostic-related modules of AML transcriptome data

[0056] Gene HR PValue FDR moduleColor FAM30A 1.011 4.35E-10 2.67E-06 red CALCRL 1.072 2.38E-08 4.32E-05 red PLA2G4A 1.095 1.55E-08 4.32E-05 grey60 TREML2 1.147 1.91E-08 4.32E-05 grey60 CCDC6 1.123 5.57E-08 8.56E-05 red SOCS1 1.167 9.58E-08 0.000105489 red SLC2A5 1.041 4.53E-07 0.000261531 red MRPL16 1.052 1.15E-06 0.000455919 red SEL1L3 1.043 1.32E-06 0.000487985 grey60 TBCC 1.100 2.00E-06 0.000630256 grey60

[0057] 3. Construction of Lasso model of core genes related to AML prognosis

[0058] The Lasso method adds a penalty term to the least squares to compress the estimated parameters. When the parameter is reduced to less than a threshold, it is set to 0, thereby selecting the independent variable that has a greater impact on the dependent variable and calculating the corresponding regression coefficient, and finally obtaining a more streamlined model. The LASSO method has obvious advantages in dealing with sample data with multicollinearity. The applicant used Lasso regression to construct a model for the 10 AML prognosis core genes screened above, and the model obtained is as follows (Figure 2A, B):

[0059] RISK=0.0145*CALCRL+0.00147*FAM30A+0.0280*MRPL16+0.0495*PLA2G4A+0.00794*SEL1L3+0.000641*SLC2A5+0.0962*SOCS1+0.0349*TBCC+0.0301*TREML2

[0060] The risk model scores (Riskscore) of 150 patients were calculated and divided into two groups according to the median (Table 4).

[0061] Table 4. Riskscore calculation and grouping of Lasso model for AML transcriptome

[0062]

[0063]

[0064]

[0065]

[0066]

[0067] 4. Higher Riskscore scores indicate a worse prognosis for AML patients

[0068] AML patients were divided into two groups according to the median Riskscore score. First, a survival analysis was performed on 150 newly diagnosed AML patients. The results showed that the overall survival (OS), event-free survival (EFS), and relapse-free survival (RFS) of AML patients with high Riskscore scores were shorter than those of patients with low scores ( Figure 3 AC). The treatment and prognosis of patients with acute promyelocytic leukemia (APL) in AML are very different from those of other types of AML. We performed a prognostic analysis on non-APL AML patients and found that the overall survival (OS), event-free survival (EFS) and relapse-free survival (RFS) of non-APL AML patients with high Riskscore scores were shorter than those of patients with low scores (Figure 3D-F). Cytogenetic abnormalities are an important basis for risk stratification of AMLf, but some AML patients do not carry abnormal karyotypes. Therefore, the prognostic assessment of these patients is still a clinical problem that needs to be solved urgently. We found that the overall survival (OS), event-free survival and relapse-free survival of AML patients with high Riskscore scores and normal karyotypes were shorter than those of patients with low scores (Figure 3D-F). Figure 3 GI). Three independent validation cohorts all showed that AML patients with high Riskscore scores had a worse prognosis. To further evaluate the evaluation performance of the RISK model in AML, we validated it in three independent validation cohorts, and the results showed that AML patients with high Riskscore scores had a worse prognosis (Figure 4).

[0069] 5. High Riskscore is an independent predictor of poor prognosis in AML

[0070] A variety of clinical factors have a significant impact on the prognosis of AML, represented by age, white blood cell (WBC) count, and cytogenetic abnormalities. In addition, many gene mutations also have an important impact on the prognosis of AML, such as DNMT3A, TP53, etc. We included the patient age Riskscore together with patient age, WBC, cytogenetic abnormalities, and FLT3, DNMT3A, TP53, and RUNX1 mutations in multivariate Cox regression. The results showed that when combined with all these prognostic factors in multivariate analysis, higher FHL1 expression was still a prognostic factor for AML overall survival, event-free survival, and relapse-free survival, and was independent of the patient's clinical characteristics and gene mutations (Table 5, Figure 5 ).

[0071] Table 5. Multivariate Cox regression showed that the RISK model can be used as an independent prognostic factor for overall survival, event-free survival, and relapse-free survival in AML.

[0072] Overall survival

[0073]

[0074]

[0075] Event-free survival

[0076]

[0077] Relapse-free survival

[0078]

[0079] 6. The RISK model shows good evaluation performance compared with other reported models

[0080] A variety of transcriptome-related models have been reported to be able to assess the prognosis of AML patients, such as the 3-gene, 7-gene, and 24-gene models. In order to compare the evaluation performance of the RISK model constructed by the applicant with the reported models, we included the RISK model and the 3-gene, 7-gene, and 24-gene models into the same Cox regression model for comparison. The results showed that the RISK model showed better evaluation effects than these three reported AML prognosis models (Table 6).

