A kit for predicting the efficacy of chemotherapy for head and neck squamous cell carcinoma and application thereof

Abstract

The present application relates to the application of reagents for detecting the expression amount of lipid metabolism genes in the preparation of a kit for predicting the prognosis model of the efficacy of chemotherapy for head and neck squamous cell carcinoma, wherein the lipid metabolism genes are selected from the following eight genes: ACSBG2, APOB, IKBKB, MAPK9, MOGAT2, PLA2G10, PIK3R3 and SREBF1. The present application also provides a prognosis model and a kit for predicting the efficacy of chemotherapy for head and neck squamous cell carcinoma. The prediction model is constructed according to the expression amount of the eight genes and combined with clinical information, and appropriate evaluation and selection criteria are provided. The score of the model is =(-2.7486)*ACSBG2+(1.7158)*APOB+(-0.3216)*IKBKB+(0.4612)*MAPK9+(-0.8421)*MOGAT2+(0.6413)*PLA2G10+(-0.2157)*PIK3R3+(-0.2355)*SREBF1. In addition, the LMRS model not only has excellent diagnostic efficiency for the efficacy evaluation and overall survival prognosis diagnosis of systemic treatment drugs for HNSCCs, but also has an important relationship with the distribution and expression of immune cells, and has potential prognostic value for HNSCC patients, providing important reference value for new drug research and development.

Description

Technical Field

[0001] This invention relates to the field of reagent kit technology, and more specifically, to a reagent kit for predicting the efficacy of chemotherapy drugs for head and neck squamous cell carcinoma and its application. Background Technology

[0002] Head and neck squamous cell carcinoma (HNSCC) is a group of malignant tumors occurring in the head and neck region, accounting for approximately 90% of all head and neck tumors. HNSCC includes neck tumors, oral and maxillofacial tumors, and ear, nose, and throat tumors, and is more common in women. Although HPV (Human Papillomavirus)-positive HNSCC patients have a relatively good prognosis, with an overall survival (OS) of 70%, stage III-IV patients still experience extensive local invasion and treatment failure, with a poor 5-year OS of approximately 40%. Treatment for HNSCC varies depending on pathological features or clinical stage and includes surgical treatment, concurrent chemoradiotherapy, targeted therapy, and immunotherapy. However, the extensive tissue resection and reconstruction, as well as the side effects of chemoradiotherapy, can severely impair swallowing and respiratory function, significantly impacting the quality of life and survival rate of HNSCC patients.

[0003] With the advent of immunosuppressants, the role of systemic therapy in head and neck tumors is constantly increasing, and the overall survival rate of HNSCCs has been significantly improved. Patients sensitive to chemotherapy (cisplatin as the primary first-line chemotherapy drug) not only have the opportunity to preserve their larynx but also achieve better overall survival and quality of life. However, most HNSCC patients are diagnosed at an advanced stage, and even with a personalized comprehensive treatment plan, treatment failure still occurs. The most common reason for treatment failure is drug resistance. Existing treatment options are limited after platinum-based resistance develops. Apart from PD-1 / PD-L1 (Programmed Cell Death 1 / Programmed Cell Death 1Ligand 1) and EGFR (Epidermal Growth Factor Receptor), there are very few other targeted drugs available. Therefore, early prediction of the efficacy of chemotherapy-based systemic therapy in HNSCC after a definitive diagnosis is of great value. Identifying potential key drug resistance molecules and mechanisms will also provide an important reference basis for the formulation of comprehensive treatment plans for HNSCC patients and the development of new drugs.

[0004] Metabolic remodeling is a key characteristic of cancer, and it plays an increasingly important role in hepatocellular carcinoma (HNSCC). Besides the Warburg effect and glutamine metabolism, lipid metabolism remodeling also significantly impacts the proliferation, metastasis, and recurrence of HNSCC tumor cells. Most lipid metabolism enzymes are elevated in HNSCC and are associated with poor prognosis. However, their impact and potential role in systemic drug administration in HNSCC remain poorly studied. Therefore, this invention analyzes the influence of lipid metabolism remodeling on systemic drug administration in HNSCC and constructs a lipid metabolism characteristic prognostic model (LMRS model).

