Molecular classification of UTUC and construction of prognostic prediction model
Through the analysis of the abundance expression characteristics of circRNAs in UTUC patients, circRNAs related to prognosis were screened and prognosis prediction models were constructed, which solved the shortcomings of UTUC molecular typing and prognosis judgment, improved the accuracy of prognosis prediction, provided biomarkers, and helped Distinguish prognostic risks in different patients.
Patent Information
- Application Number
- CN202311124225.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-01
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2043-09-01
AI Technical Summary
There are few molecular typing studies on UTUC and lack of biomarkers for auxiliary diagnosis and prognosis judgment. The existing methods are mainly based on mutations and mRNA expression characteristics, and have failed to fully explore the biological processes of disease invasion, recurrence and progress.
Molecular typing is performed based on the abundance expression characteristics of UTUC patients. Through the analysis of transcriptome sequencing data, circRNAs related to prognosis were screened, multi-factor Cox proportional hazard model was constructed, UTUC prognosis prediction model was established, and circRNA characteristic risk value was used for prognosis. predict.
Three molecular subtypes of UTUC were achieved, and the accuracy of patient prognosis prediction was improved, providing biomarkers for auxiliary diagnosis and prognosis judgment, helping to distinguish prognostic risks in different patients.
Smart Images

Figure CN117558346B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of urological oncology medicine, in particular to upper urinary tract urothelial carcinoma (UTUC), and more specifically to the molecular typing of UTUC and the construction of a prognosis prediction model. Background Art
[0002] Upper tract urothelial carcinoma (UTUC) includes renal pelvic cancer and ureteral cancer, and belongs to the same category as bladder cancer. According to research reports on European and American populations, UTUC accounts for 5% to 10% of all urothelial carcinomas, while in my country, UTUC accounts for as high as 10% to 30%. The clinical and pathological characteristics of UTUC are insidious onset, high grade, strong invasiveness, and high recurrence rate. At present, there is limited research on the molecular mechanism of the occurrence and development of UTUC, and there is a lack of biomarkers that can assist in diagnosis and prognosis.
[0003] In 2017, Moss TJ et al. from the Anderson Cancer Center published a study in the journal Eur Urol (Moss TJ, Qi Y, Xi L, et al. Comprehensive genomic characterization of upper tracturothelial carcinoma [J]. European Urology, 2017, 72 (4): 641-649). They reported the unsupervised clustering results of 31 cases of UTUC based on mRNA expression characteristics, and divided UTUC into four types. There were obvious differences in mutations and tumor grades among the four types, and they showed different prognostic characteristics. In 2019, Robinson BD et al. (Robinson BD, Vlachostergios PJ, Bhinder B, et al. Upper tract urothelial carcinoma has a luminal-papillary T-cell depleted contexture and activated FGFR3 signaling. Nat Commun. 2019; 10(1): 2977) from the Center for Pathology and Laboratory Medicine at Weill Cornell Medical College described the molecular characteristics of 37 cases of high-grade UTUC. Through integrated analysis of WES and RNA-seq sequencing data, they found that the vast majority of UTUC were luminal-papillary type. In 2021, the Seishi Ogawa research team of Kyoto University in Japan (Fujii Y, Sato Y, Suzuki H, et al. Molecular classification and diagnostics of upper urinary tract urothelial carcinoma. Cancer Cell. 2021; 39(6): 793-809) comprehensively depicted the molecular pathogenesis by integrating and analyzing the gene mutations, copy number variations, DNA methylation and gene expression characteristics of 199 UTUC samples. The research team first divided UTUC into 5 types based on the gene mutation status of TP53, MDM2, RAS and FGFR3: hypermutation type (5.5%), TP53 / MDM2 type (37.7%), RAS type (HRAS / KRAS / NRAS, 15.1%), FGFR3 type (35.2%) and triple negative type (6.5%). In addition, five expression profile subtypes, C1-C5, were identified by RNA sequencing of 158 UTUC samples and performing indifferent cluster analysis.
[0004] At present, there are few studies on the molecular typing of UTUC, and they are mainly based on mutations and mRNA expression characteristics. It is urgent to integrate other dimensions of omics information to further comprehensively explore the biological processes of disease invasion, recurrence and progression. Circular RNA (circRNA) is a type of single-stranded closed circular RNA molecule, which is produced by mRNA precursor (pre-mRNA) through alternative splicing (AS, Alternative Splicing), exon cyclization or intron cyclization. Endogenous circular RNA includes coding or non-coding RNA, does not contain 5' end cap structure and 3' end poly (A), lacks free ends, and is therefore not easily degraded by nuclease exonucleases, and is more stable than linear RNA. Due to its unique structure, stability and immunogenicity, circRNA is an ideal candidate for biomarkers. Summary of the invention
[0005] One of the technical problems to be solved by the present invention is to provide a UTUC molecular typing method, which can assist in analyzing the differences in prognosis between different UTUC patients from the circRNA level.
