A vestibular schwannoma molecular typing gene marker combination, kit and application thereof

Abstract

The application belongs to the technical field of biotechnology, and particularly relates to a vestibular schwannoma molecular typing gene marker combination, a kit and application thereof. The vestibular schwannoma molecular typing gene marker combination provided by the application comprises PILRA, STAC3 and at least two genes selected from TNFSF8, NCKAP1L, AIF1 and SLC7A7, and a matching PCR kit, which is used for determining a vestibular schwannoma molecular subtype and is stable in method technology and strong in operability.

Description

Technical Field

[0001] This invention belongs to the field of biotechnology, specifically relating to a combination of molecular subtyping gene markers for vestibular schwannomas, a reagent kit, and their applications. Background Technology

[0002] Schwannoma is a benign tumor originating from Schwann cells, accounting for approximately 8% of intracranial tumors. Vestibular schwannoma (VS) is the most common intracranial type of schwannoma. Patients typically present with hearing loss, facial paralysis, and other neurological dysfunctions. Tumor progression can compress surrounding cranial nerves and the brainstem, and in severe cases, can even be life-threatening. Due to significant differences in tumor growth rate, surgical efficacy, and long-term prognosis among VS patients, individualized treatment strategies are increasingly emphasized in clinical intervention for this disease.

[0003] Currently, the clinical diagnosis and assessment of schwannomas primarily rely on histopathological features, such as observing Antoni A / B type structures, Verocay bodies, and detecting the Ki-67 proliferation index. However, this purely morphological diagnostic method has significant limitations: firstly, it cannot distinguish the heterogeneity within the tumor, making it difficult to reflect the molecular nature of the disease; secondly, it cannot predict the patient's prognosis or disease progression; thirdly, it cannot predict the preservation of postoperative neurological function (such as facial and auditory nerve function); and fourthly, it cannot predict the risk of tumor recurrence or progression. In other words, current clinical diagnosis based solely on histopathological features lacks a precise molecular-based classification system, making it difficult to achieve accurate patient stratification and effective intervention.

[0004] In recent years, with the development of molecular biology techniques, researchers have gradually recognized the significant heterogeneity of schwannomas at the molecular level. Previous studies have systematically characterized the genome of schwannomas (see: Agnihotri S, et al. The genomic landscape of schwannoma. Nature Genetics, 2016, 48(11):1339-1348), and have reviewed and discussed the current molecular understanding of central nervous system schwannomas (see: Tsuchiya et al. Current molecular understanding of central nervous systemschwannomas. Acta Neuropathologica Communications, 2025, 13:24). Building upon this foundation, the applicant previously conducted transcriptomic sequencing analysis on a large cohort of vestibular schwannomas associated with neurofibromatosis type 2 (NF2-SWN), and for the first time globally proposed a molecular subtype diagnostic method for schwannoma based on transcriptomic information (see reference: Zhao et al. Transcriptomic profiling identifies two molecular subtypes of NF2-related schwannoma with distinct immune microenvironment characteristics. Acta Neuropathologica, 2025, 149(1):47). This study established two molecular subtypes of vestibular schwannoma: immune-enriched schwannoma (IES) and immune-depleted schwannoma (IDS), and confirmed that these two subtypes differ significantly in tumor biological characteristics and patient prognosis. This preliminary technical approach primarily employs bulk RNA sequencing, which involves extracting RNA from fresh frozen tumor samples, performing analysis using Illumina's RNA sequencing platform, and then using unsupervised clustering analysis for molecular subtyping diagnosis. This approach mainly targets NF2-related schwannomas, and its applicability and generalizability to sporadic vestibular schwannomas require further exploration.

[0005] Subsequently, the applicant further analyzed the transcriptome sequencing data of 86 sporadic vestibular schwannoma (VS) and NF2-SWN vestibular schwannoma samples, confirming that all types of vestibular schwannoma can be classified into two molecular subtypes: IES and IDS (see: Zhao et al. Divergent Schwann cell-immune celltrajectories across vestibular schwannoma subtypes mirror nerve injury-repairprograms. Journal of Neuroinflammation, 2026, online). Although the above-mentioned molecular subtyping method based on whole genome transcriptome data provides an important basis for the precise diagnosis and treatment of schwannomas, it mainly relies on bulk RNA sequencing technology. Analysis based on whole transcriptome sequencing data has high requirements for samples (requiring fresh frozen tumor tissue), and the testing cost is expensive and the data analysis process is complex, making it difficult to widely carry out and apply in routine clinical testing.

[0006] In summary, there is an urgent need in this field to develop a universal molecular subtyping scheme for vestibular schwannomas that is suitable for both convenient testing and accurate subtyping, and that can be implemented on a variety of conventional testing platforms (not just high-throughput sequencing platforms). This would overcome the limitations of traditional histopathological diagnosis and break through the bottlenecks in the clinical translation of existing transcriptome sequencing subtyping technologies, thereby providing effective technical support for individualized treatment, prognostic assessment, and neurological function protection of schwannomas. Summary of the Invention

[0007] To address the problems existing in the prior art and achieve molecular subtyping of vestibular schwannomas, this invention provides:

[0008] 1. A set of simplified gene markers for molecular subtyping of vestibular schwannoma, the set of which covers multiple gene sets with STAC3 and PILRA as the core, and by detecting the expression level of multiple genes, the same subtyping performance as large-scale gene sets can be achieved;

[0009] 2. A low-throughput, low-cost molecular typing detection kit is proposed, which is suitable for detection by routine clinical testing platform experimental methods (qPCR, ddPCR, etc.), and can shorten the detection cycle from several weeks to several hours, and reduce the detection cost by more than 90%.

[0010] 3. This invention aims to detect the transcript expression signals of multiple genes, such as PILRA, TNFSF8, AIF1, and STAC3, centered around STAC3 and PILRA, in vestibular schwannoma tissue. This is achieved through specific primer and probe amplification, targeted gene capture, high-throughput sequencing, or expression profiling microarrays. The invention designs qPCR primer and probe sets and targeted capture probe sets for molecular subtyping of vestibular schwannoma, establishing a molecular subtyping method, kit, and application for IES / IDS of vestibular schwannoma based on a frozen parameter logistic regression model. The kit and system can be used for molecular subtyping of vestibular schwannoma, prognostic assessment, treatment guidance, and subject stratification in clinical trials.

[0011] The technical solution of the present invention includes:

[0012] This invention discloses a combination of gene markers for molecular subtyping of vestibular schwannoma, including PILRA, STAC3, and a combination of at least two genes selected from TNFSF8, NCKAP1L, AIF1, and SLC7A7.

[0013] In some implementations, STAC3 and PILRA, and a combination of two genes selected from TNFSF8, NCKAP1L, AIF1, and SLC7A7 are included.

[0014] In some embodiments, the combination of genetic markers is PILRA, TNFSF8, AIF1, and STAC3.

[0015] In some embodiments, it is composed of PILRA, AIF1, STAC3 and any one of the genes selected from TNFSF8, NCKAP1L or SLC7A7.

[0016] In some embodiments, the combination of genetic markers is PILRA, TNFSF8, NCKAP1L, AIF1, and STAC3.

[0017] In some implementations, PILRA, TNFSF8, NCKAP1L, AIF1, STAC3, and SLC7A7 are included.

