Peripheral blood-based five-gene biomarker for predicting non-small cell lung cancer immunotherapy curative effect, kit and application of peripheral blood-based five-gene biomarker
A 5-gene biomarker model for NSCLC using UQCRB, NDUFA3, CDKN2D, FMNL1-DT, and APOL3 enhances prediction of immunotherapy response, improving specificity and sensitivity, and offering comprehensive, non-invasive monitoring of immune status.
Patent Information
- Application Number
- CN202411491177.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-24
- Publication Date
- 2024-12-03
Smart Images

Figure CN119061147A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of gene detection, and in particular, to a 5-gene biomarker for predicting the efficacy of immunotherapy for non-small cell lung cancer based on peripheral blood, a kit and applications thereof. Background Art
[0002] In the treatment of non-small cell lung cancer (NSCLC), immune checkpoint inhibitors (such as anti-PD-1 / PD-L1 and CTLA-4) have become an important treatment. Immune checkpoint therapy activates the body's attack on tumors by relieving the immune system's suppression of tumor cells. However, since there are significant differences in patients' responses to these treatments, it has become particularly important to predict which patients will respond well to treatment. Currently, PD-L1 expression and TMB are the two main biomarkers used to predict the response of NSCLC patients to immune checkpoint therapy. The detection of PD-L1 expression is based on immunohistochemical analysis of tumor tissue, which predicts the treatment effect by evaluating the expression level of PD-L1 protein on the surface of tumor cells. Tumor mutation burden (TMB) is another predictive criterion, which evaluates the number of mutations in tumor tissue through sequencing technology. High TMB is generally associated with a good response to immunotherapy. However, these methods have great limitations:
[0003] 1. Insufficient specificity and sensitivity: Current biomarkers used to predict the response of NSCLC patients to immune checkpoint therapy, such as PD-L1 expression and TMB, have problems with insufficient specificity and sensitivity, which means that these methods cannot accurately distinguish all potential treatment responders and non-responders.
[0004] 2. Large individual differences: PD-L1 expression and TMB, as single biomarkers, fail to fully consider the biological differences between patients, resulting in limited universality and accuracy of the prediction results.
[0005] 3. Single effect evaluation: These technologies mainly rely on the evaluation of a single or a few factors, failing to fully reflect the complex biological characteristics and immune microenvironment of the tumor, thus affecting the comprehensiveness and accuracy of the prediction results.
[0006] Therefore, it is particularly important to develop prediction methods that are universal and highly accurate. Summary of the invention
[0007] In view of the shortcomings of the prior art, the present invention aims to overcome the limitations of the prior art in predicting the response to NSCLC immune checkpoint therapy by introducing a new gene model, thereby achieving a more accurate and comprehensive prediction of the patient's treatment response. To this end, the present invention provides the following technical solutions:
[0008] First, in a first aspect, the present invention provides an application of a reagent for detecting a biomarker combination in the preparation of a kit for predicting the effect of immunotherapy for non-small cell lung cancer, wherein the biomarker combination includes UQCRB, NDUFA3, CDKN2D, FMNL1-DT and APOL3.
[0009] In a preferred embodiment, the biomarker combination further includes: PD-L1 and TMB.
[0010] In one embodiment, the immunotherapy is anti-PD(L)-1 immunotherapy.
[0011] In one embodiment, the reagent is selected from: a primer that specifically amplifies the biomarker, a probe that specifically recognizes the biomarker, and / or an immunochemical reagent.
[0012] The second aspect of the present invention provides a product for diagnosing the prognosis of immunotherapy for non-small cell lung cancer, the product comprising a reagent for detecting the expression level of biomarkers, wherein the biomarkers are UQCRB, NDUFA3, CDKN2D, FMNL1-DT and APOL3.
[0013] In one embodiment, the product is a kit, a chip or a nucleic acid membrane strip capable of detecting the expression level of the biomarker.
