Marker combination for predicting thyroid cancer metastasis and application thereof

Abstract

The invention provides a marker combination for predicting thyroid cancer metastasis and application thereof. Specifically, the invention provides application of a gene, mRNA, cDNA and protein of the risk marker combination for thyroid cancer metastasis judgment or a detection reagent of the gene, the mRNA, the cDNA and the protein, and is used for preparing / establishing a diagnostic reagent or a kit / equipment for judging the thyroid cancer metastasis occurrence risk. Researches show that the thyroid cancer metastasis risk marker combination can be used as a marker for early judgment of thyroid cancer metastasis of a thyroid cancer patient, has high sensitivity and specificity, can quickly diagnose thyroid cancer metastasis at a relatively early disease progress stage, and provides powerful assistance for early treatment intervention of diseases.

Description

Marker combination for predicting thyroid cancer metastasis and its application Technical Field

[0001] The present invention relates to the field of medical diagnosis, and more particularly to a marker combination for predicting thyroid cancer metastasis and an application thereof. Background Art

[0002] Globally, approximately 567,000 cases of thyroid cancer occur annually, ranking ninth in incidence among all cancer types. The incidence rate in women is approximately three times that of men, at 10.2 per 100,000 people. Since the 1980s, thyroid cancer incidence has been steadily increasing in many countries, primarily due to improvements in detection and diagnostic methods, particularly increased detection of papillary thyroid carcinoma (PTC).

[0003] PTC accounts for approximately 84% of thyroid cancer patients and is the most common type of thyroid malignancy. Between 1974 and 2013, the average incidence of PTC in the United States was approximately 6.66%, with an incidence-related mortality rate of 0.20%. Although PTC is generally considered a slow-acting tumor, some cancer cells will metastasize to nearby lymph nodes, especially the central and lateral cervical lymph nodes. Lymph node metastasis (LNM), which usually occurs first in the central region, is a key factor in PTC prognosis, treatment strategy, and recurrence rate, and is also associated with lower survival rates.

[0004] Several approaches have been proposed to assess the risk of LNM in PTC patients, including consideration of factors such as tumor size, location, extension, microcalcifications, and Hashimoto's disease. Some studies have also included blood markers such as TSH and TGAb. Radiomics has recently attracted attention in precision diagnosis, and radiomics-based techniques have been proposed to predict LNM in PTC patients by converting ultrasound images into analyzable data. These techniques extract features from ultrasound images, including intensity, edges, texture, and wavelets, and establish a correlation between these high-throughput features and LNM status. In previous studies, whether based on clinical data or radiomics, the prediction performance of LNM has been suboptimal due to difficulties in ensuring the integrity of image feature extraction. The area under the receiver operating characteristic (ROC) curve (AUC) for a single test set ranged from approximately 0.67 to 0.78.

[0005] Therefore, in this field, there is an urgent need to develop a method that can screen and diagnose thyroid cancer lymph node metastasis more simply, earlier, more effectively and more accurately, so as to achieve timely intervention and treatment of thyroid cancer lymph node metastasis.

[0006] Summary of the Invention

[0007] The present invention provides a method for early screening and diagnosis of thyroid cancer metastasis that is simpler, more effective, earlier and more accurate.

[0008] In a first aspect of the present invention, there is provided a use of a gene, mRNA, cDNA, protein, or a detection reagent thereof for a thyroid cancer metastasis risk marker for preparing a diagnostic reagent or kit for determining whether a thyroid cancer patient has thyroid cancer metastasis;

[0009] Wherein, the thyroid cancer metastasis risk markers include: (A1) RPS4Y1; (A2) PKHD1L1; and (A3) CRABP1.

[0010] In another preferred embodiment, the detection reagent includes a primer pair, and the primer pair includes: a marker primer pair for specifically amplifying the mRNA or cDNA of the thyroid cancer metastasis risk marker combination.

[0011] In another preferred embodiment, the kit further comprises: an internal reference primer pair for amplifying an internal reference gene.

[0012] In another preferred embodiment, the marker primer pair comprises:

[0013] (P1) Primer pair for amplifying RPS4Y1: SEQ ID NO: 1 and SEQ ID NO: 2;

[0014] (P2) Primer pair for amplifying PKHD1L1: SEQ ID NO: 3 and SEQ ID NO: 4;

[0015] (P3) primer pair for amplifying CRABP1: SEQ ID NO: 5 and SEQ ID NO: 6; and / or

[0016] The internal reference primer pair is SEQ ID NO: 7 and SEQ ID NO: 8.

[0017] In another preferred embodiment, the detection reagent further comprises a probe used in conjunction with the primer pair.

[0018] The probes include: probe 1: SEQ ID NO: 9, used in conjunction with a primer pair for amplifying RPS4Y1; probe 2: SEQ ID NO: 10, used in conjunction with a primer pair for amplifying PKHD1L1; probe 3: SEQ ID NO: 11, used in conjunction with a primer pair for amplifying CRABP1; and probe 4: SEQ ID NO: 12, used in conjunction with a primer pair for amplifying an internal reference gene.

[0019] In another preferred example, the diagnostic reagent or kit obtains an assessment result of the risk of thyroid cancer metastasis by detecting the difference ΔCt between the Ct value of each marker in qPCR amplification and the Ct value of the internal reference gene in thyroid protocarcinoma tissue, namely, three thyroid cancer metastasis risk markers (A1) RPS4Y1; (A2) PKHD1L1; and (A3) CRABP1.

[0020] In another preferred embodiment, the markers A1 to A3 are selected from Table A:

[0021] In another preferred embodiment, the gene, mRNA, cDNA, or protein of any marker listed in Table A of thyroid cancer metastasis risk markers is derived from human.

[0022] In another preferred embodiment, the diagnostic reagent or kit is used to detect thyroid cancer metastasis risk markers in thyroid proto-cancer tissue.

[0023] In another preferred embodiment, the thyroid cancer includes papillary thyroid carcinoma.

[0024] In another preferred embodiment, the thyroid cancer metastasis includes thyroid cancer lymph node metastasis;

[0025] In another preferred embodiment, the thyroid cancer lymph node metastasis includes: central thyroid cancer lymph node metastasis, lateral cervical lymph node metastasis of thyroid cancer, or a combination thereof.

[0026] In another preferred embodiment, the detection reagent includes: (a) specific antibodies or specific binding molecules for the thyroid cancer metastasis risk marker; and / or (b) primers or primer pairs, probes or chips (such as nucleic acid chips or protein chips) that specifically amplify the mRNA or cDNA of the thyroid cancer metastasis risk marker.

[0027] In another preferred embodiment, the detection reagent includes a primer pair or a probe, and the primer pair or probe is a primer pair or a probe that specifically amplifies the mRNA or cDNA of the thyroid cancer metastasis risk marker.

[0028] In another preferred embodiment, the primer pair includes a primer pair selected from the following group: a primer pair for amplifying RPS4Y1: SEQ ID NO: 1, SEQ ID NO: 2; a primer pair for amplifying PKHD1L1: SEQ ID NO: 3, SEQ ID NO: 4; a primer pair for amplifying CRABP1: SEQ ID NO: 5, SEQ ID NO: 6; or a combination thereof.

[0029] In another preferred embodiment, the detection reagent further includes the following internal reference primer pair: SEQ ID NO: 7, SEQ ID NO: 8.

[0030] In another preferred embodiment, the diagnostic reagent or kit is used to detect the expression level of a thyroid cancer metastasis risk marker in a test sample.

[0031] In a second aspect of the present invention, a kit is provided, comprising a detection reagent for detecting genes, mRNA, cDNA, proteins, or a combination thereof of thyroid cancer metastasis risk markers in a sample to be tested;

[0032] Wherein, the thyroid cancer metastasis risk markers include: (A1) RPS4Y1; (A2) PKHD1L1; and (A3) CRABP1.

[0033] In another preferred embodiment, the detection reagent includes: (a) specific antibodies or specific binding molecules for the thyroid cancer metastasis risk marker; and / or (b) primers or primer pairs, probes or chips (such as nucleic acid chips or protein chips) that specifically amplify the mRNA or cDNA of the thyroid cancer metastasis risk marker.

[0034] In another preferred embodiment, the detection reagent includes a primer pair or a probe, and the primer pair or probe includes a primer pair or a probe that specifically amplifies the mRNA or cDNA of the thyroid cancer metastasis risk marker.