[0018] The second aspect of this invention discloses a detection kit for molecular subtyping of vestibular schwannomas, comprising: a specific upstream primer, a downstream primer, and a fluorescent probe for transcripts of the gene marker combination described in the first aspect of this invention, wherein the sequences of the upstream primer, the downstream primer, and the fluorescent probe include SEQ ID NO.1~3, SEQ ID NO.13~15, SEQ ID NO.19~21, and any 2~4 sequences selected from sequences SEQ ID NO.4~6, SEQ ID NO.7~9, SEQ ID NO.10~12, and SEQ ID NO.16~18.

[0019] In some embodiments, the sequences of the upstream primer, downstream primer, and fluorescent probe include SEQ ID NO.1~3, SEQ ID NO.4~6, SEQ ID NO.10~12, and SEQ ID NO.13~15.

[0020] In some implementations, it also includes: internal reference gene detection reagent, reverse transcription reagent, probe method premix solution, positive control and negative control;

[0021] The internal reference gene detection reagent includes a primer and probe set targeting the internal reference gene GAPDH.

[0022] The reverse transcription reagent includes reverse transcriptase, random primers, dNTPs, RNase inhibitors, and reverse transcription buffer.

[0023] The probe-based premix includes Taq DNA polymerase, dNTPs, MgCl2, PCR buffer, and the aforementioned upstream primer, downstream primer, and fluorescent probe.

[0024] The negative control is a template-free control, which uses nuclease-free water instead of cDNA from the test sample as a template.

[0025] In some embodiments, the following components are also included:

[0026] RNA extraction reagents and DNase I;

[0027] Preferably, it also includes: quality control standards and typing interpretation tools;

[0028] The quality control standards include standards for each transcript in the combination of the gene markers at known concentrations;

[0029] The typing interpretation tool is configured to determine and output the molecular typing results of vestibular schwannomas based on the detection results of each transcript, and the molecular typing includes immune enrichment type and immune depletion type.

[0030] The third aspect of this invention also discloses the application of the gene marker combination for molecular subtyping of schwannomas described in the first aspect of this invention or the detection kit for molecular subtyping of vestibular schwannomas described in the second aspect of this invention in basic molecular biology research on schwannomas; or, in in vitro cell experiments for screening targeted drugs for vestibular schwannomas for non-diagnostic purposes.

[0031] Based on the foregoing technical solution, the beneficial effects of the present invention include at least the following:

[0032] (1) The gene marker combination of the present invention for the molecular subtyping of vestibular schwannomas, with STAC3 and PILRA as the core of multiple gene sets and their matching subtyping models, can simplify the molecular subtyping steps of vestibular schwannomas and can realize the molecular subtyping of vestibular schwannomas tissue samples on a conventional qPCR platform, which significantly reduces the technical threshold and detection cost of subtyping.

[0033] (2) This invention screens out 4-6 biomarker combinations, which greatly reduces the number of gene biomarkers. Taking 4 gene biomarkers as an example, the combination of this invention has an apparent AUC of 0.934 (95% CI: 0.870–0.983) in 86 training set samples, and an AUC of 0.927 (95% CI: 0.863–0.978) with leave-one-out cross-validation. It has 100% specificity, 82.6% sensitivity, and 90.7% accuracy. It takes into account both good classification performance and low detection cost.

[0034] (3) This invention has systematically verified the optimal number of genes: Existing technologies have not determined the relationship between genotyping performance and the number of genes through marginal benefit analysis. This invention, through a systematic evaluation of the number of candidate genes from 1 to all, found that the LOOCV AUC enters a plateau phase at 4-6 genes, the performance drops significantly at 3 genes or less, and overfitting occurs when there are more than 6 genes.

[0035] (4) The vestibular schwannoma molecular subtype classification model constructed in this invention has excellent generalization ability and cross-platform stability: it was validated in two completely independent external validation cohorts: the whole genome transcriptome sequencing dataset HRA002762 (n=71) and the GEO gene expression microarray datasets GSE141801, GSE108524 and GSE54934 (n=140), totaling 211 independent samples. The validation results show that the model of this invention achieves a classification efficiency of AUC>0.93 in both external cohorts. Attached Figure Description

[0036] Figure 1The ROC curves (n=86, epigenetic AUC and LOOCV cross-validation) of the 4-gene combination (PILRA, TNFSF8, AIF1 and STAC3) genotyping model of this invention are marked with the optimal decision threshold and performance indicators.

[0037] Figure 2 The AUC of the four single genes and four gene combinations (PILRA, TNFSF8, AIF1 and STAC3) of this invention are compared in the training cohort (n=86), NF2-SWN validation cohort (n=71), and GEO validation cohort (n=140).

[0038] Figure 3 The diagram shows the positional distribution of the four gene combinations (PILRA, TNFSF8, AIF1, and STAC3) in the candidate combination space (left: all 1,365 exhaustive four gene combinations in the candidate pool; right: 3,000 four gene combinations randomly selected from the candidate gene set).

[0039] Figure 4 This represents the marginal benefit curve between gene quantity and genotyping performance.

[0040] Figure 5 This document summarizes the ROC curves and frozen model AUC values ​​of the genotyping model for the four gene combinations (PILRA, TNFSF8, AIF1, and STAC3) of this invention in the training cohort and two independent validation cohorts (NF2-SWN independent validation cohort, n=71; GEO microarray cohort, n=140).

[0041] Figure 6 Comparison of ROC curves for the 4-gene combination (PILRA, TNFSF8, AIF1 and STAC3), 5-gene combination (PILRA, TNFSF8, NCKAPIL, AIFI and STAC3), and 6-gene combination (PILRA, TNFSF8, NCKAP1L, AIF1, STAC3 and SLC7A7) of this invention.

[0042] Figure 7 The heatmap of differential gene typing of vestibular schwannoma in this invention divides 86 samples into two molecular subtypes: immune enrichment type (IES, red, 46 cases) and immune depletion type (IDS, blue, 40 cases).

[0043] Figure 8 This is a flowchart illustrating the construction and verification technology of the molecular subtyping model for vestibular schwannoma of the present invention.

[0044] Figure 9 This is a flowchart of the application of the low-throughput qPCR kit for molecular typing of vestibular schwannomas of the present invention (including an example of the use of the interpretation card). Detailed Implementation

[0045] Embodiments of the present invention are described in detail below. Examples of each embodiment are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0046] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0047] In this invention, the term log2(x+1) transformation refers to a logarithmic transformation based on base 2. Specifically, the gene expression value x (x≥0) is first added by 1 to avoid log2(0) being undefined, and then the logarithm based on base 2 is taken, i.e., y = log2(x+1). This transformation is used to compress the numerical dynamic range of highly expressed genes, reduce the impact of extreme values ​​on the model, and at the same time preserve the relative differences between low-expressed genes.

[0048] In this invention, the term Z-score standardization refers to standard score standardization, specifically: for a gene expression value x, the mean μ and standard deviation σ of the expression values ​​of all samples of that gene in the training queue are transformed, and the calculation formula is z = (x - μ) / σ; after transformation, the expression value of the gene in the training queue follows a distribution with a mean of 0 and a standard deviation of 1, eliminating the differences in dimensions and numerical ranges between different genes, so that each feature gene has comparable weights in the logistic regression model.

[0049] In this invention, the term "IES" refers to immune-enriched vestibular schwannoma.

[0050] In this invention, the term "IDS" refers to immune-depleted vestibular schwannoma.

[0051] In this invention, the term "cDNA" refers to Complementary DNA, which is complementary DNA synthesized using RNA as a template via reverse transcriptase catalysis and does not contain introns.