[0014] The third aspect of the present invention provides an immunotherapy prediction model, wherein the model is:
[0015] Risk Score=0.92*UQCRB+0.84*NDUFA3-0.42*CDKN2D-0.63*APOL3-0.54*
[0016] FMNL1-DT
[0017] Each gene symbol represents the expression value of the gene. The method used is to calculate based on the numerical value of each gene expression (TPM (Transcripts Per Million)). First, the raw readings of each gene are obtained by RNA sequencing; then, these readings are normalized to the expected number of fragments per kilobase (RPK) by dividing the readings of each gene by its transcript length (in kilobases); finally, the RPK values of all genes are added to get the total value, and then the RPK value of each gene is divided by this total value and multiplied by one million to get TPM (Transcripts Per Million). Then select the expression values of the required 5 target genes from all genes. For example, assuming that the TPM values of the five genes are UQCRB=10.5, NDUFA3=8.2, CDKN2D=3.4, APOL3=5.1, and FMNL1-DT=2.9, then Risk Score=0.92*10.5+0.84*8.2-0.42*3.4-0.63*5.1-0.54*2.9=9.66+6.89-1.43-3.21-1.57=10.34.
[0018] A fourth aspect of the present invention provides a non-small cell lung cancer immunotherapy prognosis diagnosis system, the diagnosis system comprising:
[0019] (1) Detection component: The detection component is used to detect the expression levels of biomarkers UQCRB, NDUFA3, CDKN2D, FMNL1-DT and APOL3;
[0020] (2) Result judgment component: The result judgment component is used to output whether the subject responds to immune node treatment based on the expression level results of the biomarkers UQCRB, NDUFA3, CDKN2D, FMNL1-DT and APOL3 detected by the detection component, and according to the diagnostic model according to claim 5.
[0021] Compared with the prior art, the present invention has the following significant improvements:
[0022] 1. Improve the specificity and sensitivity of prediction: By introducing a model based on 11 genes, the present invention aims to improve the specificity and sensitivity of predicting the response of NSCLC patients to immune checkpoint therapy. The data showed that the model based on 11 genes had an AUC of 0.89-0.90 in the retrospective cohort and an AUC of 0.89 in the prospective validation cohort, which was significantly better than a single marker.
[0023] 2. Considering individual differences: The present invention considers the comprehensive expression of multiple genes, which can better reflect the biological differences between different patients, thereby improving the accuracy and universality of the prediction.
[0024] 3. Use peripheral blood samples to achieve dynamic and non-invasive monitoring of immune status. Dynamic monitoring of immune status is simple to operate, has high patient compliance, and can guide clinical decision-making more comprehensively. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:
[0026] Figure 1 is the relationship between the AUC value and the number of features in the recursive feature elimination process;
[0027] Figure 2 The model for peripheral blood expression and immune prediction of 5 genes in patients with non-small cell lung cancer was established and evaluated. A shows the receiver operating characteristic (ROC) curves of the 5 key genes in responders and non-responders, which is used to demonstrate the predictive efficacy of the model in the PBMCR cohort, including the area under the curve (AUC) and its 95% confidence interval. B is the survival curve analysis of the responders and non-responders predicted by the immune prediction model.
[0028] Figure 3 A model for peripheral blood expression and immune prediction of 5 genes in patients with non-small cell lung cancer was established and evaluated. A shows the receiver operating characteristic (ROC) curves of the 5 key genes in responders and non-responders, which is used to demonstrate the predictive efficacy of the model in the BC cohort, including the area under the curve (AUC) and its 95% confidence interval. B is a survival curve analysis of the responders and non-responders predicted by the immune prediction model. DETAILED DESCRIPTION
[0029] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.
[0030] Example 1 Sample collection and RNA extraction
[0031] 1. Sample source:
[0032] Three independent NSCLC patient cohorts were included: retrospective PMBCR cohort (as test set 1, n = 59), BC cohort (as test set 2, n = 44), and prospective PBMCP cohort (as validation set, n = 42). All patients were histologically confirmed to have NSCLC and were planned to receive immune checkpoint inhibitors. All patients had signed informed consent.
[0033] 2. Sample processing:
[0034] PBMC sample collection: 59 patients with advanced NSCLC who were scheduled to receive immunotherapy were enrolled. Before treatment, 8-10 ml of peripheral venous blood was drawn, anticoagulated with EDTA, and stored at 4°C. PBMCs were separated by density gradient centrifugation, washed twice with PBS, and the cell pellets were aliquoted and stored at -80°C.