[0035] In another preferred embodiment, the detection reagent includes a primer pair, and the primer pair includes: a marker primer pair for specifically amplifying the mRNA or cDNA of the thyroid cancer metastasis risk marker combination.

[0036] In another preferred embodiment, the kit further comprises: an internal reference primer pair for amplifying an internal reference gene.

[0037] In another preferred embodiment, the internal reference gene is GAPDH.

[0038] In another preferred embodiment, the marker primer pair comprises:

[0039] (P1) Primer pair for amplifying RPS4Y1: SEQ ID NO: 1 and SEQ ID NO: 2;

[0040] (P2) Primer pair for amplifying PKHD1L1: SEQ ID NO: 3 and SEQ ID NO: 4;

[0041] (P3) primer pair for amplifying CRABP1: SEQ ID NO: 5 and SEQ ID NO: 6; and / or

[0042] The internal reference primer pair is SEQ ID NO: 7 and SEQ ID NO: 8.

[0043] In another preferred embodiment, the detection reagent further comprises a probe used in conjunction with the primer pair.

[0044] The probes include: probe 1: SEQ ID NO: 9, used in conjunction with a primer pair for amplifying RPS4Y1; probe 2: SEQ ID NO: 10, used in conjunction with a primer pair for amplifying PKHD1L1; probe 3: SEQ ID NO: 11, used in conjunction with a primer pair for amplifying CRABP1; and probe 4: SEQ ID NO: 12, used in conjunction with a primer pair for amplifying an internal reference gene.

[0045] In another preferred embodiment, the primer pair comprises a primer pair selected from the group consisting of: a primer pair for amplifying RPS4Y1: SEQ ID NO: 1, SEQ ID NO: 2; a primer pair for amplifying PKHD1L1: SEQ ID NO: 3, SEQ ID NO: 4; a primer pair for amplifying CRABP1: SEQ ID NO: 5, SEQ ID NO: 6; or a combination thereof. In another preferred embodiment, the detection reagent further comprises the following internal reference primer pair: SEQ ID NO: 7, SEQ ID NO: 8.

[0046] In another preferred embodiment, the detection reagent is coupled with or carries a detectable label.

[0047] In another preferred embodiment, the detectable label is selected from the following group: a chromophore, a chemiluminescent group, a fluorophore, an isotope or an enzyme.

[0048] In another preferred embodiment, the antibody is a monoclonal antibody or a polyclonal antibody.

[0049] In another preferred embodiment, the diagnostic reagent includes antibodies, primers, probes, sequencing libraries, nucleic acid chips (such as DNA chips) or protein chips.

[0050] In another preferred embodiment, the nucleic acid chip includes a substrate and specific oligonucleotide probes spotted on the substrate, and the specific oligonucleotide probes include probes that specifically bind to any polynucleotide (mRNA or cDNA) of the thyroid cancer metastasis risk marker.

[0051] In another preferred embodiment, the protein chip comprises a substrate and specific antibodies spotted on the substrate, and the specific antibodies comprise specific antibodies against the thyroid cancer metastasis risk marker.

[0052] In another preferred embodiment, the antibody is a monoclonal antibody or a polyclonal antibody.

[0053] In another preferred embodiment, the kit contains genes, mRNA, cDNA and / or proteins of thyroid cancer metastasis risk markers as reference substances or quality control substances.

[0054] In another preferred embodiment, the kit further includes a label or instructions, which indicate that the kit is used for (a) determining the risk of thyroid cancer metastasis, and / or (b) evaluating the therapeutic effect of thyroid cancer metastasis.

[0055] In another preferred embodiment, the reagents include primers, probes, gRNA or a combination thereof, more preferably primer pairs or probes for PCR, qPCR, or RT-PCR.

[0056] In another preferred embodiment, the thyroid cancer metastasis risk marker can be detected by the following methods: sequencing, PCR, or a combination thereof.

[0057] In another preferred embodiment, the detection of the thyroid cancer metastasis risk marker can be quantitative.

[0058] In a third aspect of the present invention, a detection method is provided, comprising the steps of:

[0059] (a) providing a test sample;

[0060] (b) detecting the expression level of a thyroid cancer metastasis risk marker gene in the test sample, recorded as C1; and

[0061] (c) comparing the concentration of the thyroid cancer metastasis risk marker C1 with a control reference value C0, wherein the thyroid cancer metastasis risk marker comprises:

[0062] (A) any one marker selected from A1 to A8, or a combination thereof: (A1) RPS4Y1; (A2) PKHD1L1; and (A3) CRABP1;

[0063] If the test result of the thyroid cancer metastasis risk of the test subject meets the following conditions, it indicates that the subject has a high risk of thyroid cancer metastasis:

[0064] When the expression level of a marker in Table A in the test subject is significantly different from the reference value or standard value, the risk of thyroid cancer metastasis in the thyroid cancer patient is high.

[0065] In a fourth aspect of the present invention, a device for early screening of thyroid cancer metastasis is provided, comprising:

[0066] (a) an input module, wherein the input module is used to input expression data of characteristic genes of a thyroid cancer patient;

[0067] Wherein, the characteristic genes include: (A1) RPS4Y1; (A2) PKHD1L1; and (A3) CRABP1;

[0068] (b) a processing module or a risk assessment module, wherein the module is configured to: analyze and calculate the input characteristic genes to obtain a risk assessment; and through differential expression analysis, find that when (A1) RPS4Y1, (A2) PKHD1L1, and (A3) CRABP1 are significantly upregulated, it indicates that the risk of thyroid cancer metastasis in the thyroid cancer patient is high; otherwise, it indicates that the risk of thyroid cancer metastasis in the thyroid cancer patient is low; and

[0069] (c) an output module, wherein the output module is used to output the auxiliary screening results.

[0070] In another preferred embodiment, the risk assessment module is configured to: when performing risk assessment, use a random forest model to perform risk assessment on the input expression data.

[0071] In another preferred embodiment, the risk assessment module is configured to use the random forest model formula shown in the following formula (I) to perform risk assessment:

[0072] The random forest model is a trained and optimized model, T is the number of decision trees, and x1, x2, and x3 correspond to the ΔCt values ​​of the characteristic genes (A1) RPS4Y1; (A2) PKHD1L1; and (A3) CRABP1, respectively.

[0073] In another preferred example, in the random forest model, T=41; and / or the weights of (A1) RPS4Y1; (A2) PKHD1L1; and (A3) CRABP1 markers integrated into the model are 0.484, 0.281, and 0.235, respectively.

[0074] In another preferred example, when y≥0.5, it indicates that the thyroid cancer patient has a high risk of metastasis, otherwise it indicates that the metastasis risk is low.

[0075] In another preferred embodiment, the expression data is obtained using a detection reagent:

[0076] Wherein, the detection reagent includes:

[0077] (P1) Primer pair for amplifying RPS4Y1: SEQ ID NO: 1 and SEQ ID NO: 2 and probe 1 used in conjunction with it: SEQ ID NO: 9;

[0078] (P2) Primer pair for amplifying PKHD1L1: SEQ ID NO: 3 and SEQ ID NO: 4 and probe 2: SEQ ID NO: 10;

[0079] (P3) primer pair for amplifying CRABP1: SEQ ID NO: 5 and SEQ ID NO: 6 and probe 3 used in conjunction with it: SEQ ID NO: 11; and

[0080] The internal reference primer pair is SEQ ID NO: 7 and SEQ ID NO: 8 and the probe 4 used in conjunction therewith is SEQ ID NO: 12.

[0081] In another preferred example, the device further comprises a detection module, and the detection module is used to detect the mRNA level, protein level, or protein activity of the risk marker.

[0082] In another preferred embodiment, the sample to be tested is thyroid protocarcinoma tissue.

[0083] In another preferred embodiment, the detection module is selected from the following group: an ELISA analyzer, a PCR detector, a sequencer, or a combination thereof.

[0084] In a fifth aspect of the present invention, a method for detecting the expression level of a combination of thyroid cancer metastasis risk markers is provided, comprising the steps of:

[0085] (a) providing a test sample;

[0086] (b) extracting total RNA from the sample;

[0087] (c) reverse transcribing the product RNA obtained in step (b);

[0088] (d) performing fluorescent quantitative PCR on the reverse transcription product obtained in step (c) to obtain the expression level of the risk marker gene;

[0089] The risk marker combination includes: (A1) RPS4Y1; (A2) PKHD1L1; and (A3) CRABP1.

[0090] In another preferred embodiment, the method is a non-diagnostic and non-therapeutic method.

