Construction of glioblastoma prognosis prediction model based on multiple biomarkers

Abstract

The invention discloses construction of a glioblastoma prognosis prediction model based on multiple biomarkers, belongs to the technical field of biology and computers, and utilizes 10 biomarkers for glioblastoma prognosis evaluation and clinical data of patients to construct a patient accurate prognosis model. The method is used for future clinical diagnosis and treatment strategy optimization.

Description

Technical Field

[0001] The present invention relates to the fields of biological technology and computer technology, and in particular to a glioblastoma prognosis prediction model based on multiple biomarkers and clinical information of glioblastoma. Background Art

[0002] Glioblastoma (GBM) is a high-grade glioma (grade IV). Approximately 49% of malignant brain tumors are glioblastomas, which are the most aggressive and lethal gliomas. Clinically, GBM has a poor prognosis, with less than 5% of patients surviving for more than 5 years after diagnosis. The pathogenesis of GBM is still unclear, and due to the lack of effective treatment, it is prone to recurrence. The clinical manifestations of GBM are usually related to the location, size and growth rate of the tumor. Common symptoms include headache, epilepsy, cognitive dysfunction, movement disorders and paresthesia. Since these symptoms are also common in other diseases, accurate diagnosis usually requires imaging examinations such as magnetic resonance imaging (MRI) and computed tomography (CT). In imaging, GBM often appears as an irregular mass with clear boundaries, and edema may occur around the tumor. Tissue biopsy is currently the gold standard for clinical diagnosis, but the ambiguity of tissue morphology and the heterogeneity of tumors pose challenges to the diagnosis, prognosis prediction and treatment of GBM. The experience of histological diagnosis has produced considerable differences between observers, limiting the reproducibility of diagnosis, which has led to temporary pathological diagnosis and significant clinical confusion, resulting in insufficient prognostic accuracy. Genetic variation is a clear diagnostic molecular marker for glioblastoma, which is more reliable and objective than clinical criteria. The molecular classification of GBM based on gene spectrum and molecular biological characteristics is a necessary supplement to the traditional GBM classification based on pathology.

[0003] Molecular-based diagnosis, patient stratification, and treatment are becoming increasingly important. In 2016, the World Health Organization updated its guidelines to combine morphology and genetic variation to establish new standards for the diagnosis and treatment of glioblastoma. In 2016, the two important entities classified by the WHO based on the isocitrate dehydrogenase (IDH) gene mutation status of GBM were IDH wild-type GBM and IDH mutant GBM. Patients who could not fully evaluate IDH were classified as GBM NOS. Multi-omics studies have explained the profile of GBM in databases such as the Tumor Genome Atlas Research Network (TCGA) and the Chinese Glioma Genome Atlas (CGGA), revealing the complex genetic spectrum of GBM. These abnormal molecules include 1p and 19q coding deletions (oligodendroglioma-specific), IDH gene mutations, PTEN (phosphatase and tensin homolog) gene mutations, TP53 mutations, TERT (telomerase reverse transcriptase) gene promoter mutations, ATRX (α-thalassemia / mental retardation syndrome x-related) gene promoter mutations, and EGFR (epidermal growth factor receptor) gene amplification, which have forced clinical pathology to reconsider the treatment of GBM. Classification of GBM based on abnormal molecules has gradually shortened the time from diagnosis to treatment and significantly improved accuracy and targeting, but the prognosis of glioblastoma patients is still not optimistic. Prognostic assessment is a key component in treatment planning and patient care, and accurate biomarkers are needed to guide clinical decisions.

[0004] Biomarkers refer to biochemical indicators that can determine changes in the structure or function of systems, organs, tissues, cells, and subcellular structures. They are widely used in the diagnosis and staging of diseases, as well as in the evaluation of the safety and effectiveness of new drugs or new therapies in specific populations. For patients with glioblastoma (GBM) with large intra- and inter-individual heterogeneity, the use of new biomarkers for molecular typing has become a consensus for diagnosis. Compared with traditional pathological diagnosis, biomarkers have obvious advantages, and the use of biomarkers with significant prognostic value to screen for clinical accurate prognosis prediction has great application value. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a glioblastoma prognosis prediction model based on multiple biomarkers.