[0035] BC sample collection: 44 patients with advanced NSCLC who were to receive immunotherapy were enrolled. Before treatment, 8-10 ml of peripheral venous blood was drawn, anticoagulated with EDTA, and stored at 4°C. The blood was centrifuged at 3000 rpm for 10 min, the BC layer rich in white blood cells was aspirated, washed twice with PBS, and the cell pellet was aliquoted and stored at -80°C.
[0036] Total RNA extraction: TRIzol method was used to extract total RNA from PBMC and BC, and the concentration and purity were detected by NanoDrop (A260 / A280: 1.8-2.0). The quality was qualified by NanoDrop / Agilent 2100 detection.
[0037] Example 2 RNA sequencing and data analysis
[0038] 1. Sequencing Library Preparation
[0039] Use the Illumina TruSeq RNA Sample Preparation Kit to prepare sequencing libraries according to the manufacturer's instructions. The specific steps are as follows:
[0040] 1. Take 1 μg of total RNA for mRNA enrichment and fragmentation. Use magnetic beads combined with poly-T oligonucleotide primers to enrich mRNA and randomly cut mRNA into 200-500 bp fragments.
[0041] 2. Reverse transcription to synthesize first-strand cDNA. Use random primers and reverse transcriptase to generate first-strand cDNA.
[0042] 3. Synthesis of second-strand cDNA: DNA polymerase I and RNase H are used to degrade the RNA template and generate the second-strand cDNA.
[0043] 4. End repair, A tail addition and adapter ligation. The double-stranded cDNA is end-repaired, an A base is added to the 3' end, and then an adapter with a T base overhang is ligated.
[0044] 5. PCR amplification and library purification: Use primers for PCR amplification and purify the PCR product with AMPure XP magnetic beads to obtain the final sequencing library.
[0045] 6. Library quality control: Agilent 2100 and Q-PCR were used to evaluate the quality and concentration of the library.
[0046] 2. RNA Sequencing
[0047] Paired-end 150 bp (PE150) sequencing was performed using the Illumina NovaSeq 6000 platform according to the standard protocol provided by the manufacturer.
[0048] 3. Raw Data Processing
[0049] 1. Use Trimmomatic v0.39 software to perform quality control and filtering on raw reads based on the following parameters: ILLUMINACLIP:adapters.fa:2:30:10LEADING:3TRAILING:3SLIDINGWINDOW:4:15MINLEN:36.
[0050] 2. The quality of the filtered data was assessed using FastQC v0.11.9.
[0051] 4. Data comparison and expression quantification
[0052] 1. Use HISAT2 v2.1.0 to align the quality-controlled reads to the human reference genome GRCh38.
[0053] 2. Use featureCounts v2.0.1 with default parameters to count the number of reads mapped to gene regions and generate a gene read count matrix.
[0054] 5. Data Normalization and Differential Expression Analysis
[0055] 1. In the R 4.0 environment, the edgeR package v3.32.0 was used to perform TMM (trimmed mean of M-values) standardization on the raw counts and calculate the TPM (transcript fragments per million) expression value of each gene.
[0056] 2. The quasi-likelihood (QL) generalized linear model method in edgeR was used for differential expression analysis. A comparison group was set up to screen the differentially expressed genes with FDR < 0.05 and |log2FC| ≥ 1.
[0057] Example 3 Development of an immunotherapy prediction model
[0058] 1. Characteristic gene screening:
[0059] 1. Initial screening: We first performed a T test on the gene expression of the response group and the non-response group in the retrospective dataset PMBCR. Based on the T test results, LASSO regression was used to preliminarily screen candidate differentially expressed genes with P < 0.05 (a total of 1034 significant genes). LASSO regression can automatically perform feature selection by introducing the L1 regularization term to obtain a sparse feature set. In this step, we obtained a preliminary selection set containing dozens of potential feature genes.
[0060] 2. Recursive feature elimination: Based on the initial gene set, the recursive feature elimination (RFE) method is further used to optimize the feature set. RFE recursively removes genes with low feature importance and evaluates model performance to obtain the optimal feature combination.
[0061] The specific steps are as follows:
[0062] (1) Train the LASSO regression model based on the current feature set and evaluate its performance (such as AUC value) through cross-validation.
[0063] (2) Based on the feature weights of LASSO regression, a certain proportion (e.g., 10%) of the genes with the lowest importance are removed.