[0091] In another preferred embodiment, the method is an in vitro method.

[0092] It should be understood that within the scope of the present invention, the above-mentioned technical features of the present invention and the technical features described in detail below (such as in the embodiments) can be combined with each other to form new or preferred technical solutions. Due to space limitations, they will not be listed here one by one. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 shows a two-dimensional embedding of thyroid tumor cells (principal components 1 and 2); cells are classified based on whether they were from patients with central / neck tumor metastasis (positive) or without (negative).

[0094] Figure 2 shows the AUROC values ​​of the random forest model using different numbers of genes as input; genes were selected based on the mutual information between gene expression and class labels.

[0095] Figure 3 shows the AUROC curve of the final random forest model using 3 genes as input.

[0096] FIG4 shows the algorithm development and clinical trial process for predicting thyroid cancer metastasis of the present invention.

[0097] Figure 5 shows the distribution of ΔCt values ​​of three genes (RPS4Y1, PKHD1L1, and CRABP1) in the metastatic group and non-metastatic group in the total data set of three centers (Nantong (center B), Zhongshan (center A), and Xuzhou (center C)) (RPS4Y1: P-value < 0.00001; PKHD1L1: P-value < 0.00001; CRABP1: P-value < 0.00001).

[0098] Figure 6 shows the receiver operating characteristic (ROC) curves of the random forest model in predicting the occurrence of metastasis in the three datasets.

[0099] Figure 7 shows the matrix analysis of the random forest model on the dataset of center B.

[0100] Figure 8 shows the matrix analysis of the random forest model on the dataset of center A.

[0101] Figure 9 shows the matrix analysis of the random forest model on the dataset of center B.

[0102] FIG10 shows the receiver operating characteristic (ROC) curve of 355 thyroid papillary carcinoma samples measured using the kit of the present invention. DETAILED DESCRIPTION

[0103] After extensive and in-depth research and extensive screening, the inventors unexpectedly discovered a more effective, rapid, and accurate method and device for early screening of thyroid cancer metastasis based on characteristic genes. Specifically, the inventors found that a combination of three genes (RPS4Y1, PKHD1L1, and CRABP1) has a very high accuracy rate for early detection or screening based on the change or difference in expression level (such as ΔCt) relative to the internal reference gene. Compared with existing single-gene or multi-gene combinations, the detection of thyroid metastatic cancer is simpler, more effective, faster, and more accurate. The method of the present invention significantly reduces the detection cost of multi-gene (>3) combination testing. In addition, the present invention also uses this to construct a kit and screening device for simple, accurate, and efficient early screening of thyroid cancer metastasis. The method and screening device (or system) of the present invention can not only effectively distinguish patients with thyroid cancer metastasis from those without metastasis, but also have very excellent detection stability, resistance to endogenous interference, resistance to exogenous interference, clinical diagnostic specificity, clinical diagnostic sensitivity, and clinical reproducibility. Therefore, it is suitable for early classification and treatment intervention of patients with thyroid cancer metastasis. This is based on the present invention.

[0104] The present invention uses the tumor cell expression data of in situ thyroid cancer as input to construct a predictive model, and uses at least three genes (RPS4Y1, PKHD1L1 and CRABP1) of the present invention as input. Ultimately, the model can classify newly admitted patients with metastasis and non-metastasis with an accuracy of 92.5%, indicating strong performance. The implementation of this model may help clinicians make timely diagnosis of thyroid cancer metastasis. Therapeutic intervention can be carried out in the early stages and prevent the progression of thyroid cancer metastasis. The significance of this work will have the opportunity to solve the classification and screening of early thyroid cancer metastasis, which has great clinical significance for patients and hospitals in the management of thyroid cancer metastasis.

[0105] the term

[0106] As used herein, the term "sample" or "specimen" refers to material specifically associated with a subject from which specific information about the subject can be determined, calculated, or inferred. A sample may consist entirely or in part of biological material from a subject.

[0107] As used herein, the term "expression" includes the production of mRNA from a gene or gene portion, the production of a protein encoded by the RNA or gene portion, and the appearance of a detectable substance associated with expression. For example, cDNA, the binding of a binding ligand (such as an antibody) to a gene or other oligonucleotide, protein, or protein fragment, and the chromogenic portion of the binding ligand are all included within the scope of the term "expression." Thus, an increase in half-dot density on an immunoblot, such as a Western blot, also falls within the scope of the term "expression" based on biological molecules.

[0108] As used herein, the term "reference value" or "control reference value" refers to a value that is statistically correlated with a specific result when compared to the analysis result. In a preferred embodiment, the reference value is determined based on the mRNA expression and / or protein expression of thyroid cancer lymph node metastasis risk markers compared and statistically analyzed. Some such studies are shown in the Examples section of this article. However, research from the literature and user experience of the methods disclosed herein can also be used to produce or adjust reference values. Reference values ​​can also be determined by considering circumstances and results that are particularly relevant to the patient's ethnic group, medical history, genetics, age, and other factors.

[0109] As used herein, "Ct value" or "Cycle Threshold" refers to the number of amplification cycles corresponding to when the fluorescence signal of the amplified product reaches the set fluorescence threshold in quantitative PCR. Simply put, the Ct value represents the number of cycles required for the amplification of the starting template to reach a certain amount of product. The higher the starting template concentration, the smaller the Ct value; the lower the starting template concentration, the larger the Ct value.

[0110] As used in the present invention, "ΔCt value" refers to the value of the gene to be tested (R gene) compared with the internal reference. For example, after adding and subtracting the Ct value of GAPDH (the internal reference selected in the present invention), the ΔCt value of the R gene is Ct(R gene)-Ct(GAPDH gene).

[0111] Risk markers for lymph node metastasis in thyroid cancer

[0112] As used herein, the term "thyroid cancer metastasis risk marker of the present invention" refers to one or more markers shown in Table A.

[0113] In the present invention, the terms "thyroid cancer metastasis risk marker protein of the present invention", "thyroid cancer lymph node metastasis risk marker protein of the present invention", "protein of the present invention", "polypeptide of the present invention", "characteristic gene of the present invention" or "markers shown in Table A" are used interchangeably and refer to any one or more of the thyroid cancer lymph node metastasis risk markers of the present invention.

[0114] In the present invention, the terms "thyroid cancer metastasis risk marker gene", "thyroid cancer lymph node metastasis risk marker gene", and "thyroid cancer lymph node metastasis risk marker polynucleotide" can be used interchangeably and refer to the nucleotide sequence of any thyroid cancer metastasis risk marker shown in Table A.

[0115] It should be understood that when encoding the same amino acid, the substitution of nucleotides in the codon is acceptable. In addition, it should be understood that when the nucleotide substitution produces a conservative amino acid substitution, the change of nucleotides is also acceptable.

[0116] When information on thyroid cancer metastasis risk markers is obtained, a nucleic acid sequence encoding the marker can be constructed based thereon, and specific probes can be designed based on the nucleotide sequence. The full-length nucleotide sequence or a fragment thereof can usually be obtained using a PCR amplification method, a recombinant method, or an artificial synthesis method. For the PCR amplification method, primers can be designed based on the nucleotide sequence of the thyroid cancer metastasis risk marker disclosed in the present invention, especially the open reading frame sequence, and a commercially available cDNA library or a cDNA library prepared by conventional methods known to those skilled in the art is used as a template to amplify the relevant sequence. When the sequence is long, it is often necessary to perform two or more PCR amplifications, and then the fragments amplified each time are spliced ​​together in the correct order.

[0117] Once the relevant sequence is obtained, it can be obtained in large quantities by recombinant methods. This is usually done by cloning it into a vector, then transferring it into cells, and then isolating the relevant sequence from the propagated host cells by conventional methods.

[0118] In addition, the sequences can also be synthesized by artificial synthesis, especially when the fragment length is shorter. Usually, a long fragment can be obtained by synthesizing multiple small fragments and then connecting them.

[0119] Currently, DNA sequences encoding proteins of the present invention (or fragments or derivatives thereof) can be obtained entirely through chemical synthesis, which can then be introduced into various existing DNA molecules (such as vectors) and cells known in the art.

[0120] The polynucleotide sequence of the present invention can be used to express or produce recombinant thyroid cancer metastasis risk markers using conventional recombinant DNA technology.

[0121] RPS4Y1, PKHD1L1, and CRABP1

[0122] RPS4Y1 is a component of the ribosome and is involved in protein synthesis.