[0052] In this invention, the term "No Template Control (NTC)" refers to a negative control reaction tube in which no cDNA template or DNA template is added to the real-time quantitative PCR (qPCR) reaction system, but an equal volume of nuclease-free water or TE buffer is used instead of the template.

[0053] In this invention, the term "fluorescent probe" refers to a short synthetic oligonucleotide chain labeled with a fluorescent reporter group and a fluorescent quencher group at both ends, used in real-time quantitative PCR (qPCR) for the specific detection of target nucleic acid sequences; the fluorescent reporter group emits a detectable fluorescent signal when separated from the quencher group.

[0054] In this invention, the term "AUC" refers to the area under the ROC curve, which is a core evaluation index for quantitatively assessing the discriminative ability of molecular subtyping models. The value range is 0.5 to 1.0. The closer the value is to 1, the higher the accuracy and specificity of the neurofibroma subtype differentiation and sample discrimination model of this invention.

[0055] In this invention, the term "Youden Index" (also known as the accuracy index) is a comprehensive evaluation index for diagnostic tests. It combines the characteristics of both sensitivity and specificity, and is calculated as: True Positive Rate + True Negative Rate - 1. Its accuracy value ranges from 0 to 1, with higher values ​​being better. The Youden Index is often used to confirm the optimal threshold for diagnostic indicators using ROC curves.

[0056] In this invention, 1-specificity refers to the false positive rate, that is, the probability of a negative sample being detected as positive.

[0057] In this invention, the term "DNase I" refers to DNase I (Deoxyribonuclease I): a general term for a class of endonucleases derived from mammals (such as bovine pancreas) that are secreted or recombinantly expressed, with the International Enzyme Committee number EC3.1.21.1; it is used in the presence of divalent metal ions (such as Mg). 2+ Ca 2+In the presence of [a specific enzyme / enzyme], non-specific catalytic hydrolysis of phosphodiester bonds within single-stranded or double-stranded DNA molecules is performed to generate small oligonucleotide fragments with 5′-phosphate ends. The optimal reaction pH is typically 7.0–8.0. The DNase I described in this invention includes commercially available formulations (such as bovine pancreatic-derived DNase I, recombinant DNase I) and mutants or homologous enzymes with equivalent catalytic activity. These are primarily used to digest free or contaminated genomic DNA in samples without significantly degrading RNA.

[0058] In this invention, the term "Ct value" refers to the Cycle threshold, which is the number of PCR cycles required for the fluorescence signal to reach a set threshold. It is used to characterize the copy number of the target nucleic acid in the starting template and is a basic parameter for relative or absolute quantitative analysis.

[0059] In this invention, the term "replicates" refers to at least three parallel tests performed on the same sample in the same reaction system (or after equalization), with a standard deviation between replicates < 0.5. Specifically, in qPCR experiments, the standard deviation of the Ct values ​​obtained from multiple replicates of the same sample amplified in parallel must be less than 0.5 cycles.

[0060] In this invention, the term RIN (RNA Integrity Number) is an indicator reflecting the degree of RNA degradation, with a range of 1-10, where 10 represents completely undegraded and 1 represents completely degraded. RIN ≥ 7 means that the RNA has maintained a high degree of integrity, which can meet the precise gene detection requirements for molecular subtyping of schwannomas.

[0061] In this invention, the term "A260 / A280" refers to the absorbance ratio used in ultraviolet spectrophotometry to detect nucleic acid purity. A260 is the absorption peak value for nucleic acids (RNA / DNA), and A280 is the absorption peak value for proteins. The theoretical ratio for pure RNA is approximately 2.0. A ratio between 1.8 and 2.1 indicates high purity of the extracted total RNA. A ratio below 1.8 usually indicates protein or phenolic contamination; a ratio above 2.1 may indicate RNA degradation or residual chemical reagents such as guanidine isothiocyanate.

[0062] In this invention, the term "95% CI" refers to a 95% confidence interval.

[0063] The main reagents used in the embodiments of this invention and their specifications are as follows:

[0064] (1) qPCR detection reagents. The TaqMan probe used was the Thermo Fisher Custom TaqMan® MGB probe (5' end labeled with FAM fluorescent group, 3' end labeled with MGB-NFQ quencher group; probe sequences of biomarker gene and internal reference gene GAPDH are shown in Table 2), with a final reaction concentration of 0.2 μmol / L. The qPCR premix used was TaqPath™ ProAmp™ Multiplex Master Mix (Thermo Fisher, Cat. A15299, 2×, containing hot-start Taq DNA polymerase, dNTPs, MgCl2 and ROX reference dye), added at a reaction system of 12.5 μL / 25 μL; if additional adjustment of magnesium ion concentration was required, 25 mM MgCl2 stock solution (Thermo Fisher, Cat. AM9530G) was used. When using the SYBR Green method, iTaq™ Universal SYBR® Green Supermix (Bio-Rad, Cat. 1725124, 2×) was used.

[0065] (2) Reverse transcription reagents. The reverse transcriptase was SuperScript™ IV reverse transcriptase (Thermo Fisher, Cat. 18090010, 200 U / μL), with 1 μL added to every 20 μL of reverse transcription system; the random primers were random hexamers (Thermo Fisher, Cat. N8080127, 50 ng / μL), with a final concentration of 2.5 μmol / L; the dNTPs were dNTP Mix (Thermo Fisher, Cat. R0192, 10 mM each), with a final concentration of 0.2 mmol / L; the RNase inhibitor was RNaseOUT™ recombinant ribonuclease inhibitor (Thermo Fisher, Cat. 10777019, 40 U / μL), with a final concentration of 1 U / μL; the reverse transcription buffer was 5× First-Strand Buffer (a component of SuperScript™ IV).

[0066] (3) Sample processing reagents. RNA extraction was performed using TRIzol™ Reagent (Invitrogen, Cat. 15596026), with 1 mL of TRIzol added for every 50–100 mg of schwannoma tissue; or the RNeasy Mini Kit column extraction kit (Qiagen, Cat. 74104) was used, following the product instructions. Genomic DNA removal was performed using TURBO™ DNase (Thermo Fisher, Cat. AM2238, 2 U / μL, RNase-free), at 37°C for 30 minutes.

[0067] (4) Quality control standards. The transcript standards for the four biomarker genes (PILRA, TNFSF8, AIF1, STAC3) were all in vitro transcribed RNAs, purified by DNase treatment, and then serially diluted with buffer containing vector RNA at concentrations of 10. 8 10 7 10 6 10 5 10 4 10 3 10 2 Copy / μL, used to establish standard curves for each gene; each gene is aliquoted into 10 μL / tube and stored at -80℃ in the dark, avoiding repeated freeze-thaw cycles.

[0068] The technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems are described in detail below with specific embodiments. The following embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0069] 1. Combination of genetic markers

[0070] 1.1 Key Discovery: 4-6 Gene Classifier

[0071] The gene marker combination for vestibular schwannoma of this invention, with STAC3 and PILRA as the core, also includes at least two gene sets selected from TNFSF8, NCKAP1L, AIF1, and SLC7A7. By detecting the expression levels of the aforementioned 4 to 6 genes, the same typing performance as large-scale gene sets can be achieved. Among them, the combination of 4 genes, PILRA, TNFSF8, AIF1, and STAC3, has the best effect. This combination is the optimal combination with the highest leave-one-out cross-validation (LOOCV) AUC (0.9266) among all 15 candidate 4-gene combinations, achieving typing performance comparable to larger-scale combinations with the fewest number of genes, and achieving 100% specificity. The names and functions of each gene in this invention are summarized in Table 1.