[0064] (3) Repeat steps (1) and (2) using the updated feature set until the preset termination condition is met, such as the number of features reaching a threshold or the model performance no longer improves significantly ( Figure 1 ).
[0065] 3. Stability assessment: In order to evaluate the stability and repeatability of the feature selection results, we used a nested cross-validation method. In the outer cross-validation, the data was divided into multiple mutually exclusive training sets and test sets. On each training set, we independently performed the complete feature selection process (including inner cross-validation) and recorded the selected feature set. We compared the features selected on different training sets to evaluate their consistency.
[0066] 4. Final results: Through the above rigorous screening process, we identified 5 key genes most relevant to the efficacy of immunotherapy from 1034 candidate genes, including UQCRB, NDUFA3, CDKN2D, FMNL1-DT and APOL3, and used these 5 genes to form the final prediction model feature set.
[0067] 2. Prediction model construction and verification:
[0068] The five characteristic genes screened by LASSO regression were used to construct an immunotherapy efficacy prediction model. The risk score calculation formula of the model is:
[0069] Risk Score=0.92*UQCRB+0.84*NDUFA3-0.42*CDKN2D-0.63*APOL3-0.54*FMNL1-DT
[0070] Among them, each gene symbol represents the expression value of the gene. The positive coefficients before UQCRB and NDUFA3 indicate that they are positively associated with risk, while the negative coefficients of the other three genes indicate that they are negatively associated with risk.
[0071] The effects of characteristic genes are superimposed in a weighted summation manner. The absolute value of the coefficient reflects the degree of influence of the corresponding gene on the risk, and the sign indicates the direction of the influence (positive or negative).
[0072] The risk score of each sample is calculated according to the above formula. Specifically, the calculation is performed based on the numerical value of each gene expression (TPM (Transcripts Per Million)). First, the raw readings of each gene are obtained by RNA sequencing; then, these readings are normalized to the expected number of fragments per kilobase (RPK) by dividing the readings of each gene by its transcript length (in kilobases); finally, the RPK values of all genes are added together to obtain a total value, and then the RPK value of each gene is divided by this total value and multiplied by one million to obtain TPM (Transcripts Per Million). Then the expression values of the 5 target genes in the formula are selected from all genes. For example, assuming that the TPM values of the five genes are UQCRB=10.5, NDUFA3=8.2, CDKN2D=3.4, APOL3=5.1, and FMNL1-DT=2.9, then Risk Score=0.92*10.5+0.84*8.2-0.42*3.4-0.63*5.1-0.54*2.9=9.66+6.89-1.43-3.21-1.57=10.34.
[0073] The immune prediction score is used to determine whether the patient responds to immunotherapy. If the immune prediction score is greater than or equal to zero, it is considered a response (the patient has complete remission (CR) or partial remission (PR)), and if the immune prediction score is less than zero, it is considered a non-response (the patient has stable disease (SD) or progressive disease (PD)).
[0074] These five key genes are involved in multiple biological processes closely related to the occurrence and development of tumors, such as energy metabolism, cell cycle regulation, and immune response. By integrating the expression profiles of multiple key genes, the characteristics of the tumor immune microenvironment can be more comprehensively reflected, which helps to accurately predict the efficacy of immunotherapy. Next, the five-gene risk score (linear combination of regression coefficient × expression level) of each patient is calculated, and then a efficacy prediction model is constructed using Logistic regression.
[0075] The ROC curve was used to evaluate the predictive efficacy, and the AUC of the PMBCR cohort reached 0.90. The Bootstrap method was used for internal validation, and the specific steps were as follows: Repeated sampling 1000 times from the original data set to generate multiple subsets of the same size as the original data set, train the model on each subset, and use the samples that were not sampled (ie, "external data") to evaluate the model performance. The AUC value of each iteration was calculated through multiple iterations, and the average and confidence interval of these values were taken to correct the model performance. The results showed that the original AUC of the PMBCR cohort was 0.90, and the corrected AUC was 0.88 (95% CI, 0.78-0.95). Kaplan-Meier analysis showed that the PFS of the high and low 5-gene risk score groups was significantly different (3.8 months vs 12.6 months, P<0.0001) ( Figure 2 ). The established 5-gene prediction model was applied to the BC cohort, and the ROC curve showed an AUC of 0.89. The bootstrap method was internally validated, and the adjusted AUC was 0.86 (95% CI, 0.73-0.94). Kaplan-Meier analysis confirmed that the PFS (5.6 months vs. not reached) between the high and low 5-gene risk score groups was significantly different (P<0.001) ( Figure 3 ).