[0123] The PKHD1L1 gene is involved in regulating cell growth and differentiation, which is crucial for tissue development and the normal function of organs.

[0124] CRABP1 is involved in the metabolism of retinoic acid, a molecule that regulates cell differentiation and growth. Dysregulation of the retinoic acid pathway has been shown to influence tumor behavior in various cancers, including thyroid cancer. Several studies have suggested that altered CRABP1 expression may be associated with thyroid carcinogenesis and may affect tumor growth and response to treatment.

[0125] Detection method

[0126] Based on the differential expression of thyroid cancer metastasis risk markers in carcinoma in situ, the present invention also provides a corresponding method for judging the risk of thyroid cancer metastasis.

[0127] The present invention relates to diagnostic assays for quantitatively and positionally detecting protein or mRNA levels of thyroid cancer metastasis risk markers. These assays are well known in the art. The protein or mRNA levels of thyroid cancer metastasis risk markers detected in the assays can be used to determine (including assist in determining) whether a person has a risk of thyroid cancer metastasis.

[0128] A preferred method is to perform quantitative detection of mRNA or cDNA by PCR / qPCR / RT-PCR.

[0129] A preferred method is to quantitatively detect mRNA or cDNA by sequencing.

[0130] The polynucleotides of thyroid cancer lymph node metastasis risk markers can be used to diagnose the risk of thyroid cancer lymph node metastasis. Part or all of the polynucleotides of the present invention can be immobilized as probes on microarrays or DNA chips for differential expression analysis of genes under analysis and genetic diagnosis.

[0131] In addition, the present invention can also detect at the protein level. For example, antibodies against thyroid cancer lymph node metastasis risk markers can be immobilized on a protein chip to detect thyroid cancer lymph node metastasis risk proteins in a sample.

[0132] Detection kit

[0133] Based on the correlation between thyroid cancer metastasis risk markers and thyroid cancer metastasis risk, thyroid cancer metastasis risk markers can be used as markers for judging thyroid cancer metastasis risk.

[0134] The present invention also provides a kit for determining thyroid cancer metastasis risk, comprising a detection reagent for detecting genes, mRNA, cDNA, proteins, or combinations thereof, of thyroid cancer metastasis risk markers. Preferably, the kit comprises an antibody or immunoconjugate of the present invention, or an active fragment thereof, against a thyroid cancer metastasis risk marker; or a primer or primer pair, probe, or chip that specifically amplifies the mRNA or cDNA of a thyroid cancer metastasis risk marker.

[0135] In another preferred embodiment, the kit further comprises a label or instructions.

[0136] The main advantages of the present invention include:

[0137] (a) Compared with existing methods for detecting thyroid cancer metastasis, the marker combination established by the present invention uses fewer markers and has a very high detection accuracy, greatly optimizing the detection process.

[0138] (b) The prediction model provided by the present invention avoids multiple punctures in patients at high risk of thyroid cancer metastasis, greatly reduces the pain of patients, and saves medical resources.

[0139] (c) The prediction model provided by the present invention can more accurately guide the selection of clinical surgical plans, making the treatment plan more precise.

[0140] (d) The present invention can achieve very good detection results by using only three genes (RPS4Y1, PKHD1L1 and CRABP1), significantly reducing the detection cost, process and time.

[0141] The present invention will be further described below in conjunction with specific examples. It should be understood that these examples are intended to illustrate the present invention only and are not intended to limit the scope of the present invention. The experimental methods in the following examples, for which specific conditions are not specified, are generally performed under conventional conditions, such as those described in Sambrook et al., Molecular Cloning: A Laboratory Manual (New York: Cold Spring Harbor Laboratory Press, 1989), or according to the conditions recommended by the manufacturer. Unless otherwise stated, percentages and parts are by weight.

[0142] Materials and methods

[0143] The present invention uses the expression of EPCAM and TG genes as guide genes to screen out all tumor cells. In the process of building a predictive model, genes with highly variable expression are first selected to eliminate genes with low information content and optimize computational efficiency, based on the gene expression profile of tumor cells. Next, the mutual information score is used as a metric to select the dominant genes used to construct the random forest model classifier. AUROC is used to evaluate model performance and ultimately determine model selection. Other evaluation indicators are also calculated, including but not limited to recall and precision. All models were evaluated using a 5-fold cross-validation.

[0144] Example 1 Sequencing Sample Preparation

[0145] Tissue samples were obtained from six patients with central thyroid cancer metastases, eight patients with lateral neck metastases, and four patients with carcinoma in situ without pathologically detected metastases.

[0146] 1. For patients diagnosed with papillary thyroid cancer by pathology, after passing the hospital ethics review, a 5mm*5mm tissue sample from the center of the thyroid nodule was removed during surgery, placed in tissue preservation fluid, and stored in liquid nitrogen for subsequent flow cytometry and single-cell sequencing analysis.

[0147] 2. For patients diagnosed with papillary thyroid cancer by pathology, after passing the hospital ethics review, the central part of the thyroid nodule was removed during surgery, and ex vivo thyroid tumor tissue was obtained by ultrasound-guided puncture. The tissue was placed in tissue preservation fluid and stored in liquid nitrogen for subsequent flow cytometry and single-cell sequencing analysis.

[0148] Example 2 Single Cell Sequencing

[0149] Single-cell sequencing and V(D)J libraries were generated using the 10X Genomics Chromium Controller Instrument, Chromium Single Cell 5' Library, Gel Bead Kit, and V(D)J Enrichment Kit (10X Genomics, Pleasanton, CA). Cell suspensions were concentrated to 1,000 cells / μL, and approximately 10,000 of these were loaded into each channel to generate single-cell gel beads in emulsion (GEM). The output was mRNA barcodes for 6,000 cells per sample. After the reverse transcription step, the GEM was destroyed, and the barcoded cDNA was purified, amplified, and used to construct 5' gene expression and TCR and BCR enrichment libraries. For 5' library construction, the amplified barcoded cDNA was fragmented, A-tailed, ligated with adapters, and indexed PCR amplified. For the V(D)J library, human T cell and B cell V(D)J sequences were enriched from the amplified cDNA, fragmented, A-tailed, adapter-ligated, and indexed PCR amplified.

[0150] The final libraries were quantified using the Qubit high-sensitivity DNA assay (No. Q33231; Thermo Fisher Scientific, Waltham, MA, USA). Library size distribution was determined using a high-sensitivity DNA chip in a Bioanalyzer 2200 (Agilent Technologies, Santa Clara, CA, USA). All libraries were sequenced using a 150 bp paired-end run on an Illumina sequencer (Illumina, San Diego, CA, USA).

[0151] Example 3. qRT-PCR experiment

[0152] RNA extraction: First, extract total RNA from the thyroid cancer cell in situ samples using an RNA extraction kit (e.g., Takara Bio, Kusatsu, Japan). Follow the kit instructions to ensure high-quality RNA extraction.

[0153] cDNA synthesis: Reverse transcribe the extracted total RNA into the corresponding cDNA. Perform the reverse transcription reaction using a reverse transcription kit (e.g., Vazyme, Nanjing, China) according to the manufacturer's recommendations. This step converts the RNA into stable cDNA for subsequent PCR analysis.

[0154] qPCR preparation: Prepare a qPCR reaction to measure the expression level of the target gene. Use a SYBR Green qPCR kit (e.g., Vazyme) and follow the manufacturer's instructions to mix cDNA with appropriate primers and reagents.

[0155] Real-time PCR analysis: Place the qPCR reaction system in a real-time PCR instrument, such as the Bio-Rad Laboratories CFX96 Touch Real-Time PCR Detection System (RRID:SCR_008426). Set up an appropriate PCR program to measure changes in the fluorescence signal, which will reflect the expression level of the target gene. Perform real-time PCR analysis to obtain quantitative gene expression data.

[0156] The primers (primer F (forward) and primer R (reverse)) and probes used are shown in Table B below:

[0157] Table B

[0158] Example 4 Gene selection and clustering

[0159] After obtaining the raw sequencing data, the single-cell sequencing data were aligned and barcode demultiplexed using the Cell Ranger v.3.0.2vdj pipeline (10X Genomics). The expression matrix data were then analyzed under the Scanpy computational framework. The data were filtered according to QC criteria. Cells with <500 detected genes and genes detected in <5 cells / dataset were removed. Cells with mitochondrial gene expression accounting for >10% of the total expression level were also excluded. Double peaks were avoided by removing cells with the top 5% total transcripts (UMIs). Scrublets software was also used to remove double peaks.