[0006] The technical solution of the present invention is:

[0007] The invention discloses an application of a detection reagent in the preparation of a kit for predicting the prognosis of glioblastoma, wherein the detection reagent is composed of reagents for detecting the expression levels of 12 genes, and the 12 genes are RPS3A gene, CLASP2 gene, NDUFA11 gene, KLHL23 gene, COX7A2L gene, BTF3 gene, IL4R gene, ARPC1B gene, RCC1L gene and CUL1 gene.

[0008] The application of a biomarker combination in the construction of a prognostic risk prediction model for glioblastoma, the biomarker combination being the expression levels of 10 genes, and the 10 genes being RPS3A gene, CLASP2 gene, NDUFA11 gene, KLHL23 gene, COX7A2L gene, BTF3 gene, IL4R gene, ARPC1B gene, RCC1L gene and CUL1 gene.

[0009] Furthermore, the application is: using the age (Age), gender (Gender), whether radiotherapy (radiation therapy), whether temozolomide treatment (chemo tmz), whether afatinib treatment (chemo alk), whether targeted molecular therapy (targeted molecular therapy) and the expression levels of the above-mentioned 10 genes of the samples in the glioblastoma database as covariates, and taking survival time and survival status as dependent variables, a Cox regression model is used to construct a glioblastoma prognosis risk prediction model.

[0010] Furthermore, the formula of the glioblastoma prognosis risk prediction model is:

[0011] Risk Score=0.0279×Age-0.1793×Gender-5.8572×chemo alk+6.6536×chemo tmz-0.2183×radiation therapy-0.9607×targeted moleculartherapy+0.7293×RPS3A-0.1213×CLASP2-0.2685×NDUFA11+0.9958×KLHL23-1. 5604×COX7A2L+0.9382×BTF3+1.0253×IL4R-0.2268×ARPC1B+0.6410×RCC1L-0.8122×CUL1

[0012] Among them, Gender was assigned as 0 for male and 1 for female; chemo alk was assigned as 1 for yes and 0 for no; chemo tmz was assigned as 1 for yes and 0 for no; radiation therapy was assigned as 1 for yes and 0 for no; targeted molecular therapy was assigned as 1 for yes and 0 for no; RPS3A, CLASP2, NDUFA11, KLHL23, COX7A2L, BTF3, IL4R, ARPC1B, RCC1 and CUL1 were calculated as the log2(FPKM+1) of the corresponding genes.

[0013] In the present invention, the RPS3A gene is located in the 31.3th segment of the long arm (q arm) of human chromosome 4 (data source: NCBI Gene database: https: / / www.ncbi.nlm.nih.gov / gene / 6189); the CLASP2 gene is located in the 22.3th segment of the short arm (p arm) of human chromosome 3 (data source: NCBI Gene database: https: / / www.ncbi.nlm.nih.gov / gene / 23122); the NDUFA11 gene is located in the 13.3th segment of the short arm (p arm) of human chromosome 19 (data source: NCBI Gene database: https: / / www.ncbi.nlm.nih.gov / gene / 126328); the KLHL23 gene is located in the 31.1 segment of the long arm (q arm) of human chromosome 2 (data source: NCBIGene database: https: / / www.ncbi.nlm.nih.gov / gene / 151230); the COX7A2L gene is located in the 21st segment of the short arm (p arm) of human chromosome 2 (data source: NCBI Gene database: https: / / www.ncbi.nlm.nih.gov / gene / 9167); the BTF3 gene is located in the 13.2 segment of the long arm (q arm) of human chromosome 5 (data source: NCBI Gene database: https: / / www.ncbi.nlm.nih.gov / gene / 689); the IL4R gene is located in the 12.1 segment of the short arm (p arm) of human chromosome 16 (data source: NCBI Gene database: https: / / www.ncbi.nlm.nih.gov / gene / 3566); the ARPC1B gene is located in the 22.1 segment of the long arm (q arm) of human chromosome 7 (data source: NCBI Gene database: https: / / www.ncbi.nlm.nih.gov / gene / 10095); the RCC1L gene is located in the 11.23 segment of the long arm (q arm) of human chromosome 7 (data source: NCBIGene database: https: / / www.ncbi.nlm.nih.gov / gene / 81554); the CUL1 gene is located in the 36.1 segment of the long arm (q arm) of human chromosome 7 (data source: NCBI Gene database: https: / / www.ncbi.nlm.nih.gov / gene / 8454).

[0014] Compared with the prior art, the present invention has the following beneficial effects:

[0015] 1. Efficient data integration and processing: The authoritative data set TCGA-GBM was used to ensure sample diversity and data reliability; molecular label information was effectively integrated through Cox regression and other methods, improving the depth and breadth of data analysis.