[0006] In order to solve the above technical problems, the UTUC molecular typing method of the present invention performs molecular typing of upper urinary tract urothelial carcinoma based on the circRNA abundance expression characteristics of UTUC patients, and divides UTUC into three molecular subtypes: type I, type II and type III. The specific typing method includes the following steps:
[0007] 1) Obtain transcriptome sequencing data and clinical information of tumor tissue samples from UTUC patients;
[0008] 2) Identifying the transcriptome sequencing data, obtaining circRNA, calculating the abundance expression value of circRNA and performing standardization;
[0009] 3) Screening circRNAs that are associated with UTUC prognosis and have large changes in abundance expression values in all samples as candidate typing features;
[0010] 4) Based on the abundance expression matrix of the candidate circRNA typing features obtained in step 3), all samples were clustered to obtain the UTUC molecular typing results.
[0011] In the above step 1), the transcriptome sequencing data is a fastq format file aligned to the human reference genome hg19 and has undergone quality control. The clinical information includes disease-specific survival time and disease-specific survival status.
[0012] In the above step 2), the CPM standardization method is preferably used, and the CPM standardization conversion formula is:
[0013]
[0014] Among them, C is the number of reads mapped to gene A, and N is the total number of reads mapped to all genes.
[0015] After normalization, circRNAs with an abundance expression value of 0 in 90% of the samples were filtered out.
[0016] In the above step 3), preferably, circRNA is screened using a univariate Cox proportional hazard model and absolute median difference.
[0017] In the above step 4), preferably a consensus clustering analysis method is used.
[0018] The second technical problem to be solved by the present invention is to provide a method for screening UTUC prognosis prediction markers. After screening out the candidate typing features, the method further screens out the UTUC prognosis prediction markers according to the following steps:
[0019] 1) Perform log2 transformation on the normalized circRNA abundance expression values;
[0020] 2) All samples were randomly divided into a training set and a test set. In the training set, the log2-transformed circRNA abundance expression value obtained in step 1) was used as the feature to construct a lasso cross-validation analysis, and the optimal lambda value and the coefficient of each feature under the optimal lambda value were obtained, and the features with non-zero coefficients were retained;
[0021] 3) taking the intersection of the features obtained in step 2) and the candidate typing features, and constructing a multi-factor Cox proportional hazard model;
[0022] 4) Screen out circRNA abundance expression features that meet the PH hypothesis and have a p-value less than 0.05 as UTUC prognosis prediction markers.
[0023] The third technical problem to be solved by the present invention is to provide a group of UTUC prognosis prediction markers screened according to the above screening method, the circRNA id and host gene name of the group of markers are as follows: chr11:117115161|117117664, chr14:106208230|106235896 (host gene name: IGHV3-20, IGHV3-21), chr17:39767106|39777131 (host gene name: IKZF3), chr3:195352202|195447954 (host gene name: ACAP2, Y_RNA), chr19:3979360|3979982 (host gene name: EEF2), chr20:43954808|43955168 (host genename: TOX2), chr2:96517442|96525822, chr5:180217594|180217765, chr10:22616766|22616910(host genename: PIP4K2A), chr11:105962052|105962204(host genename: GRIA4).
[0024] The fourth technical problem to be solved by the present invention is to further construct a UTUC prognosis prediction model based on the above-mentioned UTUC prognosis prediction marker screening method. The method for constructing the prognosis prediction model includes: based on the above-screened UTUC prognosis prediction markers, establishing an optimized multifactor Cox proportional risk model, and calculating the circRNA characteristic risk value of each sample according to the coefficient of each prognosis prediction marker in the model, and then establishing the UTUC prognosis prediction model with the circRNA characteristic risk value as a variable.
[0025] The calculation formula of the circRNA characteristic risk value is:
[0026]
[0027] In the formula, S is the circRNA characteristic risk value of the sample, n is the number of UTUC prognosis prediction markers, β is the coefficient of the UTUC prognosis prediction marker, X is the abundance value of the UTUC prognosis prediction marker corresponding to the sample, and i is the number of the UTUC prognosis prediction marker.