[0081] Table 6. The RISK model constructed by the applicant is significantly superior to the reported 3-gene, 7-gene and 24-gene models in terms of AML prognosis assessment performance.

[0082]

[0083] 7. The RISK model outperforms age, WBC count, and cytogenetic abnormalities in assessment

[0084] To further clarify the evaluation value of the RISK model, we used the receiver operating characteristic (ROC) curve to compare the impact of the RISK model on patient survival with the important clinical factors currently used in AML evaluation. We found that the RISK model showed a better prognostic effect on AML than the patient's age, white blood cell count, and cytogenetic abnormalities in terms of overall survival, event-free survival, and progression-free survival. Figures 6 - 8 ). The above results show that the RISK model is an effective and independent prognostic indicator for AML.

[0085] 8. Design of RISK-based test kit

[0086] Based on the above preliminary results, the applicant designed a kit based on the RISK model, which mainly includes: SYBRGreen, water without DNA and 9 gene primers included in the RISK model, whose nucleotide sequences are shown in SEQ ID No. 1-18.

[0087] 9. Construct a nomogram based on the RISK model for clinical application

[0088] In order to further facilitate the clinical promotion of the RISK model, the applicant constructed a nomogram model based on the RISK model (Figure 9A). The cDNA of AML patients was subjected to real-time fluorescence quantitative PCR using the above-mentioned RISK kit to obtain the corresponding cycle number and standardize it, calculate the Riskscore, match the corresponding score of the constructed nomogram, and obtain the patient's final prognostic score in combination with the patient's age and TP53 mutation status. The patient's corresponding 1-year, 3-year and 5-year survival rates in the nomogram. The calibration curve shows that this nomogram has a relatively accurate predictive efficacy ( Figure 9 B). Based on the above results, the RISK model proposed by the applicant can accurately predict the prognosis of AML patients and has good clinical application prospects.

[0089] The above is only a preferred implementation of this patent. It should be pointed out that for ordinary technicians in this technical field, several improvements and substitutions can be made without departing from the technical principles of this patent. These improvements and substitutions should also be regarded as the scope of protection of this patent. Sequence Listing <110> Shandong University <120> Construction of a novel 9-gene RISK acute myeloid leukemia prognostic model <141> 2021-12-17 <160> 18 <170> SIPOSequenceListing 1.0 <210> 1 <211> 25 <212> DNA <213> Artificial Sequence <400> 1 ttgaatagag tagttccttg cgctg 25 <210> 2 <211> twenty three <212> DNA <213> Artificial Sequence <400> 2 ggctacttca cccagctgtc tag 23 <210> 3 <211> twenty one <212> DNA <213> Artificial Sequence <400> 3 cagcaagcaa cagaacatgg a 21 <210> 4 <211> twenty one <212> DNA <213> Artificial Sequence <400> 4 tgccaagcga gataagcagt g 21 <210> 5 <211> twenty three <212> DNA <213> Artificial Sequence <400> 5 taccagcaca ttatagtgga gca 23 <210> 6 <211> twenty one <212> DNA <213> Artificial Sequence <400> 6 gctgtcaggg gttgtagaga t 21 <210> 7 <211> twenty three <212> DNA <213> Artificial Sequence <400> 7 cctataaggg ctacaaaaac cgc 23 <210> 8 <211> 19 <212> DNA <213> Artificial Sequence <400> 8 cccatcaagg ggtacagga 19 <210> 9 <211> 19 <212> DNA <213> Artificial Sequence <400> 9 cacgcacttc cgcacattc 19 <210> 10 <211> twenty one <212> DNA <213> Artificial Sequence <400> 10 taagggcgaa aaagcagttc c 21 <210> 11 <211> twenty one <212> DNA <213> Artificial Sequence <400> 11 acgttgctgt ggtctgtaac c 21 <210> 12 <211> twenty two <212> DNA <213> Artificial Sequence <400> 12 cattaagatc gcaggcacga ta 22 <210> 13 <211> 20 <212> DNA <213> Artificial Sequence <400> 13 tgctggcgta aagacactgc 20 <210> 14 <211> twenty three <212> DNA <213> Artificial Sequence <400> 14 tcttactttt ggcacaagtg gtg 23 <210> 15 <211> twenty two <212> DNA <213> Artificial Sequence <400> 15 ttcagaagta gcattcccgt gt 22 <210> 16 <211> twenty one <212> DNA <213> Artificial Sequence <400> 16 tttccgcctt tcaacttcca g 21 <210> 17 <211> 20 <212> DNA <213> Artificial Sequence <400> 17 caggaccgga gacatggagt 20 <210> 18 <211> twenty one <212> DNA <213> Artificial Sequence <400> 18 tttccgcctt tcaacttcca g 21