[0005] Currently, there are no reports on the prognostic model of lipid metabolism characteristics constructed in this invention. Summary of the Invention

[0006] The first objective of this invention is to provide a kit for predicting the efficacy of chemotherapy drugs in head and neck squamous cell carcinoma, addressing the shortcomings of the prior art.

[0007] The second objective of this invention is to provide the application of reagents for detecting lipid metabolism gene expression levels in the preparation of prognostic model kits.

[0008] The third objective of this invention is to provide a prognostic model system for predicting the efficacy of chemotherapy in head and neck squamous cell carcinoma.

[0009] To achieve the first objective mentioned above, the technical solution adopted by the present invention is as follows:

[0010] A kit for predicting the prognostic model of chemotherapy efficacy in head and neck squamous cell carcinoma, the kit comprising reagents for detecting lipid metabolism gene expression levels and a lipid metabolism characteristic risk score.

[0011] More preferably, the lipid metabolism gene is a combination of the following eight genes: ACSBG2, APOB, IKBKB, MAPK9, MOGAT2, PLA2G10, PIK3R3, and SREBF1.

[0012] More preferably, the kit is a risk score containing lipid metabolism characteristics: Risk score = (-2.7486)*ACSBG2+(1.7158)*APOB+(-0.3216)*IKBKB+(0.4612)*MAPK9+(-0.8421)*MOGAT2+(0.6413)*PLA2G10+(-0.2157)*PIK3R3+(-0.2355)*SREBF1.

[0013] To achieve the second objective mentioned above, the technical solution adopted by the present invention is as follows:

[0014] The application of reagents for detecting lipid metabolism gene expression levels in the preparation of a kit for predicting the prognostic effect of chemotherapy in head and neck squamous cell carcinoma. The lipid metabolism genes are a combination of the following eight genes: ACSBG2, APOB, IKBKB, MAPK9, MOGAT2, PLA2G10, PIK3R3, and SREBF1. The prognostic model is a risk score including lipid metabolism features: Risk score = (-2.7486)*ACSBG2 + (1.7158)*APOB + (-0.3216)*IKBKB + (0.4612)*MAPK9 + (-0.8421)*MOGAT2 + (0.6413)*PLA2G10 + (-0.2157)*PIK3R3 + (-0.2355)*SREBF1.

[0015] To achieve the third objective mentioned above, the technical solution adopted by the present invention is as follows:

[0016] A prognostic model system for predicting the efficacy of chemotherapy in head and neck squamous cell carcinoma, the system comprising information acquisition, risk score calculation, and prognostic assessment.

[0017] More preferably, the information acquisition refers to acquiring gene expression information of each head and neck squamous cell carcinoma patient.

[0018] More preferably, the risk score is calculated by substituting the expression level of lipid metabolism mRNA into the prognostic model formula.

[0019] More preferably, the prognostic assessment is based on grouping patients according to risk scores, with a median of -2.15128 for the low-score group and a median of -1.09682 for the high-score group.

[0020] More preferably, the prognostic model system can also be used to predict immune cell infiltration and immune checkpoint expression.

[0021] The advantages of this invention are:

[0022] 1. A better understanding of the origins and interrelationships of heterogeneity in HNSCC is a key objective in head and neck oncology, and has broad implications for diagnosis and treatment.

[0023] 2. Our invention further supports the role of lipid metabolism remodeling, particularly fatty acid-related metabolism, in the survival prognosis of HNSCC patients, especially in their response to systemic therapy. The model achieved satisfactory results by integrating risk score, age, sex, tumor stage, pathological type, and smoking status. It possesses both short-term and medium-term diagnostic value, exhibiting better diagnostic thresholds compared to commonly used clinical indicators such as sex and age.

[0024] 3. The model detection panel established by this invention provides appropriate assessment and selection criteria to screen out patients who can obtain the greatest benefit from systemic treatment and allow patients who are unlikely to benefit to quickly transition to other therapies. Attached Figure Description

[0025] Appendix Figure 1 This is the design flowchart for this method.

[0026] Appendix Figure 2 The results show a significant enrichment of lipid-related genes after screening the CRISPR / Cas9 library.

[0027] Appendix Figure 3 To classify the prognostic outcomes of systemic treatment for HNSCC patients using lipid metabolism genes.

[0028] Appendix Figure 4 To establish and evaluate diagnostic prognostic models for eight lipid-related genes.