[0028] The fifth technical problem to be solved by the present invention is to provide a UTUC prognosis prediction model constructed by the above-mentioned model construction method, wherein the prognosis prediction model comprises a nomogram, and the variables of the nomogram include circRNA characteristic risk values.
[0029] The sixth technical problem to be solved by the present invention is to provide the application of the above-mentioned UTUC molecular typing method and prognosis prediction model in the prognosis prediction of UTUC patients.
[0030] The present invention utilizes the transcriptome sequencing data of UTUC patients to analyze the circRNA abundance expression data of UTUC patients. By screening and clustering circRNAs related to UTUC prognosis, three molecular typings of UTUC based on circRNA abundance expression characteristics are obtained, and a UTUC prognosis prediction model with circRNA as a marker is further constructed. This model helps to improve the accuracy of prognosis prediction for UTUC patients. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 is the consensus clustering diagram when k=3 in the embodiment of the present invention;
[0032] Figure 2 The three molecular typing of UTUC based on circRNA in Example 1 of the present invention are significantly correlated with disease-specific survival rate;
[0033] Figure 3 is a nomogram of the prognosis prediction model constructed in Example 2 of the present invention;
[0034] Figure 4 It is the internal validation AUC value of the prognosis prediction model constructed in Example 2 of the present invention. DETAILED DESCRIPTION
[0035] In order to have a more specific understanding of the technical content, characteristics and effects of the present invention, the technical solution of the present invention is further described in detail in conjunction with the accompanying drawings and specific embodiments:
[0036] Example 1 UTUC molecular typing based on circRNA
[0037] 1. Collect circRNA abundance expression data and clinical data of UTUC
[0038] Transcriptome sequencing (RNA-seq) data of tumor tissues of 156 UTUC patients with data number EGAD00001007667 were downloaded from the EGA public database. The data type was a BMA file aligned to the human reference genome hg19. The sequencing data was converted into a fastq format file using the samtools tool. The fastq sequencing data was quality controlled using trim galore software to remove the adapter sequence and filter out low-quality fragments. For the sequencing data after quality control, the CIRIquant tool was used to identify circRNAs, and abundance quantification and CPM normalization were performed. A total of 161,510 circRNAs were identified. The CPM standardization conversion formula is:
[0039]
[0040] Among them, C is the number of reads mapped to gene A, and N is the total number of reads mapped to all genes.
[0041] After filtering out the circRNA features with expression values of 0 in 90% of the samples, 966 circRNAs remained. In addition, the clinical information (i.e., prognostic information) corresponding to the above 156 samples was collected, including disease-specific survival time and disease-specific survival status.
[0042] 2. Screening of circRNAs associated with UTUC prognosis
[0043] The above 966 circRNAs were screened using the univariate Cox proportional hazard model and the absolute median difference in turn: first, the circRNAs with a statistical test p value less than 0.05 in the univariate Cox proportional hazard model analysis were screened, and for these circRNAs related to prognosis, the absolute median difference was calculated, and the top 50 circRNAs in terms of absolute median difference were selected as candidate typing features, and 50 circRNAs related to prognosis and with large changes in abundance expression were obtained (see Table 1, 11 of which are novel circRNAs).
[0044] Table 1 50 circRNAs associated with UTUC prognosis and with significant changes in abundance expression
[0045]
[0046]
[0047]
[0048] 3. Consensus clustering to obtain molecular typing
[0049] Based on the 50 circRNAs related to UTUC prognosis obtained from the above screening, consensus clustering was performed on all patient samples using the CancerSubtypes package. The clustering algorithm was set to "pam", the distance calculation method was set to "pearson", the number of clusters was set to 2-4, and the relative change of the area under the consistent cumulative distribution function curve corresponding to types 2-4 was calculated. The K value (k = 3) with the smallest relative change was determined as the optimal cluster number, and the molecular typing results of UTUC based on circRNA were obtained (see Figure 1 Among them, 80 patients belonged to type I, 49 patients belonged to type II, and 27 patients belonged to type III.
[0050] 4. Validation of the prognostic stratification capability of molecular typing
[0051] Based on the above-mentioned UTUC molecular classification results, the prognostic differences among the three classifications were calculated using univariate Kaplan-Meier survival analysis. Figure 2 As shown, type I has the best prognosis, followed by type III, and type II has the worst prognosis. The p value of the difference between the three types in the logrank test is 0.0098, which is less than 0.05, indicating that there are significant differences in the disease-specific survival rates between UTUC patients of different types. The molecular typing of this example can distinguish the prognostic risks of UTUC patients.