[0072] The genes listed in Table 1 are all relatively highly expressed in immune-enriched schwannomas (IES) samples, and their combined expression profiles can stably distinguish between the two molecular subtypes, IES and IDS.

[0073] Table 1: Composition of the gene marker combination described in this invention

[0074]

[0075] As shown in Table 1, all six genes met the following criteria: the average expression level of the IES group was higher than that of the IDS group, and the log2FC ranged from 0.72 to 1.02 (expression level 1.6 to 2.0 times higher). All differences were highly statistically significant (P < 5 × 10⁻⁶). -10 ).

[0076] 2. Low-throughput molecular typing detection kit

[0077] 2.1 Reagent Kit Design Principles

[0078] (1) Low throughput: The kit only detects a few (e.g., 4 to 6) gene marker combinations used for molecular subtyping of schwannomas (e.g., when the gene marker combination used for molecular subtyping of schwannomas is 4 target genes, only 4 target genes are detected), with a short detection cycle and low cost;

[0079] (2) Multi-platform compatibility: Can be implemented on conventional platforms such as qPCR and ddPCR;

[0080] (3) Clinical application: Applicable to fresh frozen tissues, meeting the needs of routine clinical testing;

[0081] (4) Easy to operate: Equipped with standardized interpretation tools (interpretation cards or simple software), no complex bioinformatics analysis is required;

[0082] (5) Complete quality control: including positive control, negative control, internal reference gene and batch-to-batch quality control standards.

[0083] 2.2 Kit Components (qPCR Implementation Plan)

[0084] 2.2.1 Core Components:

[0085] (1) Specific primer and probe set: The kit of the present invention contains a combination of gene markers for specific molecular subtypes of schwannomas. For example, when the combination of gene markers is 4 to 6 genes (PILRA, TNFSF8, NCKAP1L, AIF1, STAC3, SLC7A7), the specific upstream primer, downstream primer and fluorescent probe (such as TaqMan probe) of the transcript are provided. The primer and probe sequences of each gene are shown in Table 2.

[0086] Table 2: qPCR primer and probe sequences for 4–6 biomarker genes and the internal reference gene (GAPDH)

[0087]

[0088] (2) Internal reference gene detection reagent: primer and probe set for GAPDH (see Table 2), used for expression level normalization;

[0089] (3) Reverse transcription reagents: The reverse transcriptase was SuperScript™ IV reverse transcriptase (Thermo Fisher, Cat. 18090010, 200 U / μL), with 1 μL added to every 20 μL of reverse transcription system; the random primers were random hexamers (Thermo Fisher, Cat. N8080127, 50 ng / μL), with a final concentration of 2.5 μmol / L; the dNTPs were dNTP Mix (Thermo Fisher, Cat. R0192, 10 mM each), with a final concentration of 0.2 mmol / L; the RNase inhibitor was RNaseOUT™ recombinant ribonuclease inhibitor (Thermo Fisher, Cat. 10777019, 40 U / μL), with a final concentration of 1 U / μL; the reverse transcription buffer was 5× First-Strand Buffer (a component of SuperScript™ IV);

[0090] (4) qPCR premix: The TaqMan probe used was the Thermo Fisher Custom TaqMan® MGB probe (5' end labeled with FAM fluorescent group, 3' end labeled with MGB-NFQ quencher group; probe sequences of 6 biomarker genes and internal reference gene GAPDH are shown in SEQ ID NO: 3, 6, 9, 12, 15, 18 and 21, respectively), with a final reaction concentration of 0.2 μmol / L. The qPCR premix used was TaqPath™ ProAmp™ Multiplex Master Mix (Thermo Fisher, Cat. A15299, 2×, containing hot-start Taq DNA polymerase, dNTPs, MgCl2 and ROX reference dye), added at a reaction system of 12.5 μL / 25 μL; if additional adjustment of magnesium ion concentration was required, 25 mM MgCl2 stock solution (Thermo Fisher, Cat. AM9530G) was used. When using the SYBR Green method, use iTaq™ Universal SYBR® Green Supermix (Bio-Rad, Cat. 1725124, 2×).

[0091] (5) Negative control: template-free control (NTC, nuclease-free water).

[0092] 2.2.2 Auxiliary components:

[0093] (6) Sample processing reagents: RNA extraction reagents (such as TRIzol™ Reagent (Invitrogen, Cat. 15596026) or RNeasy Mini Kit column extraction kit (Qiagen, Cat. 74104), DNase I;

[0094] (7) Quality control standards: Taking a combination of four gene markers as an example, the transcript standards of the four marker genes (PILRA, TNFSF8, AIF1, STAC3) are all in vitro transcribed RNAs, which are purified by DNase I and then serially diluted with buffer containing vector RNA in a concentration series of 10. 8 10 7 10 6 10 5 10 4 10 3 10 2 Copy / μL, used to establish standard curves for each gene; each gene is aliquoted into 10 μL / tube and stored at -80℃ in the dark, avoiding repeated freeze-thaw cycles;

[0095] (8) Typing interpretation tool: A reading card or simple analysis software with all frozen parameters of the typing model of this invention (regression coefficients of each gene, standardized mean and standard deviation) and the judgment threshold (P=0.7088). The reading card is in paper or spreadsheet form. Clinical operators only need to substitute the ΔCt value or relative expression level obtained by qPCR detection to obtain the IES / IDS typing results, without the need for professional bioinformatics background.

[0096] 2.3 Quality Control Standards

[0097] (1) Internal control gene quality control: GAPDH Ct value should be between 15 and 25, and the standard deviation between replicates should be <0.5;

[0098] (2) Negative control quality control: NTC showed no amplification signal (Ct>35 or no signal);

[0099] (3) Inter-batch quality control: Each batch of testing must run both a positive control and a negative control simultaneously. Only when the control results meet expectations can the batch of testing be deemed valid.

[0100] 3. Classification Method Flow

[0101] 3.1 Method and Steps

[0102] like Figure 8 As shown, taking four gene markers as an example, the method for molecular subtyping of schwannomas using this invention includes the following steps:

[0103] S1. Sample Acquisition: Obtain fresh frozen schwannoma tissue samples from the subjects;

[0104] S2. RNA Extraction and Quality Control: Extract total RNA from tissues and test its integrity. Fresh tissues should have a RIN ≥ 7. Test its purity. The A260 / A280 ratio should be between 1.8 and 2.1. Qualified samples will proceed to subsequent tests.

[0105] S3. Reverse transcription: Total RNA is reverse transcribed into cDNA;

[0106] S4. Expression level detection: The expression levels of transcripts of four gene markers and internal reference genes were detected by qPCR;

[0107] S5. Data preprocessing: The relative expression levels of the four genes were calculated using the ΔΔCt method or the standard curve method, and then transformed by log2(x+1). The data were then standardized by Z-score according to the mean μ and standard deviation σ of each gene listed in Table 2.

[0108] S6. Subtyping score: Substitute the standardized expression value into the logistic regression subtyping model shown in the formula to calculate the probability P that the sample belongs to the IES type;

[0109] S7. Subtype determination: When P≥0.7088, it is determined to be immune enrichment type (IES), and when P<0.7088, it is determined to be immune exhaustion type (IDS). This determination threshold is determined in the training queue according to the principle of maximizing Youden index.