[0029] Appendix Figure 5 The results are used to assess the prognosis of each gene and the DCA diagnostic curve.

[0030] Appendix Figure 6 This is the result of an immune-related assessment of a gene model.

[0031] Appendix Figure 7 The results of model validation and functional analysis of HNSCC data in the GEO database.

[0032] Appendix Figure 8 This serves as a validation result for the prognostic diagnosis of the LMRS model in clinical specimens. Detailed Implementation

[0033] As a preferred example, the following embodiments are in accordance with the appendix. Figure 1 The steps in the design flowchart are performed. The invention is further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the description of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0034] It should be noted that the sequencing service in this embodiment was performed by Anson Technology Co., Ltd., and the sequenced genes were sgRNA sequences of the entire human genome, contained in the GECKO 2.0 library. The gene values ​​in the risk score were the mRNA expression values ​​from the TCGA database. The TPM format data of each gene were extracted from the database, then normalized to log2(TPM+1), and finally, samples with RNAseq data and clinical information were retained.

[0035] Example 1

[0036] 1. Materials and Methods

[0037] 1.1 Instruments and Reagents

[0038] instrument

[0039] Fluorescence microscopy Zeiss zen3.3 incubator Ultra-clean bench

[0040] reagents

[0041]

[0042]

[0043] 1.2 Model Establishment

[0044] 1.2.1 CRISPR / Cas9 library screening

[0045] CRISPR / Cas9 library screening was performed according to the reference (Tian X, Wang X, Cui Z, Liu J, Huang X, Shi C, et al. A Fifteen-Gene Classifier to Predict Neoadjuvant Chemotherapy Responses in Patients with Stage IB to IIB Squamous Cervical Cancer. Advancedscience (2021) 8(10): 2001-978.). Lentiviral products were obtained from Heyuan Biotechnology (Shanghai) Co., Ltd. To determine the drug resistance of cells after gene suppression, we screened the hypopharyngeal cancer cell lines Fadu and Detriot-562 (purchased from the Cell Bank of the Chinese Academy of Sciences) from human primary hypopharyngeal carcinoma and human pleural effusion metastasis. 500,000 cells were seeded in 100 mm culture dishes, starved for 12 hours the next day, and transfected with viral libraries at infection efficiencies (MOI) of 0.3, 0.4, and 0.5. 12-18 hours later, cells were screened for 3 days with 2ug / ml puromycin to identify cells with successfully knocked-out genes. The knocked-out genes were divided into control and drug screening groups, with the IC20 concentration of the drug used as the screening concentration: cisplatin (Detroit-562: 5ug / ml; Fadu: 1ug / ml) and 5-FU (Detroit-562: 5ug / ml; Fadu: 1ug / ml), with 3 replicates per group. Fresh high-glucose DMEM medium containing the drugs was changed every 3 to 5 days. After 2 weeks, cells were collected to extract gDNA for PCR amplification, and the sgRNA coding region was sequenced for analysis. MAGeCK (v0.5.7) was used to align the sgRNA coding primers with the human reference genome (hg19). MAGeCK's robust rank aggregation (RRA) analysis was used to analyze the differential sgRNAs among the cisplatin, 5-FU screening, and control groups. The selection criteria for enriched sgRNAs for further analysis were as follows: 1. All independently replicated sgRNAs in the DDP and 5-FU groups with an FDR < 0.25 and a p-value < 0.05, or 2. Any independently replicated sgRNA in the DDP and 5-FU groups with a fold change < 0.05 and a p-value < 0.05. sgRNA data analysis after CRISPR / Cas9 library screening was performed by AzentaLife Science.

[0046] 1.2.2 CRISPR / Cas9 library screening

[0047] In the R language environment, the cluster profile (version 3.14.3) and org.Hs.eg.db (version 3.10.0) packages were used to perform GO and KEGG enrichment analyses on differentially expressed sgRNAs screened from CRISPR / Cas9 libraries to analyze differences in biological processes (BP), cellular components (CC), molecular functions (MF), and pathways. Multiple tests and corrections were performed using the BH method.