[0016] 2. Accurate subtype utilization: Provide accurate prognostic feedback based on the patient's classification and gene expression, and use a variety of tumor molecular markers to assist in clinical decision-making for glioblastoma.

[0017] 3. Verification of survival analysis: Survival analysis was performed on the 10 screened prognostic core genes, and it was found that high expression of RPS3A, CLASP2, NDUFA11, KLHL23, COX7A2L and BTF3 genes was associated with better prognosis, while high expression of IL4R, ARPC1B, RCC1L and CUL1 genes was associated with poor prognosis, verifying the prognostic prediction ability of the core genes.

[0018] The present invention provides a systematic and scientific process for constructing a prognostic model using GBM prognostic assessment biomarkers. Through multi-step detailed analysis, the reliability of the screened markers and the effectiveness of the prognostic model are ensured, providing a valuable reference for clinical practice. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The glioblastoma prognostic model constructed by the present invention is based on ten molecular markers of RPS3A gene, CLASP2 gene, NDUFA11 gene, KLHL23 gene, COX7A2L gene, BTF3 gene; IL4R gene, ARPC1B gene, RCC1L gene and CUL1 gene and clinical data of patients' radiotherapy, chemotherapy, targeted therapy, etc.

[0020] Figure 2 It is the time ROC curve evaluation of the prognostic model implemented by the present invention. DETAILED DESCRIPTION

[0021] The experimental methods in the following examples are conventional methods unless otherwise specified. The experimental materials used in the following examples are purchased from commercial channels unless otherwise specified.

[0022] Example 1 Construction of a glioblastoma prognosis prediction model

[0023] S1. Download TCGA-GBM patient transcriptome data as a test dataset.

[0024] In this embodiment, the TCGA-GBM patient transcriptome data were downloaded from the transcriptome sequencing data of the glioma project of The Cancer Genome Atlas Program (TCGA), the data types were RNA-seq and patient clinical follow-up data, the number of patients was 140, hereinafter referred to as TCGA-GBM.

[0025] S2. Download TCGA-GBM patient clinical data as a test dataset.

[0026] In this embodiment, the clinical data of TCGA-GBM patients are downloaded from the Cancer Genome Atlas Program (TCGA) glioma project, and the data types include the patient's age, gender, chemotherapy execution status, radiotherapy execution status, targeted therapy execution status, survival time, etc.

[0027] S3. Normalize gene expression values ​​and integrate data

[0028] The FPKM values ​​of ten genes (RPS3A gene, CLASP2 gene, NDUFA11 gene, KLHL23 gene, COX7A2L gene, BTF3 gene, IL4R gene, ARPC1B gene, RCC1L gene and CUL1 gene) were extracted using R language, together with the clinical data, overall survival (OS) and survival status (i.e., whether the patient survived) of each sample, and the FPKM values ​​were converted into log2 (FPKM+1) form. The gene expression data and survival data were read using R's read.csv() function.

[0029] The rows and columns were transposed, the row names of the gene expression data were set to sample IDs, the transcriptome data of the patients were matched with the clinical data by ID number, and patients with missing items were deleted, and finally 140 patients were screened.

[0030] S4. Multivariate Inputs into Prognostic Models

[0031] Extract and integrate the 16 variables included in the TCGA-GBM cohort, and read the clinical data and gene expression data using the read.csv() or read.table() function of R. Set the file path to the corresponding data file.

[0032] The following 16 variables were selected from the clinical data:

[0033] Age: Age at diagnosis.

[0034] Gender: Patient gender (male / female set to 0 / 1).

[0035] Radiation therapy: whether the patient received radiation therapy (yes / no, set to 1 / 0).

[0036] Chemotherapy status: whether receiving temozolomide treatment (chemo tmz) (yes / no set to 1 / 0)

[0037] Whether receiving afatinib treatment (chemo alk) (yes / no set to 1 / 0)

[0038] Targeted molecular therapy: whether to receive targeted molecular therapy (yes / no set to 1 / 0)

[0039] Gene expression level: The log2 (FPKM+1) gene expression levels of C1-GBM, ABRAXAS1, OSMR, STAT3, MYD88, IL6ST, SOCS3, UBE2V2, PSMF1, PSMA8, and KAT5 were extracted.