[0110] 3.2. Fractal Model

[0111] The typing model of this invention is a logistic regression model based on the standardized expression values ​​of four genes, and the calculation formula is as follows:

[0112] z_i = (log2(E_i +1)-μ_i) / σ_i

[0113] S = β0 + Σ β_i · z_i

[0114] P(IES) = 1 / (1 + e^(-S))

[0115] In the formula: E_i is the original expression level of the i-th gene; μ_i and σ_i are the mean and standard deviation of the log2 transformed expression value of the gene in the training queue (n=86), respectively; β0 is the intercept, and β_i is the regression coefficient of the i-th gene.

[0116] The model parameters are shown in Table 3.

[0117] Table 3: Parameters of the 4-gene combination (PILRA+TNFSF8+AIF1+STAC3) model

[0118]

[0119] Note: The standardized mean μ and standard deviation σ are the same as those for the 6-gene model, calculated based on the log2(FPKM+1) expression level in the training cohort. The epigenetic AUC of this 4-gene model in the training cohort is 0.9342.

[0120] 3.3. Parting performance:

[0121] As shown in Table 4 and Figure 1 As shown, in the training cohort of 86 cases, the epigenetic AUC of the described 4-gene typing model (PILRA, TNFSF8, AIF1, STAC3) was 0.934 (95% CI: 0.870-0.983), with a sensitivity of 82.6% (38 / 46), a specificity of 100% (40 / 40), an accuracy of 90.7% (78 / 86), a positive predictive value of 100%, and a negative predictive value of 83.3%. Using leave-one-out cross-validation (LOOCV), the AUC was 0.927 (95% CI: 0.863-0.978), with a sensitivity of 78.3% (36 / 46), a specificity of 100% (40 / 40), and an accuracy of 88.4% (76 / 86). The confusion matrix showed that all 40 IDS-type samples were correctly identified, with no false positives. All four genes showed relatively high expression in the IES-type samples.

[0122] The diagnostic efficacy of single-gene biomarkers has significant limitations: the AUC of the four biomarker genes used individually in the training cohort ranged from 0.885 to 0.927. While they possess some diagnostic value, they suffer from three main drawbacks: First, sensitivity and specificity are difficult to balance. At their respective optimal Youden thresholds, PILRA, with the highest sensitivity, only reached 0.826 (specificity 0.950); while STAC3, although achieving a specificity of 1.000, had a sensitivity of only 0.783, meaning that more than one-fifth of IES subtype patients would be missed. Second, cross-platform stability is poor. For example... Figure 2As shown, the AUC of single genes fluctuated significantly across different cohorts: the cross-cohort fluctuations of PILRA, TNFSF8, AIF1, and STAC3 in the training cohort, NF2-SWN validation cohort, and GEO validation cohort were 0.046, 0.052, 0.090, and 0.085, respectively; while the cross-cohort AUC fluctuation of the four-gene combined model of this invention was only 0.015 (0.931–0.946), showing significantly better stability than any single gene. Third, a single indicator cannot characterize the multidimensional molecular features of a subtype. Schwannoma IES and IDS subtypes involve synergistic changes in multiple biological processes such as immune infiltration and extracellular matrix remodeling. The expression level of a single gene cannot fully reflect the molecular essence of the subtype and is easily affected by individual differences and sample heterogeneity. Model performance indicators are shown in Table 4 and... Figure 1 , Figure 2 .

[0123] Table 4: Performance metrics of the 4-genotyping model in the training cohort (n=86)

[0124]

[0125] 3.4. System verification of combinatorial optimality:

[0126] To verify the optimality of the biomarker combination screening process, this invention systematically evaluated all candidate biomarker combinations. The candidate biomarkers were derived from the candidate gene set obtained in Example 2, which was progressively reduced from 27,095 quantitative genes using subtype tags as a reference and differential expression analysis combined with random forest / Boruta feature screening.

[0127] (1) Comparison of combinations of the same scale: The top 15 genes with the highest single-gene genotyping AUC values ​​were used as the candidate pool (in descending order of AUC value: STAC3, PILRA, PARVG, LGALS9, RIPK3, ARHGAP9, GALNT6, CD37, BIN2, PAK1, SIGLEC7, CAPG, ARL11, FBP1, NCKAP1L). Four genes were randomly selected from these to form all C(15,4) = 1,365 possible combinations, and the genotyping AUC was evaluated one by one. The epigenetic AUC of the four-gene combination of this invention (i.e., the AUC of the training set after resubmission and without cross-validation, 0.9342) ranked in the top 2.3% of all combinations, which is better than 97.8% of other combinations (see Figure 3 (Left); Furthermore, 3,000 four-genome combinations were randomly selected from the above candidate gene set as controls. The AUC of the combination of the present invention was superior to 99.2% of the random combinations (see...). Figure 3 right).

[0128] (2) Equivalent to the theoretically optimal combination: The 4-gene combination with the highest apparent AUC in the candidate pool is STAC3+PILRA+ARHGAP9+BIN2, with an apparent AUC of 0.938, which seems to be better than the combination of the present invention (0.934). However, after leave-one-out cross-validation (LOOCV) correction, the AUC of this combination drops back to 0.925, which is lower than the 0.927 of the combination of the present invention. This indicates that the apparent advantage of this combination stems from overfitting to the training data, and its true generalization performance is not higher than that of the combination of the present invention—the combination of the present invention represents the upper limit of performance for the 4-gene scale of the candidate space.

[0129] (3) Validation of the marginal benefit of the number of genes: The performance of combinations with 1, 2, 3, 4, 5, 6, 7, 8, 10, 15, 20, 30 genes up to all candidate genes was evaluated. LOOCV validation showed that the AUC plateaued at 4-6 genes (0.919-0.926); after more than 6 genes, it gradually decreased (7 genes: 0.920, 8 genes: 0.918, 20 genes: 0.908, all candidate genes: 0.902), while the model AUC continued to increase with the number of genes (reaching 1.000 for all candidate genes), showing typical overfitting. This indicates that increasing the number of genes does not bring additional discrimination benefits, and 4-6 genes is the optimal combination size window that balances genotyping performance and detection cost. See Table 5 and Figure 4 Based on the combined plateau phase and overfitting analysis, this invention ultimately selected the four-gene combination PILRA+TNFSF8+AIF1+STAC3, which has the fewest genes and the highest LOOCV AUC during the plateau phase, as the biomarker combination.

[0130] Table 5: Discriminant performance of different combinations of gene numbers (training queue, n=86)

[0131]

[0132] Note 1. Apparent AUC is the AUC of the training set, which has not been cross-validated and may have overfitting bias.

[0133] 2. LOOCV AUC is the leave-one-out cross-validation AUC, which is an unbiased estimate and reflects the true generalization performance;

[0134] 3. The "overfitting" described in this invention is determined according to the following objective criteria:

[0135] (1) The difference between the epigenetic AUC (training set resubmission AUC) and the LOOCV AUC (leave-one-out cross-validation AUC) increased significantly. The difference was 0.012 to 0.013 when there were 4 to 6 gene combinations, and increased to more than 0.045 when there were ≥15 genes.

[0136] (2) The LOOCV AUC fell below the lower limit of the plateau period of 4-6 genes (0.919).