[0048] 1.2.3 TCGA and GEO Data Acquisition

[0049] Download clinical data, TCGA RNA-seq data, and probe annotation files for HNSCC patients from the TCGA dataset (https: / / portal.gdc.com), excluding samples without clinical data. Use the R package "GEOquery" to download the GSE32877 and GSE10300 datasets from the Geneexpression Omnibus (GEO) database in R.

[0050] 1.2.4 Construction of Gene Prediction Model

[0051] Gene count data from TCGA (The Cancer Genome Atlas) HNSCC (Head and neck squamous cell cancer) patients receiving systemic therapy were converted to TPM (Total Productive Motion), and the data were normalized to log2(TPM+1), while samples containing clinical information were retained. Finally, a total of 173 samples were available for subsequent analysis.

[0052] 1.2.5 Subgroup Analysis

[0053] ConsensusClusterPlus R package (v1.54.0) was used for consistency analysis, with two clusters. 80% of the total samples were plotted 100 times, with clusterAlg="hc" and innerLinkage="ward.D2". Cluster heatmaps were generated using the R package pheatmap (v1.0.12). Gene expression heatmaps retained genes with an expression SD > 0.1. If the number of input genes exceeded 1000, the top 25% of genes were extracted after sorting by SD. Principal component analysis was performed using the prcomp function in R.

[0054] 1.2.6 Construction of the LMRS (Lipid Metabolism Related Score) Feature Model

[0055] A total of 751 lipid metabolism-related genes were identified between the two drug screening groups. Based on the CRISPR / Cas9 library screening results, only the top 50 lipid metabolism-related genes were used. Feature selection was then performed using the Least Absolute Contraction Selection Operator (LASSO) regression algorithm and 10-fold cross-validation, and the R package glmnet was used for analysis. Finally, a risk score for each patient was calculated using a formula derived from the weighted analysis of regression coefficients, and the samples were divided into high-risk and low-risk groups based on the median risk score.

[0056] 1.3 Model Validation

[0057] 1.3.1 Prognostic Diagnostic Analysis

[0058] For Kaplan-Meier curves, p-values ​​and hazard ratios (HRs) for the 95% confidence interval (CI) are derived using a log-rank test. Univariate and multivariate Cox regression analyses are performed to determine the appropriate conditions for constructing the nomogram infographic. Forest plots are displayed using the 'forestplot' R package, showing the p-values, HRs, and 95% CIs for each variable.

[0059] Based on the results of multivariate Cox proportional hazards analysis, a nomogram risk prediction model was established to predict overall relapse at 1, 3, and 5 years. The nomogram provides a graphical representation of each factor and can be used to calculate the relapse risk for each patient. The relevant scores for each risk factor were calculated using the "rms" R package.

[0060] Predictive accuracy was evaluated using time-dependent receiver operating characteristic (ROC) curves and area under the curve (AUC). The timeROC(v 0.4) package was used to analyze and compare the predictive accuracy of individual genes, subgroups, LMRS models, and risk scores. An AUC > 0.7 was considered to indicate good diagnostic efficiency.

[0061] Decision curve analysis R package-ggDCA was used to construct 1-year, 3-year, and 5-year diagnostic models.

[0062] 1.3.2 Analysis of immune-related functions

[0063] To obtain a reliable immune score assessment result, immunoeeconv is used, which is an R package that integrates six state-of-the-art algorithms, including TIMER, xCell, MCP-counter, CIBERSORT, EPIC, and quanTIseq. These algorithms have all been benchmarked; each has its unique advantages.

[0064] SIGLEC15, TIGIT, CD274, HAVCR2, PDCD1, CTLA4, LAG3, and PDCD1LG2 were selected as immune checkpoint-related transcripts, and their expression values ​​were extracted from the TCGAHNSCC cohort. The TIDE algorithm was used to predict the potential ICB response. The correlation between gene expression and immune scores was plotted using the ggstatsplot package in R, and multi-gene correlations were plotted using the pheatmap package in R.

[0065] The correlation analysis between gene expression and immune scores used Spearman correlation analysis to describe the non-normal distribution correlation between quantitative variables.

[0066] 1.3.2 GEO Validation

[0067] Expression profiles of characteristic genes were extracted from the GEO datasets GSE32877, GSE10300, and GSE41613 for validation of the characteristic model. Risk scores were calculated for each group, and their relationship with systemic treatment response and survival outcomes was further analyzed.