Claims

1. A method for constructing a novel 9-gene RISK acute myeloid leukemia prognosis model, characterized in that The following steps are involved: STP1. Screening of gene modules related to disease prognosis by weighted gene co-expression network analysis of acute myeloid leukemia transcriptome; STP2. Univariate Cox regression analysis combined with WGCNA module analysis to determine the core genes related to AML prognosis; STP3. The above-screened AML prognostic core genes were used to construct a model using Lasso regression to obtain a new 9-gene RISK acute myeloid leukemia prognostic model: RISK =0.0145*CALCRL + 0.00147*FAM30A + 0.0280*MRPL16 + 0.0495*PLA2G4A +0.00794*SEL1L3 + 0.000641*SLC2A5 + 0.0962*SOCS1 + 0.0349*TBCC + 0.0301*TREML2; STP4. Survival curve analysis was performed. The risk model score of the novel 9-gene RISK acute myeloid leukemia prognostic model showed that a higher risk model score indicated a worse prognosis for AML patients. STP5. Multivariate Cox regression validation of the independence of the novel 9-gene RISK acute myeloid leukemia prognostic model and other existing validation factors; STP6. Compare the new 9-gene RISK acute myeloid leukemia prognostic model with other existing gene prognostic models to verify the evaluation efficacy; STP7. Use receiver operating curves to compare the impact of the RISK model with important clinical factors in AML assessment on patient survival and verify the evaluation performance.

2. The method for constructing a novel 9-gene RISK acute myeloid leukemia prognosis model according to claim 1, characterized in that: In the STP1, five modules were identified to have statistically significant associations with AML patient survival, namely: darkgrey module, green module, grey60 module, lightyellow module and red module.

3. The method for constructing a novel 9-gene RISK acute myeloid leukemia prognosis model according to claim 1, characterized in that: In the STP2, univariate Cox regression analysis was performed on the whole genome expression data of AML genes, and the FDR-corrected P value < 0.001 was used as the statistical boundary to obtain 50 genes that were significantly closely related to AML prognosis; The genes screened by univariate Cox regression were intersected with the prognosis-related module genes obtained by the above WGCNA, and 10 core genes related to the prognosis of AML patients were obtained, namely FAM30A, CALCRL, PLA2G4A, TREML2, CCDC6, SOCS1, SLC2A5, MRPL16, SEL1L3 and TBCC.

4. The method for constructing a novel 9-gene RISK acute myeloid leukemia prognosis model according to claim 1, characterized in that: In the STP3, the Lasso regression model construction uses the expression matrix of 10 intersection genes of the module genes closely related to the survival of AML patients obtained by WGCNA and the prognosis-related genes with univariate Cox regression FDR<0.001 as the input file, and uses the "GLMNET" package of R software to perform Lasso regression to establish a new 9-gene RISK acute myeloid leukemia prognosis model.

5. The method for constructing a novel 9-gene RISK acute myeloid leukemia prognosis model according to claim 1, characterized in that: In the STP4, the survival curve analysis was performed on three independent validation cohorts, namely: AML patients, non-APL AML patients, and AML patients with normal karyotype.

6. The method for constructing a novel 9-gene RISK acute myeloid leukemia prognosis model according to claim 1, characterized in that: In the STP5, multivariate Cox regression showed that the RISK model can be used as a prognostic factor for overall survival, event-free survival and relapse-free survival of AML, independent of patient age, white blood cell count WBC, cytogenetic stratification, FLT3 mutation, DNMT3A mutation, RUNX1 mutation and TP53 mutation.

7. The method for constructing a novel 9-gene RISK acute myeloid leukemia prognosis model according to claim 1, characterized in that: In the STP6, the RISK model and the 3-gene, 7-gene, and 24-gene models were included in the same Cox regression model for comparison; the results showed that the RISK model showed better evaluation effects compared with these three reported AML prognostic models.

8. The method for constructing a novel 9-gene RISK acute myeloid leukemia prognosis model according to claim 1, characterized in that: In the STP7, the evaluation performance of the RISK model was superior to that of age, WBC count and cytogenetic abnormalities.

9. A kit based on the RISK model according to any one of claims 1 to 8, characterized in that: Contains SYBRGreen, water without DNA, and 9 gene primers included in the RISK model. The sequences are shown in the table below. The nucleotide sequences are shown in SEQ ID No. 1-18. 。 10. An application of a nomogram model based on a RISK model, wherein the cDNA of an AML patient is subjected to real-time fluorescence quantitative PCR using the kit of the RISK model according to claim 9, the corresponding cycle number is obtained and standardized, the Riskscore is calculated, the corresponding score of the constructed nomogram is matched, and the patient's final prognostic score is obtained in combination with the patient's age and TP53 mutation status, and the corresponding 1-year, 3-year and 5-year survival rates of the patient in the nomogram are obtained; and the accuracy of the nomogram is evaluated using a calibration graph.

Images