[0137] (3) Clinical safety requirement of specificity below 95%. If two or more of the above three conditions are met, it is considered significant overfitting;

[0138] 4. Specificity refers to the proportion of IDS-type samples that are correctly identified, with a clinical requirement of ≥95% (false positive rate ≤5%).

[0139] 5. The specificity of the 1-3 gene model is <95%. The false positive rate is >5%, which does not meet the clinical safety requirements.

[0140] 6. For LOOCV with 7 or more genes, the AUC begins to decline, indicating decreased specificity, which is a typical sign of overfitting;

[0141] 7. The epigenetic AUC of all candidate genes was 1.0 (perfect fit), but the AUC of LOOCV was only 0.902 and the specificity was 82.5%, which is a serious overfit and has no clinical value.

[0142] 4. Multi-platform external verification:

[0143] The four-genotyping model was validated in multiple independent cohorts: the frozen model AUC in the NF2-SWN independent transcriptome sequencing validation cohort (HRA002762, n=71) was 0.946, and the in-cohort normalized AUC was 0.944; the frozen model AUC in the GEO expression profiling microarray cohorts (GSE141801, GSE108524, and GSE54934, n=140) was 0.931, and the in-cohort normalized AUC was 0.932, indicating that the model has good cross-platform and cross-cohort generalization ability (see details). Figure 5 (Table 6-2).

[0144] The present invention will be explained below with reference to specific embodiments. Those skilled in the art will understand that the following embodiments are for illustrative purposes only and should not be considered as limiting the scope of the invention. Where specific techniques or conditions are not specified in the embodiments, they are performed according to the techniques or conditions described in the literature in the field or according to the product instructions. Reagents or instruments whose manufacturers are not specified are all conventional products that can be obtained commercially.

[0145] Example 1: Establishment of a transcriptome sequencing and molecular typing system for a schwannoma training cohort

[0146] Samples: Tissue samples from 86 histopathologically confirmed schwannomas (including sporadic schwannomas and NF2-SWN schwannomas) were collected from patients surgically removed at Beijing Tiantan Hospital, Capital Medical University. Immediately after surgical resection, samples were flash-frozen in liquid nitrogen or preserved in RNA-protective solution. This study was approved by the Ethics Committee of Beijing Tiantan Hospital, Capital Medical University, and all patients signed informed consent forms.

[0147] Transcriptome sequencing: 50 mg of total RNA was extracted from each tissue sample using TRIzol™ Reagent (Invitrogen, Cat. 15596026). The RNA integrity index (RIN value) of all total RNA samples was detected using an Agilent 2100 bioanalyzer. Samples with a RIN value ≥ 7 were identified as qualified RNA samples. Library construction was carried out using the qualified RNA samples, followed by sequencing library construction. Paired-end sequencing was performed using a high-throughput sequencing platform. After quality control and alignment with the adult human reference genome, the expression levels of 27,095 genes were quantified using FPKM to obtain the gene expression matrix.

[0148] Molecular typing: The obtained gene expression matrix was first filtered for low-expression genes, normalized for sequencing depth, and transformed using log2. The coefficient of variation of each gene expression was calculated, and the top 2000 hypervariable genes were screened. Unsupervised clustering was performed using nonnegative matrix factorization (NMF) in R language, with k set to values ​​of 2-6 and each parameter repeated 50 times. The cophenetic coefficient of the clustering results of each group was calculated to determine the optimal number of clusters k=2. The 86 samples were divided into two molecular subtypes: immune enrichment type (IES, 46 cases) and immune exhaustion type (IDS, 40 cases). Subsequently, the independent validation cohort dataset was retrieved, and the same data preprocessing workflow was used for typing. Silhouette coefficient analysis was performed to evaluate the subtype differentiation effect and verify the stability of typing. Figure 7 In this embodiment, NMF clustering is used to generate corresponding subtype classification labels for each sample, namely IES subtype labels and IDS subtype labels. All sample subtype labels correspond one-to-one with the gene expression matrix sample information, serving as the grouping reference for subsequent differential expression analysis of different subtypes.

[0149] Example 2: Screening of combinations of genotyping markers

[0150] Using the IES and IDS subtype classification labels corresponding to the 86 samples in Example 1 as a reference, the following steps were taken to screen the genotyping markers: (1) Differential expression analysis: The FPKM expression matrix of all 27,095 quantitative genes was transformed by log2(FPKM+1). After removing the low-expression genes with less than 10 samples of standardized expression level ≥10 in each subtype group, the Mann-Whitney U test was used to perform differential expression analysis between IES and IDS subtypes. The significance threshold was set by using the multiple test P value (padj) < 0.05 after Benjamini-Hochberg correction. (2) Machine learning feature screening: Random forest (R language randomForest package, v4.7-1.1) combined with Boruta algorithm (R language Boruta package, v8.0.0) was used to evaluate the importance of features and perform iterative screening. The random forest parameters are set as follows: number of decision trees ntree = 1000, number of features randomly selected mtry = √p (p is the total number of features, rounded down) per split, using the Mean DecreaseGini (MDG) index to evaluate the feature importance of each gene; the Boruta algorithm parameters are set as follows: maximum number of iterations maxRuns = 500, significance level P = 0.01, the genes with the judgment result of "Confirmed" and "Tentative" are included in the candidate range; after gradual reduction, the subtype-related candidate gene set is obtained; (3) Combination exhaustive search and performance comparison: the top 15 genes in the candidate gene set with single gene typing AUC value are used to form a candidate pool, and 4 genes are randomly selected from them to exhaust all C(15,4)=1,365 possibilities. For each combination, log2 transformation and Z-score standardized expression value are used to establish a logistic regression classification model. Leave-one-out cross-validation (LOOCV) and 10-fold cross-validation are used to calculate the area under the receiver operating characteristic curve (AUC), sensitivity, specificity and Youden index. The cross-validation AUC is maximized and the specificity is ≥95% as the comprehensive screening criteria. After iterative comparison, the combination of 4 genes, PILRA, TNFSF8, AIF1 and STAC3, has the highest LOOCV AUC (0.9266), the fewest number of genes and the lowest detection cost. Finally, it is determined to be the biomarker combination of the present invention (Table 1). The epigenetic AUC of this 4-gene combination in the training cohort was 0.934 (95% CI: 0.870–0.983), the LOOCV AUC was 0.927 (95% CI: 0.863–0.978), and the 10-fold cross-validation AUC was 0.923. At the decision threshold P = 0.7088 determined by maximizing the Youden index, the sensitivity was 82.6% (38 / 46), the specificity was 100% (40 / 40), and the Youden index was 0.826.All four genes were relatively highly expressed in the IES-type samples (IES / IDS expression ratio 1.6–2.0 fold), and the single-gene discrimination AUC ranged from 0.885 to 0.927 (STAC3 was the highest at 0.927). Figure 2 However, single-gene models struggle to balance sensitivity and specificity (the highest sensitivity is only 82.6%, and sensitivity drops to 78.3% when specificity reaches 100%), and they exhibit poor cross-cohort stability (e.g., Figure 2 As shown, the AUC of a single gene fluctuates across cohorts by 0.04–0.09, while the fluctuation of the 4-gene combination in this invention is only 0.015, indicating insufficient clinical applicability; in contrast, the LOOCV AUC of the 4–6 gene combination is ≥0.919 (e.g., Figure 1 As shown, the four-gene combination of this invention achieves a sensitivity of 82.6% and a specificity of 100%, with AUC fluctuations of <0.01 between different sub-combinations. The model stability is significantly better than that of the single-gene model, demonstrating the synergistic effect and clinical necessity of multi-gene combinations.