[0076] To further confirm the universality of the model, the PBMCP cohort was used to externally validate the model generalization ability, and the AUC reached 0.89. Kaplan-Meier survival analysis showed significant differences in PFS between the high-risk and low-risk groups (4.2 months vs 13.8 months, P = 0.005). Cox regression confirmed that the risk score was an independent prognostic factor (HR = 3.45, P < 0.001).
[0077] Example 45 Comparison of the predictive effects of the gene model and existing traditional biomarkers (PD-L, TMB)
[0078] The patients in the validation set described in Example 1 (n=42) were tested for PD-L1 expression (IHC method) and TMB (next-generation sequencing), and the ROC curves and AUC values of the three in predicting the efficacy of immunotherapy were compared. PD-L1 detection uses conventional IHC methods to evaluate the PD-L1 expression level of tumor cells, and the sampling volume is the tumor tissue section of each patient; TMB detection uses second-generation sequencing technology to analyze the mutation load of the tumor genome, and the sampling volume is the tumor tissue DNA of each patient. The results showed that the predictive efficiency of the 5-gene signature (AUC=0.89) was significantly better than PD-L1 (AUC=0.58) and TMB (AUC=0.62). In order to further improve the predictive efficiency, an integrated model was constructed by combining the three. The specific method is to combine the prediction results of PD-L1, TMB and 5-gene signatures, and conduct a comprehensive analysis through a multivariate logistic regression model. Finally, the AUC in the PMBCP cohort was increased to 0.91, but the difference with the use of the 5-gene signature alone (AUC=0.89) was not significant.
[0079] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. Use of a reagent for detecting the expression level of a biomarker in a sample in the preparation of a product for diagnosing the prognosis of immunotherapy for non-small cell lung cancer, characterized in that: The biomarkers are UQCRB, NDUFA3, CDKN2D, FMNL1-DT and APOL3.
2. The use according to claim 1, characterized in that: The reagent is selected from: a primer that specifically amplifies the biomarker, a probe that specifically recognizes the biomarker, and / or an immunochemical reagent.
3. A product for diagnosing the prognosis of immunotherapy for non-small cell lung cancer, characterized in that: The product comprises reagents for detecting the expression levels of biomarkers, which are UQCRB, NDUFA3, CDKN2D, FMNL1-DT and APOL3.
4. The product according to claim 3, characterized in that The product is a kit, a chip or a nucleic acid membrane strip capable of detecting the expression level of the biomarker.
5. A prognostic diagnostic model for immunotherapy of non-small cell lung cancer, characterized in that: The model is: Risk Score=0.92*UQCRB+0.84*NDUFA3-0.42*CDKN2D-0.63*APOL3-0.54* FMNL1-, Among them, each gene represents the TPM (Transcripts Per Million) value of the gene; The model is used in the following way: first, the raw reads of each gene are obtained through RNA sequencing; then, these reads are standardized to the expected number of fragments per kilobase (RPK) by dividing the reads of each gene by its transcript length (in kilobases); finally, the RPK values of all genes are added together to obtain a total value, and then the RPK value of each gene is divided by this total value and multiplied by one million to obtain the TPM of the gene. The calculated immune prediction score is used to determine whether the patient responds to immune node treatment. The judgment criteria are: if the immune prediction score is greater than or equal to zero, there is a response; if the immune prediction score is less than zero, there is no response.
6. A non-small cell lung cancer immunotherapy prognosis diagnosis system, characterized in that: The diagnostic system comprises (1) Detection component: The detection component is used to detect the expression levels of biomarkers UQCRB, NDUFA3, CDKN2D, FMNL1-DT and APOL3; (2) Result judgment component: The result judgment component is used to output whether the subject responds to immune node treatment based on the expression level results of the biomarkers UQCRB, NDUFA3, CDKN2D, FMNL1-DT and APOL3 detected by the detection component, and according to the diagnostic model according to claim 5.