[0068] 1.3.3 Sample collection and immunohistochemistry

[0069] In accordance with the regulations of the Ruijin Hospital Review Committee, the primary tissue samples (HNSCC, n=20) were obtained anonymously. After being fixed in 4% paraformaldehyde, the samples were embedded in paraffin blocks and then cut into 4μm thick sections. After HE staining, another tumor section was incubated overnight at 4°C with antibodies against SREBF1 (Sterol Regulatory Element Binding Transcription Factor 1), PIK3R3 (Phosphoinositide-3-Kinase Regulatory Subunit 3), MAPK9 (Mitogen-Activated Protein Kinase 9), IKK2, APOB (Apolipoprotein B), ACSBG2 (Acyl CoA synthetase bubblegum family member 2), MOGAT2 (Monoacylglycerol-O-acyltransferase 2), and PLA2G10 (Phospholipase A2 Group X) diluted to 1:100 and incubated with the secondary antibody for 2 hours. The slices were viewed in ZeissZen 3.3 and then analyzed using Image-J analysis software.

[0070] 1.4 Data Analysis

[0071] Statistical analysis and ggplot2 (v3.3.2) were performed using R software v4.0.3 (R Foundation for Statistical Computing, Vienna, Austria). P < 0.05 was considered statistically significant.

[0072] 2 Experimental Results

[0073] 2.1 CRISPR / Cas9 library screening identified lipid metabolism-related pathways enriched in chemotherapy resistance genes.

[0074] Using CRISPR / Cas9 library screening technology, differentially expressed genes significantly associated with systemic therapy for HNSCC were identified and functional and pathway enrichment analyses were performed. The high-throughput CRISPR / Cas9 knockout library GeCKO 2.0 (covering an average of 3-6 sgRNAs per gene) based on whole-genome mRNA was transfected into the human pharyngeal carcinoma orthotopic cell line Fadu and the metastatic cell line Detroit-562 for screening for loss of function. Figure 2 A). Repeated screening was performed using cisplatin / 5-fluorouracil, narrowing down potential candidate genes to a total of 6848 genes associated with drug resistance. Figure 2 Among the 731 genes with significant differences (639 positively selected and 92 negatively selected; P < 0.05 and FDR < 0.25), the highest-ranking genes included several genes related to tumor development and drug resistance, including AKT1, KLF6, NCOR2, and several genes significantly overexpressed in head and neck tumors, such as KRT5 and TGFBRAP1. These were among the genes with the most severe expression changes identified through in vitro screening. Functional enrichment and KEGG pathway analysis of these candidate genes revealed significant enrichment in lipid metabolism-related functions. Figure 2 CF), especially genes involved in the PPAR pathway, glycerol and fatty acid metabolism. PPI (Protein-Protein Interaction) network analysis revealed multiple potential central node genes among lipid metabolism-related genes. Figure 2 The lipid-related gene heatmap shows that although lipid metabolism genes are significantly enriched, their functions may vary depending on the cell line and chemotherapeutic drugs. Figure 2 H).

[0075] 2.2 Construction of the Lipid Metabolism Characteristic Prognostic Model LMRS

[0076] Cases (N=173) in the TCGA-HNSCC cohort who had received systemic therapy were divided into two subtypes, C1 and C2, based on differences in the expression of lipid metabolism-related genes (751 genes) (Figure 3A). Survival analysis combined with clinical information showed that group C1 exhibited a poorer clinical prognosis, with a median survival of only 2.3 years, while the survival of group C2 was improved to 6.1 years. Figure 3 B). Group C1 (N=117) and Group C2 (N=56) showed opposite trends in the expression of multiple genes. Figure 3 C). Furthermore, combined with clinical information, it was found that the proportion of female patients was higher in group C1. Figure 3 In pathological classification, the proportion of undifferentiated type is higher, the proportion of moderately differentiated type is lower, and poorly differentiated type is almost non-existent. Figure 3 E). Grouping was not correlated with other clinical indicators. Figure 3 (D; Table 1).