[0160] Genes were then selected for dimensionality reduction and clustering based on their variability. All data were merged, and the top 2,000 variable genes were selected. The total UMI count and mitochondrial gene expression ratio per cell were regressed out to eliminate the influence of these factors on clustering. The BBKNN method was used to perform batch correction using the donor as the batch key. Leiden clustering was performed in Scanpy with a resolution of 1.

[0161] Example 5 Model establishment

[0162] The construction of the model of the present invention adopts a basic method that relies on a machine learning algorithm, with thyroid in situ cancer cells as the model input.

[0163] (1) Exploring the feasibility of disease predictors: Using EPCAM and TG dual gene expression as guide genes, all tumor cells were screened from single-cell transcriptome data. After feature screening, a new set of 400 human thyroid carcinoma in situ pathological puncture specimens were used for mRNA extraction. After qPCR experiments, the data were integrated and PCA (principal component analysis) was performed (Figure 1). The results showed that these tumor cells showed spatial differences in PCA analysis, confirming the feasibility of the method.

[0164] (2) Gene sorting: Genes are sorted according to their mutual information scores and the number of genes is limited to a relatively small range.

[0165] (3) Classifier construction and performance evaluation: Random forest model classifiers were constructed based on different numbers of features (genes) and their performance was evaluated. It was observed that when the number of feature genes reached 4, the model performance evaluation showed an inflection point (Figure 2).

[0166] (4) Classifier model of the top three genes: The top three genes with the largest mutual information scores were selected to construct the classifier model. Under this model, the AUROC area reached approximately 0.975 (Figure 3).

[0167] (5) Gene Expression Plot: The expression of these three genes in tumor cells from the metastatic and non-metastatic groups was plotted, and it was found that their expression was differentially expressed based on the Wilcoxon test. These three genes included the three genes in Table A: (A1) RPS4Y1, (A2) PKHD1L1, and (A3) CRABP1. The weight of each gene in the model is shown in Table A.

[0168] Table A

[0169] Example 6 Model Verification

[0170] The model used three specific genes from Table A as input parameters: (A1) RPS4Y1, (A2) PKHD1L1, (A3) CRABP1.

[0171] The model was validated 100 times using a 4:1 split sample set (80% test set, 20% training set), achieving an average AUROC value of approximately 0.975. These results (see Table C for experimental data) demonstrate the robustness of the model and support the robustness and reliability of its conclusions.

[0172] Table C Verification Result Data

[0173] Example 6 Comparative Example

[0174] In step (2) of Example 4, the top four genes were: RPS4Y1, PKHD1L1, CRABP1 and MT1G.

[0175] Classifier models were constructed using the four genes or any combination of three of the four genes. The classification results shown in Table 3 were obtained under each model.

[0176] Table 3

[0177] From the results in Table 3, it can be seen that the marker gene RPS4Y1 is the most important and the three-gene (RPS4Y1, PKHD1L1 and CRABP1) classifier model of the present invention performs best in the test AUROC, that is, it has a very good detection effect.

[0178] Example 7

[0179] Test organization preparation and data collection

[0180] In order to perform transcriptional analysis, the present invention selected 157 patients from Center A, including 107 patients with lymph node metastasis (49 cases of cervical metastasis and 58 cases of central metastasis) and 50 patients without metastasis. The primary tumor tissue and its paired surrounding tissue (greater than 0.5 cm from the tumor edge) were surgically removed and transferred to liquid nitrogen for storage. All patients came to the hospital for the first time and did not receive any anti-cancer treatment before surgery. In addition to collecting paired specimens, detailed clinical information was also collected, including gender, age, tumor size, degree of differentiation, lymph node metastasis, TNM stage and other relevant factors. The acquisition of all patient samples was approved by the hospital ethics committee, and all participants signed informed consent.

[0181] The inclusion criteria for patient selection were: (1) first-time thyroid surgery; (2) postoperative pathological confirmation of papillary thyroid carcinoma; and (3) no history of other malignant tumors.

[0182] Participants used to develop the prediction model were extracted from the electronic health record database, and a total of 458 patients were selected for the development of the prediction model (Center B dataset). In the clinical validation, 366 participants were recruited from two centers (Center A and C datasets). Of these, 349 participants underwent qRT-PCR testing to assess RNA expression. Due to insufficient histopathological examination information, 17 participants were excluded from further analysis; an additional 10 participants were also excluded due to insufficient RNA concentration. Therefore, a total of 339 participants were included in the final analysis. These 339 participants were categorized by metastatic status, with 185 participants in the non-metastatic group and 154 participants in the metastatic group. In the metastatic group, 24 participants were further analyzed for metastatic location. Of these, 17 participants were classified as having central metastases and 7 as having cervical metastases. However, due to missing metastatic location information, 130 participants were not included in the metastatic location analysis.

[0183] The algorithm development and clinical trial process for predicting thyroid cancer metastasis are shown in Figure 4. The distribution of ΔCT values ​​of three genes (RPS4Y1, PKHD1L1, and CRABP1) in the metastatic and non-metastatic groups in the Nantong (center B), Zhongshan (center A), and Xuzhou (center C) datasets is shown in Figure 5.

[0184] Development and evaluation of algorithmic methods combined with AI

[0185] To predict thyroid cancer metastasis, five machine learning algorithms were employed: random forest (RF), XGBoost, support vector machine (SVM), deep learning (DL), and logistic regression. These models were selected for their excellent performance in classification tasks and their ability to handle complex relationships in the dataset. The development process began by splitting the dataset into training and validation sets. For each model, hyperparameters were tuned using 5-fold cross-validation to optimize performance and prevent overfitting. The model was trained using the training data, and its performance was subsequently evaluated on the validation data. After hyperparameter tuning, model performance was further evaluated using clinical data, which served as the test set.

[0186] Model evaluation uses several key metrics, including accuracy, precision, recall, F1 score, and area under the receiver operating characteristic curve (AUROC). Accuracy measures the overall correctness of the model and calculates the proportion of correctly classified instances. Precision represents the proportion of true positives among all samples predicted to be positive, reflecting the model's ability to avoid false positives. Recall measures the proportion of true positive cases correctly identified, indicating the model's sensitivity. The F1 score is the harmonic mean of precision and recall, balancing the trade-off between the two. AUROC reflects the model's ability to distinguish different cases at all classification thresholds, providing a comprehensive performance assessment.

[0187] These metrics were used to compare each machine learning model on each independent dataset to assess its generalization ability and robustness in predicting thyroid cancer metastasis. The best-performing model was identified based on the highest score in these evaluation metrics, particularly the AUROC value.

[0188] Experimental results

[0189] The performance evaluation results of five machine learning models—random forest (RF), XGBoost, support vector machine (SVM), deep learning (DL), and logistic regression—for predicting thyroid cancer metastasis on three different datasets (center B, center A, and center C) are shown in Tables 4-6 and Figures 6-9. Evaluation metrics include accuracy, precision, recall, F1 score, and area under the receiver operating characteristic (ROC) curve (AUROC).

[0190] Table 4 Test results of each model on the data of center B

[0191] Table 5 Test results of each model on the data of center A

[0192] Table 6 Test results of each model on the data of center C

[0193] As shown in Tables 4-6, the random forest model outperformed the other models on all datasets. In the Center B dataset, random forest achieved the highest accuracy (0.949), precision (0.896), recall (0.976), F1 score (0.934), and AUROC value (0.992). Similarly, random forest performed well in the Center A and Center C datasets, with accuracy of 0.894 and 0.886, and AUROC values ​​of 0.911 and 0.953, respectively. In contrast, logistic regression and deep learning models performed poorly overall across all three datasets, particularly in terms of recall and AUROC.

[0194] Figure 6 shows the receiver operating characteristic (ROC) curves of the random forest model for predicting metastasis in the three datasets. The area under the curve (AUC) for the Center B dataset was the highest, reaching 0.992, followed by Center C (0.953) and Center A (0.911). The high AUC values ​​across all datasets indicate that this model (the AI ​​algorithm model based on the combination of the three genes) has a strong ability to distinguish between metastatic and non-metastatic cases.

[0195] Figure 7 shows the matrix analysis of the random forest model on the dataset of center B. The results show that the model correctly classified 271 non-metastatic pathologies and 164 metastatic cases, with only a small number of misclassifications (19 false positives and 4 false negatives).