[0151] Example 3: Construction and Internal Validation of the Fractal Model

[0152] The expression levels of the four genes were transformed using log2(FPKM+1) and standardized by Z-score according to the mean μ and standard deviation σ listed in Table 2. A logistic regression genotyping model was trained (see above for specific models), and the decision threshold P=0.7088 was determined by maximizing the Youden index. In the training cohort of 86 cases, the model's apparent AUC was 0.934 (95% CI: 0.870–0.983), sensitivity was 82.6% (38 / 46), specificity was 100% (40 / 40), and accuracy was 90.7% (78 / 86). After leave-one-out cross-validation (LOOCV), the AUC was 0.927 (95% CI: 0.863–0.978), sensitivity was 78.3% (36 / 46), specificity was 100% (40 / 40), and accuracy was 88.4% (76 / 86). All 40 IDS samples in the confusion matrix were correctly identified (Table 2). Figure 1 , Figure 2 ).

[0153] Example 4: Systematic verification of the optimality of marker combination

[0154] Using the top 15 genes with the highest single-gene AUC values ​​as a candidate pool (in descending order of AUC value: STAC3, PILRA, PARVG, LGALS9, RIPK3, ARHGAP9, GALNT6, CD37, BIN2, PAK1, SIGLEC7, CAPG, ARL11, FBP1, NCKAP1L), all 1,365 possible 4-gene combinations (C(15,4)=1,365) were exhaustively enumerated. The epigenetic AUC of the combination in this invention ranks in the top 2.3%, which is better than 97.8% of the combinations. Then, the combination was randomly selected from the above candidate gene pool. 3,000 four-gene combinations were selected as controls (total combinations exceed 670,000, making exhaustive testing impossible, so random sampling was used). The combination of this invention outperformed 99.2% of these combinations. The epitome of the best four-gene combination in the candidate pool (STAC3+PILRA+ARHGAP9+BIN2) after LOOCV correction decreased to 0.925, lower than the combination of this invention (0.927), indicating that its epitome advantage stemmed from overfitting. Marginal benefit analysis of the number of genes from 1 to all candidate genes showed that the LOOCV AUC reached a plateau at 4-6 genes and gradually decreased after exceeding 6 genes. Figure 3 , Figure 4 , Figure 6 (Table 5). Based on the above verification and marginal benefit analysis, a 4-6 gene combination with a small number of genes and a high LOOCV AUC was selected during the plateau period.

[0155] Example 5: Multi-platform verification of external queues

[0156] The generalization ability of the 4-gene genotyping model was validated in two independent cohorts: the NF2-SWN transcriptome sequencing cohort (HRA002762, n=71, 34 cases of IES type and 37 cases of IDS type) and the GEO expression profiling array cohort (GSE141801, GSE108524 and GSE54934, n=140, 61 cases of IES type and 79 cases of IDS type). The expression data of each cohort were transformed by log2 and standardized, and then entered into formula (1) to calculate the genotyping probability. The AUC of genotyping in each cohort was greater than 0.93, indicating that the model can be stably implemented across cohorts and sequencing platforms.

[0157] In the NF2-SWN independent validation cohort (n=71): the frozen model AUC calculated directly using the training cohort parameters was 0.946; the AUC after in-cohort standardization was 0.944. At a fixed decision threshold (P=0.7088), the sensitivity was 79.4% (27 / 34), specificity was 91.9% (34 / 37), and accuracy was 85.9% (61 / 71) (Table 6). Figure 5 ).

[0158] In the GEO expression profiling chip cohort (n=140, 61 cases of IES type and 79 cases of IDS type): the frozen model AUC was 0.931; the in-cohort standardized AUC was 0.932, with a sensitivity of 82.0% (50 / 61), specificity of 92.4% (73 / 79), and accuracy of 87.9% (123 / 140) at a fixed threshold. The optimal threshold determined by the Youden index in this cohort was 0.7251, which is close to the fixed threshold of 0.7088 determined in the training cohort, both falling within the range of 0.70–0.73, indicating that the decision threshold has good cross-cohort stability. The genotyping performance of the two cohorts was highly consistent (Tables 6-1, 6-2, and 6-3). Figure 5 It should be noted that the freezing threshold varies systematically across different sequencing platforms. In practical applications, it is recommended to perform intra-batch standardization for each batch of included control samples to correct for baseline differences between platforms.

[0159] Additionally, it should be noted that when the gene marker combination consists of 6 genes, the subtype determination threshold is P≥0.7165 for immune-enriched schwannomas and P<0.7165 for immune-depleted schwannomas.

[0160] When the gene marker combination consists of 5 genes, the subtype determination threshold is P≥0.7081 for immune-enriched schwannomas and P<0.7081 for immune-depleted schwannomas.

[0161] When the gene marker combination consists of 4 genes, the subtype determination threshold is P≥0.7088 for immune-enriched schwannomas and P<0.7088 for immune-depleted schwannomas.

[0162] Table 6-1 Gene performance parameters of 86 training sets

[0163]

[0164] Table 6-2. Performance parameters of genes 4-6 in the NF2-SWN validation group

[0165]

[0166] Table 6-3. Performance parameters of genes 4-6 in the GEO validation cohort.

[0167]

[0168] It should also be noted that the difference between the 4-gene combination 1 (PILRA, STAC3, AIF1, and TNFSF8) (hereinafter referred to as the 4 genes of this invention) and the 4-gene combination 2 (TNFSF8, NCKAP1L, PILRA, AIF1) (hereinafter referred to as the 4 genes of the control group) in the table above is that the control group 4 genes lack STAC3, while the 4 genes of this invention use PILRA and STAC3 as the core genes, which work synergistically with AIF1 and TNFSF8 to achieve optimal performance parameters. As can be seen from the table, in the results of multi-platform validation in the external cohort, the frozen AUC, in-cohort AUC, in-cohort LOOCV, sensitivity, specificity, and accuracy of the 4-gene combination 1 (PILRA, STAC3, AIF1, and TNFSF8) of this invention are all higher than those of the control group 4 genes (TNFSF8, NCKAP1L, ...). PILRA and AIF1) indicate that the 4-gene combination of the present invention is exactly at the optimal point of "information content-redundancy" - adding 1-2 genes will cause overfitting and lead to a decrease in cross-validation performance, while removing the core genes will result in insufficient information and a greater decrease in performance. Therefore, the specific 4 genes of the present invention are the optimal and concise combination.

[0169] Example 6A: Design and Validation of qPCR Primers

[0170] Primer design: Taking four gene biomarkers as an example, primer-probe combinations for the TaqMan probe method were designed using Primer Express v3.0 software for the four biomarker genes (PILRA, STAC3, AIF1, and TNFSF8) and the internal reference gene GAPDH. Design principles: Amplicon length 80-150 bp, spanning exon-exon junctions to avoid genomic DNA contamination; primer Tm 57.7-62°C; probe Tm 65-70°C; GC content 40-65%. Primer-probe sequences are shown in Table 2 above. For the four biomarker genes, the primers and probes are SEQ ID NO.1~3, SEQ ID NO.4~6, SEQ ID NO.10~12, and SEQ ID NO.13~15.

[0171] Primer verification:

[0172] (1) Specificity verification: Using cDNA from schwannoma tissue as a template, qPCR amplification was performed and melting curve analysis was performed. All primers showed a single melting peak, and there were no primer dimers or non-specific amplification.