[0077] Combining the results of previous CRISPR / Cas9 library screening with LASSO (Least Absolute Shrinkage and Selection Operator) regression analysis in bioinformatics, we further narrowed down the number of genes involved in this evaluation model. Feature selection was performed using the LASSO regression algorithm, with 10-fold cross-validation, ultimately yielding eight lipid metabolism-related genes (ACSBG2, APOB, IKBKB, MAPK9, MOGAT2, PLA2G10, PIK3R3, SREBF1) as features for this model (Figure 4A, B). The model calculation formula is as follows:

[0078] Risk score = (-2.7486)*ACSBG2 + (1.7158)*APOB + (-0.3216)*IKBKB + (0.4612)*MAPK9 + (-0.8421)*MOGAT2 + (0.6413)*PLA2G10 + (-0.2157)*PIK3R3 + (-0.2355)*SREBF1.

[0079] 2.3 Evaluation of LMRS prognostic models

[0080] For patients in TCGA-HNSCC who had received systemic therapy, risk scores were calculated for each individual based on the status of each gene in the model, and they were divided into a high-score LMRS-High group and a low-score LMRS-Low group according to their scores. Figure 4 (C, D; Table 1). Survival analysis combined with clinical data revealed results similar to the previous subgroup classification of 751 genes: the high-scoring LMRS-High group was associated with a higher mortality rate (median survival = 2.1 years), while the low-scoring LMRS-Low group showed better survival (median survival = 6.1 years). Figure 4 The prognostic assessment model had a diagnostic efficacy AUC greater than 0.7 at 1, 3, and 5 years, and showed better efficacy in the longer-term group. Figure 4 F).

[0081] Table 1. Characteristics of TCGA type head and neck squamous cell carcinoma patients (receiving chemotherapy)

[0082]

[0083]

[0084]

[0085] 2.4 Comparison of the diagnostic value of the model with single molecules and clinical indicators

[0086] In all cases of TCGA-HNSCC, not all individual genes in the model showed significant expression differences; only ACSBG2, MAPK9, PIK3R3, and SREBF1 showed differential expression, and all of them showed elevated expression. Figure 4 The prognostic nomogram constructed using these 8 genes showed good concordance, with C-index = 0.672 (0.651-0.692). Figure 4 The results of univariate and multivariate Cox analyses showed that ACSBG2, IKBKB (Inhibitor of kappa lightpolypeptide gene enhancer in B-cells, kinase epsilon), and PIK3R3 were protective factors in HNSCC, while APOB and MAPK9 were risk factors. Figure 4 (JK). The molecules MOGAT2, PLA2G10, and SREBF1 were not statistically significant. Among individual molecules, only ACSBG2 and SREBF1 showed diagnostic efficacy AUC > 0.7 in TCGA-HNSCC. Figure 4 L).

[0087] Furthermore, among these eight molecules, only MAPK9, MOGAT2, and PIK3R3 showed significant differences in survival prognosis analysis in the TCGA-HNSCC data. Figure 5 (AH). Furthermore, low expression of both MOGAT2 and PIK3R3 indicates a worse prognosis. This not only demonstrates that using a single gene from the model is insufficient for accurate prognostic assessment of TCGA-HNSCC, but also reaffirms that tumors with high expression of lipid metabolism-related genes have relatively better survival prognoses. Combined with previous analysis results, this suggests that patients with high expression of lipid metabolism-related genes may benefit more from systemic therapy. In this research model, multiple genes have negative weights, indicating that relatively high gene expression is more likely to receive lower scores, consistent with the prediction results. The 1- to 5-year DCA decision curve results also show that, compared to other commonly used clinical indicators such as age, sex, clinical stage, pathological type, and smoking status, the 8-gene prognostic diagnostic model with lipid metabolism characteristics has better evaluation performance. Especially within 3 years, the model's diagnostic decision benefits are higher ( Figure 5 Therefore, this 8-gene prognostic evaluation model with lipid metabolism characteristics also has a good evaluation effect, consistent with the evaluation effect of 751 lipid-related genes.