[0196] Figure 8 shows the matrix analysis of the random forest model on the dataset of center A. The results show that the model correctly classified 17 true negative and 34 true positive cases, with only a small number of misclassification errors (3 false positives and 3 false negatives).

[0197] Figure 9 shows the matrix analysis of the random forest model on the dataset of center C. The results show that the model correctly classified 151 non-metastatic cases and 99 metastatic cases, with 14 false positives and 18 false negatives.

[0198] The above results show that the random forest model of the present invention has high accuracy in different clinical settings under the three-gene index.

[0199] Example 8 Human RPS4Y1, PKHD1L1, CRABP1 Gene Detection Kit (qPCR-Fluorescent Probe Method)

[0200] This embodiment provides a gene detection kit developed by the inventors and prepared based on the three gene markers of the present invention. This kit is based on a real-time fluorescence PCR platform.

[0201] This kit is used to qualitatively determine the expression of RPS4Y1, PKHD1L1, CRABP1, and the internal reference gene GAPDH in formalin-fixed, paraffin-embedded (FFPE) tumor tissue specimens from patients with papillary thyroid carcinoma (PTC).

[0202] This kit is equipped with an internal reference GAPDH gene, whose amplification status can indicate the quality of the added DNA template and monitor the DNA amplification status in each reaction well. dUTP and UNG enzyme are also added to prevent contamination.

[0203] This kit establishes a random forest model and uses the amplification of RPS4Y1, PKHD1L1, and CRABP1 genes to predict the metastasis of thyroid tumors.

[0204] The main components of the kit are shown in Table 7 below.

[0205] Table 7

[0206] Detection method

[0207] Preparation of PCR pre-reaction solution

[0208] According to the reaction sample volume, place the qPCR reaction solution and primer probe mixture in the kit on ice to melt. After complete melting, gently shake to mix, and centrifuge at low speed for 10 seconds.

[0209] Each qPCR reaction requires 15 μL of qPCR reaction buffer, 1 μL of enzyme mix, and 4 μL of primer-probe mix. Add the appropriate proportions of these reagents to a centrifuge tube based on the number of samples. Vortex the qPCR pre-reaction mixture and briefly centrifuge to remove any residual liquid from the tube walls.

[0210] Preparation of PCR reaction plates

[0211] Add 20 μl of qPCR pre-reaction solution to the wells of the selected 8-tube strips. Then add 5 μl of extraction template to the corresponding wells of the PCR tubes.

[0212] Seal the tube with a cap and centrifuge at low speed until the mixture flows to the bottom of the tube and no bubbles appear. The sealed PCR tube can be placed at 2-8°C for no more than 2 hours.

[0213] The reaction system of each reaction during PCR detection is shown in Table 8.

[0214] Table 8

[0215] PCR instrument

[0216] Applied Biosystems 7500 PCR instrument

[0217] 1) qPCR reaction plate loading

[0218] The PCR pre-reaction solution does not contain ROX or other dyes, so select "None" for reference fluorescence. Select the FAM fluorescence channel for RPS4Y1, the VIC fluorescence channel for PKHD1L1, the ROX fluorescence channel for CRABP1, and the Cy5 channel for GAPDH. Set up the reaction program as shown in Table 2:

[0219] Table 9: Reaction Procedure

[0220] 2) Analysis condition settings

[0221] Analyze the run results. In most cases, the baseline start and end cycle numbers automatically set by the instrument can be used. If baseline problems occur, manual adjustments must be made on a channel-by-channel basis. (For example, if the PKHD1L1 target baseline in the VIC channel is not flat, select Analysis Setting > CT Setting, select the VIC channel, uncheck Automatic Baseline, and make changes to Baseline Start Cycle and End Cycle, such as changing Start Cycle from 3 to 2. Return to the instrument analysis interface, click Analyze in the upper-right corner, and reanalyze to see if the VIC channel baseline is flat.)

[0222] The threshold lines for RPS4Y1, PKHD1L1, CRABP1, and GAPDH were set to 25,000 (fine-tuning can be performed depending on the specific situation).

[0223] Macrostone PCR analysis system SLAN-96P is installed

[0224] 1) qPCR reaction plate loading

[0225] The PCR pre-reaction solution does not contain ROX or other dyes. Select the FAM fluorescence channel for RPS4Y1, the VIC fluorescence channel for PKHD1L1, the ROX fluorescence channel for CRABP1, and the Cy5 channel for GAPDH. Set up the reaction program as shown in Table 2.

[0226] 2) Analysis condition settings

[0227] Analyze the run results. In most cases, the baseline starting point and end cycle number automatically set by the instrument can be used. When baseline problems occur, manual adjustment of each channel is required.

[0228] The threshold lines for RPS4Y1, PKHD1L1, CRABP1, and GAPDH were set to 0.12 (fine-tuning can be performed depending on the specific situation).

[0229] Applied Biosystems Q5 PCR instrument

[0230] 1) qPCR reaction plate loading

[0231] The PCR pre-reaction solution does not contain ROX or other dyes, so select "None" for the reference fluorescence. Select the FAM fluorescence channel for RPS4Y1, the VIC fluorescence channel for PKHD1L1, the ROX fluorescence channel for CRABP1, and the Cy5 channel for GAPDH. Set up the reaction program as shown in Table 2.

[0232] 2) Analysis condition settings

[0233] Analyze the run results. In most cases, the baseline starting point and end cycle number automatically set by the instrument can be used. When baseline problems occur, manual adjustment of each channel is required.

[0234] Set the threshold lines of RPS4Y1, PKHD1L1, CRABP1 and GAPDH to 15,000 (fine-tuning can be performed depending on the specific situation).

[0235] Test results

[0236] The ΔCT values ​​of RPS4Y1, PKHD1L1, CRABP1 targets and GAPDH of each sample were detected.

[0237] The ΔCt value of each marker was calculated as follows:

[0238] ΔCt value = Ct value of the test gene - Ct value of the internal reference. The Ct value of the test gene refers to the Ct value corresponding to the gene signal detected in the sample; the internal reference Ct value refers to the Ct value of the internal reference signal corresponding to the sample.

[0239] Calculation based on the random forest model of the test results

[0240] Random forest model calculation:

[0241] Random forest is composed of multiple decision trees, each decision tree (f t ) Using the ΔCt value (X), output the prediction result The ΔCt value (X = [x1, x2, x3]) consists of the ΔCt values ​​of RPS4Y1, PKHD1L1, and CRABP1 genes.

[0242] The final prediction of the random forest is obtained by the above formula, where T is the number of decision trees. In the present invention, T = 41. If the average positive probability output by all decision trees is greater than 0.5, the prediction is transfer, otherwise it is predicted not to transfer.

[0243] Detection effect of the kit

[0244] It is stipulated that the thyroid cancer metastasis calculation software provided with this kit should be used to input the ΔCt values ​​of RPS4Y1, PKHD1L1, and CRABP1 genes to obtain the positive probability.

[0245] Positive result: If the positive probability is greater than 0.5, the sample is judged to be positive.

[0246] Negative result: If the positive probability is less than or equal to 0.5, the sample is judged to be negative.

[0247] Invalid result: If the average Ct value of GAPDH is greater than 32.0, the result is invalid. Re-examination is recommended for invalid results.

[0248] The above kit was used to measure 355 samples of papillary thyroid carcinoma, and the results were analyzed using a receiver operating characteristic (ROC) curve. The analysis results are shown in FIG10 .

[0249] When the positive probability threshold was 0.5, the sensitivity, specificity and AUC value of papillary thyroid carcinoma metastasis were 0.843, 0.952 and 0.926 respectively.

[0250] Example 9 Kit Performance Test

[0251] The data of the kit in Example 8, such as specimen type, intra-laboratory precision, inter-batch precision, detection limit, anti-interference performance, stability, cross-contamination, clinical or diagnostic sensitivity and specificity, and clinical reproducibility, are shown below.

[0252] 9.1 Specimen Type

[0253] The clinical samples of the thyroid cancer metastasis detection kit are mainly thyroid tissue sections.

[0254] Standard formalin-fixation and paraffin-embedding procedures often result in nucleic acid fragmentation. To minimize the possibility of RNA fragmentation, sample processing should follow the following steps:

[0255] Tissue should be immersed in 4%-10% formalin solution as soon as possible after excision;

[0256] The best fixation time is 14-24 hours. Longer fixation time will lead to more severe RNA fragmentation, which is not conducive to downstream experiments.

[0257] Specimens must be thoroughly dehydrated before coating.