[0040] S5. Construction of a prognostic model for glioblastoma

[0041] In this embodiment, a Cox regression model is constructed for 16 key genes such as age, gender, radiotherapy, chemotherapy, etc., in which age, gender, radiotherapy, chemotherapy and gene expression level are used as independent variables, and the patient's survival time and survival status are used as dependent variables.

[0042] The formula for constructing the Cox model is in the form of Surv(time,status)~X1+X2+...+Xn, where X1,X2,...,Xn are variables in the input data (biomarkers, age, gender, treatment status, etc.).

[0043] Fit the Cox proportional hazard model and output the results to obtain the hazard ratio coefficient (Coef) for each variable. The coefficient represents the effect of the variable on the survival risk. A positive coefficient indicates an increased risk and a negative coefficient indicates a decreased risk. Calculate the median survival time of the survival model and calculate the survival function for subsequent predictions.

[0044]

[0045]

[0046] The calculated regression risk function is:

[0047] Risk Score=0.0279×Age-0.1793×Gender-5.8572×chemo alk+6.6536×chemo tmz-0.2183xradiation therapy-0.9607×targeted molecular therapy+0.7293×RPS3A-0.1213×CLASP2-0.2685×NDUFA 11+0.9958×KLHL23-1.5604×COX7A2L+0.9382×BTF3+1.0253×IL4R-0.2268×ARPC1B+0.6410×RCClL-0.8122×CUL1

[0048] Use the nomogram to generate a prognostic model ( Figure 1 ), showing the 1-year and 3-year survival probabilities. surv(365, x) and surv(365*3, x) calculate the survival probabilities of patients at specific time points (1 year and 3 years).

[0049] S6. Construction of time ROC curve for model evaluation

[0050] To construct the ROC (receiver operating characteristic) curve, first use the Cox proportional hazard model (f_cph) to perform survival prediction on the training data set (training_dataset), calculate the linear prediction value of each sample (pred_f_training), and organize the survival time, status and prediction score into a data frame (data_table).

[0051] Check if there are missing values ​​in the data frame to ensure data integrity, and terminate the program if there are missing values. Then, use the timeROC function to calculate the ROC curve, set the survival time, state, linear prediction score, event type of interest, and calculation time point, and obtain the true positive and false positive values. The results are organized into a new data frame (time_ROC_df) for plotting.

[0052] Use ggplot2 to draw the ROC curve ( Figure 2 ), set different colors for the results of 1 year and 3 years, and add AUC value annotations to the graph. Finally, improve the readability of the graph by adjusting the graph theme and axis labels, and save the graph as a PDF file.

Claims

1. Use of a detection reagent in the preparation of a kit for predicting the prognosis of glioblastoma, the detection reagent being composed of reagents for detecting the expression levels of 12 genes, the 12 genes being RPS3A gene, CLASP2 gene, NDUFA11 gene, KLHL23 gene, COX7A2L gene, BTF3 gene, IL4R gene, ARPC1B gene, RCC1L gene and CUL1 gene.

2. The application of biomarker combination in the construction of prognostic risk prediction model for glioblastoma, which is characterized by: The biomarker combination is the expression levels of 10 genes, and the 10 genes are RPS3A gene, CLASP2 gene, NDUFA11 gene, KLHL23 gene, COX7A2L gene, BTF3 gene, IL4R gene, ARPC1B gene, RCC1L gene and CUL1 gene.

3. The use according to claim 2, characterized in that: The application is: using the age (Age), gender (Gender), whether radiotherapy (radiation therapy), whether temozolomide treatment (chemo tmz), whether afatinib treatment (chemo alk), whether targeted molecular therapy (targeted molecular therapy) and the expression levels of the 10 genes described in claim 2 of the samples in the glioblastoma database as covariates, and using the survival time and survival status as dependent variables, using the Cox regression model to construct a glioblastoma prognosis risk prediction model.

4. The use according to claim 3, characterized in that: The formula of the glioblastoma prognostic risk prediction model is: Among them, Gender was assigned as 0 for male and 1 for female; chemo alk was assigned as 1 for yes and 0 for no; chemo tmz was assigned as 1 for yes and 0 for no; radiation therapy was assigned as 1 for yes and 0 for no; targeted molecular therapy was assigned as 1 for yes and 0 for no; RPS3A, CLASP2, NDUFA11, KLHL23, COX7A2L, BTF3, IL4R, ARPC1B, RCC1 and CUL1 were calculated as the log2(FPKM+1) of the corresponding genes.

Images