[0173] (2) Amplification efficiency verification: qPCR amplification was performed using cDNA standards diluted 10-fold (1:10, 1:100, 1:1000, 1:10000), and a standard curve was plotted. The results showed that the amplification efficiency of each gene primer was between 93-107%, with R2>0.99, which met the MIQE guidelines.

[0174] (3) Repeatability verification: Six repeated tests were performed on the same sample. The coefficient of variation (CV) of Ct values ​​of each gene was <3%, and the CV of the internal reference gene GAPDH was <2%.

[0175] Example 6B: Composition and Use of Low-Throughput qPCR Kit

[0176] The kit consists of the following components as described in Section 2.2 of the technical solution: 4 gene-specific primer and probe sets, GAPDH internal control primer and probe sets, reverse transcription reagent, qPCR premix, negative control (NTC), RNA extraction reagent, quality control standards and genotyping card.

[0177] Usage instructions (e.g.) Figure 9 (as shown)

[0178] (1) RNA extraction: Take about 10-30 mg of schwannoma tissue (freshly frozen) and extract total RNA using the TRIzol method or a column extraction kit. RNA quality was assessed as follows: fresh tissue RIN ≥ 7; A260 / A280 = 1.8-2.1;

[0179] (2) Reverse transcription: Take 500 ng of total RNA and use the reverse transcription reagent provided in the kit to reverse transcribe under the following conditions: 25℃ for 10 minutes, 50℃ for 30 minutes, 85℃ for 5 minutes, and store at 4℃;

[0180] (3) qPCR detection: After diluting the cDNA 5 times, perform the qPCR reaction according to the following system:

[0181] 2x qPCR premix: 10 uL

[0182] Upstream primer (10 uM): 0.5 uL

[0183] Downstream primer (10 uM): 0.5 uL

[0184] Probe (5 uM): 0.5 uL

[0185] cDNA template: 2 uL

[0186] Nuclease-free water: Add to a final volume of 20 uL

[0187] Reaction conditions: 95℃ for 30 seconds for pre-denaturation; 95℃ for 5 seconds, 60℃ for 30 seconds (40 cycles).

[0188] Three replicates were set up for each gene in each sample, and a negative control was run simultaneously.

[0189] (4) Data analysis: The relative expression levels of each gene were calculated using the ΔΔCt method (with GAPDH as the internal reference and IDS-type positive control as the calibration sample), and log2(x+1) transformation and Z-score standardization were performed.

[0190] (5) Subtyping interpretation: Substitute the standardized expression values ​​into the formula to calculate P(IES), and determine the subtype according to the threshold of 0.7088. Use the matching interpretation card: map the Z-score values ​​of the 4 genes to the scale on the interpretation card, calculate the weighted sum to obtain the S value, and look up P(IES) and the subtyping results in the table.

[0191] The calculation formula is as follows:

[0192] z_i = (log2(E_i + 1) - μ_i) / σ_i

[0193] S = β0 + Σ β_i · z_i

[0194] P(IES) = 1 / (1 + e^(-S))

[0195] In the formula: E_i is the original expression level of the i-th gene; μ_i and σ_i are the mean and standard deviation of the log2 transformed expression value of the gene in the training queue (n=86), respectively; β0 is the intercept, and β_i is the regression coefficient of the i-th gene.

[0196] The model parameters are shown in Table 3 above.

[0197] The epigenetic AUC of this 4-gene model in the training cohort was 0.9342.

[0198] The subtype determination threshold is P≥0.7088 for immune-enriched vestibular schwannoma and P<0.7088 for immune-depleted vestibular schwannoma.

[0199] (6) Output the classification report.

[0200] Experimental results show that the detection time is approximately 4 hours.

[0201] The above embodiments demonstrate that the gene marker combination and its typing model provided by the present invention can accurately and stably achieve molecular typing of vestibular schwannomas. The model calculation process is simple and completely reproducible, and the low-throughput reagent kit design makes routine clinical testing possible.

[0202] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-described technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A combination of gene markers for molecular subtyping of vestibular schwannomas, characterized in that, It includes PILRA, STAC3, and combinations of at least two genes selected from TNFSF8, NCKAP1L, AIF1, and SLC7A7.

2. The gene marker combination for molecular subtyping of vestibular schwannomas according to claim 1, characterized in that, It includes STAC3 and PILRA, as well as a combination of two genes selected from TNFSF8, NCKAP1L, AIF1, and SLC7A7.

3. The gene marker combination for molecular subtyping of vestibular schwannomas according to claim 2, characterized in that, The combination of gene markers is PILRA, TNFSF8, AIF1, and STAC3.

4. The gene marker combination for molecular subtyping of vestibular schwannomas according to claim 1, characterized in that, It is composed of PILRA, AIF1, STAC3 and any one of the genes selected from TNFSF8, NCKAP1L or SLC7A7.

5. The gene marker combination for molecular subtyping of vestibular schwannomas according to claim 4, characterized in that, The combination of gene markers is PILRA, TNFSF8, NCKAP1L, AIF1, and STAC3.

6. The gene marker combination for molecular subtyping of schwannomas according to claim 1, characterized in that, Including PILRA, TNFSF8, NCKAP1L, AIF1, STAC3, and SLC7A7.

7. A detection kit for molecular subtyping of vestibular schwannomas, characterized in that, It comprises: a specific upstream primer, a downstream primer, and a fluorescent probe for the transcript of the combination of gene markers according to any one of claims 1 to 6, wherein the sequences of the upstream primer, the downstream primer, and the fluorescent probe include SEQ ID NO. 1 to 3, SEQ ID NO. 13 to 15, SEQ ID NO. 19 to 21, and any 2 to 4 sequences selected from sequences SEQ ID NO. 4 to 6, SEQ ID NO. 7 to 9, SEQ ID NO. 10 to 12, and SEQ ID NO. 16 to 18.

8. The detection kit for molecular subtyping of vestibular schwannomas according to claim 7, characterized in that, Also includes: Internal control gene detection reagent, reverse transcription reagent, probe method premixed solution and negative control; The internal reference gene detection reagent includes a primer and probe set targeting the internal reference gene GAPDH. The reverse transcription reagent includes reverse transcriptase, random primers, dNTPs, RNase inhibitors, and reverse transcription buffer. The probe-based premix includes Taq DNA polymerase, dNTPs, MgCl2, PCR buffer, upstream primer, downstream primer, and fluorescent probe. The negative control is a template-free control, which uses nuclease-free water instead of cDNA from the test sample as a template.

9. The detection kit for molecular subtyping of vestibular schwannomas according to claim 8, characterized in that, It also includes the following components: RNA extraction reagents and DNase I; Preferably, it also includes: quality control standards and typing interpretation tools; The quality control standards include standards for each transcript in the combination of the gene markers at known concentrations; The typing interpretation tool is configured to determine and output the molecular typing results of vestibular schwannomas based on the detection results of each transcript, and the molecular typing includes immune enrichment type and immune depletion type.

10. The combination of gene markers for molecular subtyping of vestibular schwannoma as described in any one of claims 1 to 6 or the detection kit for molecular subtyping of vestibular schwannoma as described in any one of claims 7 to 9, in basic molecular biological research on schwannoma; or, in in vitro cell experiments for screening targeted drugs for schwannoma for non-diagnostic purposes.

Images