[0258] Fresh FFPE tissue sections should be used, with a thickness of no more than 10 μm. Thick sections may result in low RNA yields. The number of sections used for each preparation should not exceed 8, and the surface area should not exceed 250 mm. 2 .

[0259] If there is no starting sample information, it is recommended that the number of slices used for the initial preparation should not exceed 2 slices. The number of slices used for the next preparation can be adjusted based on the RNA yield and purity, but should not exceed 8 slices.

[0260] In addition, fresh thyroid tissue can also be used as clinical samples for this kit.

[0261] 9.2 In-laboratory precision

[0262] To determine the repeatability of thyroid cancer metastasis detection at different times, this study mainly analyzed the intra-laboratory precision and calculated the CV value within each group.

[0263] Thyroid metastasis detection kit batches: X4410010, X4411020, X4412030;

[0264] Instruments: ABI 7500, SLAN, ABIQ5.

[0265] The CV value of intra-laboratory precision must be ≤10%.

[0266] Methods: The precision of the results was measured in the same laboratory by the same operator (group of operators) on the same instrument, using the same method and the same type and batch of reagents, for the same test samples (high, medium and low precision reference materials) over a period of one month.

[0267] The specific design is as follows: the laboratory uses one kit batch to evaluate the precision of samples at multiple concentration levels, running two analytical batches per day (morning and afternoon), with two samples of each concentration processed in parallel for 20 days (non-consecutive days). This is 20 × 2 × 2, ultimately obtaining 80 data points for each concentration.

[0268] Results: The CV values ​​of 80 CT values ​​obtained from the 20-day test of three batches were all ≤10%.

[0269] 9.3 Inter-batch precision

[0270] In order to determine the reproducibility of thyroid cancer metastasis detection between different batches, this study mainly analyzed three batches and calculated the CV value within each group.

[0271] Thyroid metastasis detection kit batches: X4410010, X4411020, X4412030;

[0272] Instruments: ABI 7500, SLAN, ABIQ5.

[0273] The CV value of inter-batch precision must be ≤10%.

[0274] Results: The CV values ​​of the three batches of reagents in laboratory A within 5 days were all ≤10%.

[0275] 9.4 Detection Limit

[0276] Since the determination method of this kit is ΔCT value, the conventional clinical detection limit is not applicable. In order to determine the detection limit of thyroid cancer metastasis detection, the present invention mainly verifies the detection limit of the kit from the perspective of reagent sensitivity.

[0277] Thyroid metastasis detection kit batches: X4410010, X4411020, X4412030;

[0278] Instruments: ABI 7500, SLAN, ABIQ5.

[0279] The detection limit of each target of the reagent is ≥1 copies / μL. The detection limit is determined by the following method:

[0280] The experiments were performed using pseudovirus templates: lenti RPS4Y1 pseudovirus product, lenti PKHD1L1 pseudovirus product, and lenti CRABP1 pseudovirus product.

[0281] Preliminary detection limit testing: Using the pseudovirus calibration data, dilute the pseudoviruses for the three targets using Nucleic Acid Releaser V2. Each target was diluted to four concentrations: 2 copies / μL, 1 copy / μL, 0.5 copies / μL, and 0.25 copies / μL. Detection of these templates was tested using the first batch of reagents. Three replicates were tested for each template, and the detection rate was calculated.

[0282] The preliminary test results of the detection limit are shown in Table 9.

[0283] Table 9

[0284] The lowest template concentration with a detection rate of 100% was selected as the initial test reagent sensitivity for subsequent experiments, i.e., 0.5 copies / μL.

[0285] In-depth detection limit testing: Pseudoviruses for three targets were diluted using Nucleic Acid Releaser V2 according to the pseudovirus calibration data. Each target was diluted to four concentrations: 1 copy / μL, 0.5 copies / μL, and 0.25 copies / μL. Detection of the above templates was tested using two batches of reagents. Twenty replicates were performed for each template, and the detection rate was calculated.

[0286] The results of the in-depth detection limit test are shown in Table 10.

[0287] Table 10

[0288] The lowest template concentration with the above detection rate of 95% was selected as the sensitivity of the preliminary test reagent for subsequent experiments, that is, 1 copies / μL.

[0289] Detection limit confirmation testing: Using calibration data, three target pseudoviruses were diluted using Nucleic Acid Releaser V2. Each target was diluted to four concentrations: 2 copies / μL, 1.5 copies / μL, 1 copy / μL, and 0.5 copies / μL. Detection of the above templates was tested using three batches of reagents. Five replicates were tested for each template, and the detection rate was calculated.

[0290] The detection limit confirmation test results are shown in Table 11 below.

[0291] Table 11

[0292] In summary, the detection limits of the three targets RPS4Y1, PKHD1L1 and CRABP1 were all 1 copies / μL.

[0293] Three targets were tested and confirmed at 1 copy / μL using three batches of reagents, using the ABI 7500, SLAN, and ABIQ5 instruments, respectively. The results are shown in Table 12.

[0294] Table 12

[0295] Results: 20 replicates of the three targets, RPS4Y1, PKHD1L1, and CRABP1, were tested with a detection rate of ≥95% at 1 copy / μL. Therefore, the sensitivity of this kit can reach 1 copy / μL.

[0296] 9.5 Anti-interference

[0297] 9.5.1 Anti-endogenous interference

[0298] 2mg / mL hemoglobin, 37mM / L triglycerides, and 60mg / mL albumin were used to simulate endogenous interfering substances to test whether this kit can accurately detect in the presence of endogenous interfering substances.

[0299] Thyroid metastasis detection kit batches: X4410010, X4411020, X4412030;

[0300] Instruments: ABI 7500, SLAN, ABIQ5.

[0301] It is required that the CT value of the sample diluted with interfering substances after extraction and testing should be within 0.5 compared with the control.

[0302] Methods: 2 mg / mL hemoglobin, 37 mM / L triglyceride and 60 mg / mL albumin were used as endogenous interfering substances. The above three substances were mixed according to their concentrations to form an endogenous interfering substance, named GRW-1.

[0303] T7 RNA templates were diluted to high, medium, and low concentrations using GRW-1 and Nucleic Acid Releaser V2. Total RNA was extracted using a paraffin-embedded tissue section total RNA extraction kit and assayed using this kit. Each concentration was assayed in triplicate, and the CT values ​​for the three targets, RPS4Y1, PKHD1L1, and CRABP1, were calculated.

[0304] Results: 2 mg / mL hemoglobin, 37 mM / L triglycerides, and 60 mg / mL albumin were used as endogenous interfering substances. These three substances were mixed at different concentrations to form endogenous interfering substances. RNA was extracted from paraffin-embedded tissue sections using a total RNA extraction kit, and the extracted product was tested using this kit. Compared to the control without interfering substances, the difference in ΔCT for the high, medium, and low templates was ≤0.5. For example, the results of the SLAN endogenous interfering substance test for batch 1 are shown in Table 13.

[0305] It can be seen that the kit of the present invention has very good resistance to endogenous interference.

[0306] Table 13

[0307] 9.5.2 Anti-exogenous interference

[0308] 21.7mmol / L ethanol, equal volume of paraffin, equal volume of formalin (4%-10%) and 35mmol / L xylene were used as exogenous interfering substances to test whether the kit can accurately detect in the presence of exogenous interfering substances.

[0309] Thyroid metastasis detection kit batches: X4410010, X4411020, X4412030;

[0310] Instruments: ABI 7500, SLAN, ABIQ5.

[0311] It is required that the CT value of the sample diluted with interfering substances after extraction and testing should be within 0.5 compared with the control.

[0312] Methods: 21.7mmol / L ethanol, equal volume of paraffin, equal volume of formalin (4%-10%) and 35mmol / L xylene were used as exogenous interfering substances. The above four substances were mixed according to their concentrations to form an exogenous interfering substance, named GRW-2.

[0313] T7 RNA template was diluted to high, medium, and low concentrations using GRW-2 and Nucleic Acid Releaser V2, respectively. Total RNA was extracted from paraffin-embedded tissue sections using a total RNA extraction kit, and the extracted product was tested using this kit.

[0314] Results: 21.7 mmol / L ethanol, an equal volume of paraffin, an equal volume of formalin (4%-10%), and 35 mmol / L xylene were used as exogenous interfering substances. Total RNA was extracted from paraffin-embedded tissue sections using a total RNA extraction kit. The extracted product was then tested using this kit. Compared to the control without interfering substances, the difference in ΔCT for the high, medium, and low templates was ≤0.5. For example, the results of the SLAN exogenous interfering substance test for batch 1 are shown in Table 14.