[0052] Example 2 Construction of UTUC prognosis prediction model
[0053] 1. Screening of prognostic markers for UTUC
[0054] 156 samples were randomly divided into training set and test set, and the sample proportions of training set and test set were 70% and 30%, respectively. Based on the 966 circRNA abundance expression features obtained in Example 1, CPM-standardized, log2 conversion was performed. In the training set, 966 circRNA abundance expression features after log2 conversion were constructed for lasso cross-validation analysis, the optimal lambda value was obtained, and the coefficients of each feature under the optimal lambda value were obtained, and the circRNA abundance expression features with non-zero coefficients were retained. The intersection of these circRNA abundance expression features and 50 candidate typing features in Table 1 was taken to construct a multifactor Cox proportional hazard model, and the PH hypothesis (i.e., proportional hazard hypothesis) and p value in the model were screened, and 10 circRNA abundance expression features that met the PH hypothesis and had a p value less than 0.05 were retained as UTUC prognosis prediction markers, and an optimized multifactor Cox proportional hazard model was established based on these prognosis prediction markers (see Table 2).
[0055] Table 2 UTUC prognosis prediction markers and their coefficients
[0056]
[0057] 2. To construct a UTUC prognosis prediction model based on UTUC prognosis prediction markers;
[0058] According to the coefficients in the optimized multi-factor Cox proportional hazard model, the circRNA characteristic risk value of each sample was calculated respectively, and the calculation formula was:
[0059]
[0060] In the formula, S is the circRNA characteristic risk value of the sample, n is the number of UTUC prognosis prediction markers, β is the coefficient of the UTUC prognosis prediction marker, X is the abundance value of the UTUC prognosis prediction marker corresponding to the sample, and i is the number of the UTUC prognosis prediction marker.
[0061] Using the circRNA characteristic risk value as a variable, a prognostic prediction model for upper tract urothelial carcinoma was established to predict the disease-specific survival rate of patients within 1 to 5 years. The specific method is: call the nomogram function of the R package rms to establish the nomogram corresponding to the model, such as Figure 3 As shown in the nomogram, the left side shows the names of the variables used in the prediction model, including the circRNA characteristic risk value. The line segment corresponding to the variable represents the value range of the variable, which can correspond to the score and the patient's 1- to 5-year survival rate.
[0062] Internal validation of the UTUC prognostic prediction model
[0063] According to the circRNA characteristic risk value calculation formula, the circRNA prognostic prediction markers and their coefficients were substituted to calculate the circRNA characteristic risk value of each sample in the test set. The corresponding nomogram can predict the disease-specific survival risk of each sample at different time points, which is compared with the true disease-specific survival risk of the sample to verify the performance of the prognostic prediction model. Specifically, the Score function of the riskRegression package of the R language was called, and 500 cross-validations were set. Finally, the AUCs of the model predicting the disease-specific survival rate of 1 to 5 years were 0.98, 0.86, 0.88, 0.88 and 0.83, respectively (see Figure 4 ).
[0064] The above embodiments are merely feasible or preferred embodiments of the present invention and are used to illustrate the present invention rather than to limit the scope of the patent application of the present invention. Therefore, all equivalent changes and modifications made according to the scope of the patent application of the present invention should fall within the scope covered by the patent of the present invention.
Claims
1. UTUC molecular typing method, which is not used for the diagnosis and treatment of the disease, It is characterized in that Based on the abundance expression characteristics of circRNA in UTUC patients, consensus clustering analysis was performed using the CancerSubtypes package. The clustering algorithm was set to pam, the distance calculation method was set to pearson, and the number of clusters was set to 2-4. Molecular typing of upper tract urothelial carcinoma was performed, and UTUC was divided into three molecular subtypes: type I, type II, and type III. The circRNAs included the 50 circRNAs shown in the following table:
2. Screening methods for prognostic markers, It is characterized in that The following steps are involved: 1) Download the transcriptome sequencing data of UTUC patient tumor tissue samples with data number EGAD00001007667 from the EGA public database; 2) Identifying the transcriptome sequencing data, obtaining circRNA, calculating the abundance expression value of circRNA and performing standardization; 3) Screening circRNAs that are associated with UTUC prognosis and have large changes in abundance expression values in all samples as candidate typing features; 4) Log the normalized circRNA abundance expression value obtained in step 2) 2 conversion; 5) All samples were randomly divided into a training set and a test set. In the training set, the circRNA abundance expression value obtained in step 4) was used as the feature to construct a lasso cross-validation analysis, and the optimal lambda value and the coefficient of each feature under the optimal lambda value were obtained, and the features with non-zero coefficients were retained; 6) taking the intersection of the features obtained in step 5) and the candidate typing features obtained in step 3), and constructing a multi-factor Cox proportional hazard model; 7) Screen out circRNA abundance expression features that meet the PH hypothesis and have a p-value less than 0.05 as UTUC prognosis prediction markers.