[0088] 2.5 Assessment of LMRS model risk factors and HNSCC-related immune function

[0089] Analysis of immune cell infiltration revealed that this model also had a good predictive effect on immune cell infiltration. Especially in T cells, CIBERSORT results showed that the low-scoring LMRS-Low group was associated with higher infiltration of T-helper, Treg, and M2 macrophages. Figure 6 (A, B). Not every gene in the model can accurately predict the infiltration of immune cells. Figure 6 Therefore, it is necessary to integrate all eight genes for joint evaluation when applying the model. Expression prediction of immune checkpoint-related genes in the model also revealed that PDCD1 (immunosuppressive receptor) and TIGIT (T cell regulation) were expressed more in the LMRS-Low group. PDCD1LG2 (which interacts with PDCD1 to inhibit T cell proliferation) was expressed less in the LMRS-Low group. Figure 6 The TIDE score also showed consistent results, with the LMRS-High group exhibiting a higher score, poorer response to immune checkpoint inhibitor therapy, and shorter post-treatment survival. Figure 6 (E). Specifically, higher risk scores are associated with more CD4+ T cells, fewer CTLs, M2 macrophages, NK cells, and Treg cells. Figure 6 The results (FJ) indicate that the high-scoring LMRS-High group is less likely to benefit from immunotherapy. These results further demonstrate that this model can effectively predict the efficacy of systemic treatments, including immunotherapy, in HNSCC, thereby assessing the prognosis of HNSCC.

[0090] 2.6 Validation of the prognostic value of the LMRS model on the external HNSCC dataset

[0091] Three head and neck tumor datasets from GEO, including: a study evaluating the efficacy of chemotherapy in HNSCC patients (GSE32877, N=2); Figure 7 A), a study that included survival prognostic information for HNSCC patients (GSE10300, N=42); Figure 7 ,B) and a study that included treatment and survival prognostic information (GSE41613, N=97; Figure 7 (C) were selected as the research subjects. After obtaining the gene expression information of each case in the database, the corresponding LMRS risk score was calculated for each case according to the calculation formula of the model and then analyzed. Figure 7The results showed that the group with a better response to chemotherapy (N=13) in the chemotherapy-receiving cases had a lower LMRS risk score, and the surviving cases (N=27) also had a lower LMRS risk score. Combined with the GSE41613 survival prognostic results, this further demonstrates that patients with higher LMRS risk scores have a poorer prognosis. These results further demonstrate that the 8-gene LMRS model can effectively assess the efficacy of systemic therapy for HNSCC and serve as a diagnostic tool for prognostic assessment.

[0092] 2.7 Prognostic validation of the LMRS model in clinical specimens

[0093] In the collected HNSCC specimens (n=6; sensitive group=3, resistant group=3), we also verified the test efficiency of the LMRS model again. Figure 8 As shown, combining the results of MRI and electronic laryngoscopy, we defined patients whose tumor volume shrank by <50% after two cycles of induction chemotherapy as the chemotherapy-sensitive group, and those whose tumor volume shrank by >50% as the drug-resistant group. The IHC or IF staining of eight molecules in the LMRS model of both groups of specimens was consistent with the mRNA expression in the TCGA and GEO databases. Figure 8 B). That is, the LMRS score was lower in the chemotherapy-sensitive group and higher in the relatively resistant group. Figure 8 (CD).

[0094] In summary, this invention, through a combination of CRISPR / Cas9 library screening technology, database mining, and in-depth bioinformatics analysis, identified the functional and genetic factors influencing systemic treatment of HNSCC, thus determining the impact of lipid metabolic remodeling. The model detection panel established in this invention provides appropriate assessment and selection criteria to screen for patients who can obtain the greatest benefit from systemic treatment, while allowing patients who are unlikely to benefit to rapidly transition to other therapies.

[0095] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements and additions without departing from the method of the present invention, and these improvements and additions should also be considered within the scope of protection of the present invention.

Claims

1. The application of reagents for detecting lipid metabolism gene expression levels in the preparation of kits for predicting the prognostic effects of chemotherapy in head and neck squamous cell carcinoma, characterized in that... The lipid metabolism genes are a combination of the following eight genes: ACSBG2, APOB, IKBKB, MAPK9, MOGAT2, PLA2G10, PIK3R3, and SREBF1. The prognostic model is a risk score that includes lipid metabolism features: Risk Score = (-2.7486)*ACSBG2 + (1.7158)*APOB + (-0.3216)*IKBKB + (0.4612)*MAPK9 + (-0.8421)*MOGAT2 + (0.6413)*PLA2G10 + (-0.2157)*PIK3R3 + (-0.2355)*SREBF1. The risk score is calculated by substituting the expression levels of lipid metabolism genes into the prognostic model formula.

Images