3. Methods for constructing prognostic prediction models, It is characterized in that The following steps are involved: 1) Download the transcriptome sequencing data of UTUC patient tumor tissue samples with data number EGAD00001007667 from the EGA public database; 2) Identifying the transcriptome sequencing data, obtaining circRNA, calculating the abundance expression value of circRNA and performing standardization; 3) Screening circRNAs that are associated with UTUC prognosis and have large changes in abundance expression values in all samples as candidate typing features; 4) Log the normalized circRNA abundance expression value obtained in step 2) 2 conversion; 5) All samples were randomly divided into a training set and a test set. In the training set, the circRNA abundance expression value obtained in step 4) was used as the feature to construct a lasso cross-validation analysis, and the optimal lambda value and the coefficient of each feature under the optimal lambda value were obtained, and the features with non-zero coefficients were retained; 6) taking the intersection of the features obtained in step 5) and the candidate typing features obtained in step 3), and constructing a multi-factor Cox proportional hazard model; 7) Screen out circRNA abundance expression features that meet the PH hypothesis and have a p value less than 0.05 as UTUC prognosis prediction markers, establish an optimized multifactorial Cox proportional hazard model based on the prognosis prediction markers, and calculate the circRNA characteristic risk value of each sample according to the coefficients in the model; the calculation formula of the circRNA characteristic risk value is: In the formula, S is the circRNA characteristic risk value of the sample, n is the number of UTUC prognosis prediction markers, β is the coefficient of the UTUC prognosis prediction marker, X is the abundance value of the UTUC prognosis prediction marker corresponding to the sample, and i is the number of the UTUC prognosis prediction marker; 8) Using the circRNA characteristic risk value as a variable, the nomogram function of the R package was called to establish a UTUC prognosis prediction model.
4. The method according to claim 2 or 3, It is characterized in that Step 1), the transcriptome sequencing data is a fastq format file aligned to the human reference genome hg19 and is quality controlled.
5. The method according to claim 2 or 3, It is characterized in that Step 2), using the CPM standardization method, the CPM standardization conversion formula is: Among them, C is the number of reads mapped to gene A, and N is the total number of reads mapped to all genes.
6. The method according to claim 2 or 3, It is characterized in that Step 2), after normalization, filter out circRNAs with an abundance expression value of 0 in 90% of the samples.
7. The method according to claim 2 or 3, It is characterized in that Step 3) CircRNAs were screened using the univariate Cox proportional hazard model and absolute median difference.
8. The method according to claim 2 or 3, It is characterized in that Step 3), the candidate typing features include 50 circRNAs shown in the following table:
9. The method according to claim 2 or 3, It is characterized in that Step 7), the UTUC prognosis prediction markers include the 10 circRNAs shown in the following table:
10. The method according to claim 3, It is characterized in that The following steps are also included: 9) Based on the UTUC prognosis prediction markers and their coefficients, the circRNA signature risk value of each sample in the test set was calculated to internally verify the performance of the UTUC prognosis prediction model.
11. UTUC prognosis prediction markers, It is characterized in that Including the 10 circRNAs shown in the table below:
12. UTUC prognosis prediction model, It is characterized in that The prognostic prediction model comprises a nomogram, the variables of the nomogram comprise a circRNA characteristic risk value, and the calculation formula of the characteristic risk value is: In the formula, S is the circRNA characteristic risk value of the sample, n is the number of UTUC prognosis prediction markers, β is the coefficient of the UTUC prognosis prediction marker, X is the abundance value of the UTUC prognosis prediction marker corresponding to the sample, and i is the number of the UTUC prognosis prediction marker; The UTUC prognosis prediction markers and their coefficients include 10 circRNAs and their corresponding coefficients shown in the following table: The nomogram was constructed using the nomogram function of the R package.
13. Use of the UTUC molecular typing method of claim 1 in predicting the prognosis of UTUC patients, wherein the use is not for the purpose of diagnosis and treatment of the disease.
14. Use of the prognostic prediction model of claim 12 in predicting the prognosis of UTUC patients, wherein the use is not for the purpose of diagnosis and treatment of the disease.