[0315] It can be seen that the kit of the present invention has very good resistance to endogenous interference.

[0316] Table 14

[0317] 9.6 Stability

[0318] The purpose of the study is to ensure the stability of all components of the kit (including positive quality control products and negative quality control products) during the use period.

[0319] Thyroid metastasis detection kit batches: X4410010, X4411020, X4412030;

[0320] Instrument: ABI 7500.

[0321] The kit is required to be stored for a certain period of time under relevant storage conditions, and its linear R 2 ≥0.99, 90%≤ amplification efficiency≤110%; the accuracy of its positive and negative reference material detection is 100%; its precision detection CV value is ≤10%; its reagent detection limit is at least 1 copy / μL.

[0322] Methods: Three batches of the original kit were stored at 2-8°C. The initial time of exposure to these conditions was set as zero. Samples were collected on days 0, 3, and 7 from the date of exposure, and the reaction solutions were prepared according to the manufacturer's instructions. Nucleic acid was extracted from the manufacturer's reference material using the Tiangen Paraffin-Embedded Tissue Section Total RNA Extraction Kit according to the manufacturer's instructions. Five replicates were added to each reference material for nucleic acid testing. After template addition, amplification was performed on an ABI 7500 fluorescence quantitative PCR instrument. A threshold of 25,000 was set for each target, and the performance of the three batches of reagents during post-opening use was analyzed for linearity, limit of detection, accuracy, and precision.

[0323] Results: The three batches of reagents were stored at 2-8°C (4°C) in the dark for 7 days to test the linearity. The correlation coefficients of R for each target of RPS4Y1, PKHD1L1, and CRABP1 were 2 The kit passed the stability linearity test with a 90% amplification efficiency of ≤0.99 and ≤110% amplification efficiency. The detection rate for each target, RPS4Y1, PKHD1L1, and CRABP1, against the detection limit reference standard (UL) reached over 95% in 20 wells. The kit passed the stability detection limit test. The ΔCt values ​​for each target, RPS4Y1, PKHD1L1, and CRABP1, compared to the internal standard, fell within the negative and positive interpretation ranges, with a 100% negative / positive agreement rate. The kit passed the stability accuracy test. The coefficient of variation (CV) for each target, measured 10 times against the precision reference standard, was ≤5% (Table 15). The kit passed the stability precision test.

[0324] Table 15

[0325] 9.7 Cross-contamination

[0326] To test whether cross contamination occurs during the extraction and loading process of the thyroid cancer metastasis detection kit.

[0327] Thyroid metastasis detection kit batches: X4410010, X4411020, X4412030;

[0328] Instrument: ABI 7500.

[0329] The test kit is required to test positive samples as positive and negative samples as negative, that is, the positive-negative consistency rate is 100%.

[0330] Methods: 20 positive and 20 negative samples were mixed and placed crosswise. Total RNA was extracted from paraffin-embedded tissue sections using the Tiangen Total RNA Extraction Kit according to the manufacturer's instructions. After extraction, RNA was detected using the kit and the results were statistically analyzed.

[0331] Results: The results are shown in Table 16. The test values ​​of 20 positive samples and 20 negative samples were consistent with their positive and negative characteristics, and there was no cross contamination.

[0332] Table 16

[0333] 9.8 Clinical Diagnostic Specificity

[0334] Clinical samples were used to validate the specificity of the thyroid cancer metastasis detection kit.

[0335] Thyroid metastasis detection kit batch number: X4410010;

[0336] Instrument: ABI 7500.

[0337] Tested on the following confirmed negative clinical specimens:

[0338] (1) 150 thyroid tissue sections from patients without thyroid cancer;

[0339] (2) 150 thyroid tissue sections from patients with thyroid cancer but without metastasis;

[0340] (3) 50 samples of thyroid tissue from patients without thyroid cancer;

[0341] (4) Thyroid tissues from 50 patients with thyroid cancer but without metastasis.

[0342] Specimens used for specific detection were obtained from Zhongshan Hospital Affiliated to Fudan University.

[0343] Requirements: Specificity greater than or equal to 95%.

[0344] Results: The results of this study are shown in Table 17.

[0345] Table 17

[0346] The test results in Table 17 show that no metastasis was detected in any of the samples. Therefore, the specificity of the kit of the present invention for thyroid tissue sections or fresh tissues of healthy subjects is greater than 95%.

[0347] 9.9 Clinical Diagnostic Sensitivity

[0348] For thyroid tissue sections or fresh tissue from healthy individuals, the specificity is greater than 95%.

[0349] Thyroid metastasis detection kit batch number: X4410010;

[0350] Instrument: ABI 7500.

[0351] Tested on the following clinical samples with confirmed positive / negative status:

[0352] (1) 150 thyroid tissue sections from patients with thyroid cancer and metastasis;

[0353] (2) 100 cases of thyroid tissue from patients with thyroid cancer and metastasis.

[0354] Specimens used for specific detection were obtained from Zhongshan Hospital Affiliated to Fudan University.

[0355] Requirements: Detection rate greater than or equal to 95%.

[0356] Results: The results of this study are shown in Table 18.

[0357] Table 18

[0358] The test results in Table 18 show that metastasis was detected in all samples. Therefore, the detection rate of the kit of the present invention for thyroid tissue sections or fresh tissues of patients is greater than 95%.

[0359] 9.10 Clinical Reproducibility

[0360] Clinical samples were used to validate the reproducibility of the thyroid cancer metastasis detection kit.

[0361] Thyroid metastasis detection kit batch number: X4410010;

[0362] Instrument: ABI 7500.

[0363] Methods: Serial sections of 20 known samples were prepared by the central laboratory and shipped to four other qPCR laboratories. Samples were shipped within 7 days and stored at -20°C upon receipt. Extraction and testing were performed within one week. Each qPCR laboratory used the same batch of this kit for the experiments, and the consistency of the results from the five qPCR laboratories was compared.

[0364] Requirements: Accuracy greater than or equal to 95%.

[0365] Results: The experimental results are shown in Table 19.

[0366] Table 19

[0367] The test results in Table 19 show that all five locations can correctly identify positive samples. Therefore, the kit of the present invention can accurately identify thyroid tissue slice samples from patients at all five locations.

[0368] discuss

[0369] In the present invention, an evaluation method is introduced for analyzing the potential risk of thyroid cancer metastasis, and an analysis device capable of predicting the risk of thyroid cancer metastasis is further designed.

[0370] In the present invention, the inventors unexpectedly identified a new set of biomarkers for thyroid cancer metastasis risk, including the following: (A1) RPS4Y1, (A2) PKHD1L1 and (A3) CRABP1.

[0371] In the implementation of the present invention, by selecting the biomarkers in the above-mentioned thyroid carcinoma in situ cells, an objective assessment of the risk of thyroid cancer metastasis is achieved, showing high sensitivity and specificity in the diagnosis of thyroid cancer metastasis.

[0372] Specifically, the present invention used single-cell sequencing for feature screening, and qPCR for retraining and feature contribution ranking. The three genes with the highest mutual information scores (RPS4Y1, PKHD1L1, and CRABP1) were selected to construct a model, which was then validated. This three-gene classification model demonstrated excellent classification and detection performance, with higher detection accuracy than other three-gene combinations, and significantly lower detection costs than combinations with more genes (>3).

[0373] Thyroid cancer is one of the most common endocrine malignancies, and its incidence has increased significantly worldwide in recent years [1,2]. Papillary thyroid carcinoma (PTC) accounts for approximately 90% of all thyroid cancers and is a well-differentiated type of cancer. Although PTC has an overall good prognosis, it is often accompanied by lymph node metastasis (LNM), which complicates surgical management and increases the risk of recurrence. The incidence of central lymph node metastasis in PTC is reported to be 30% to 65%, while the incidence of lateral lymph node metastasis is approximately 20% to 30%, but may be as high as 50% in some more aggressive lesions.

[0374] Preoperative evaluation of LNM primarily relies on neck ultrasound. However, ultrasound has limited diagnostic accuracy, particularly in detecting central cervical lymph node metastases. Studies have shown that central and lateral metastases were misreported or missed in 78.6% and 42.3% of cases, respectively, leading to a change in surgical plan in 65.4% of patients based on intraoperative ultrasound performed by the surgeon. This low accuracy often results in unnecessary fine needle aspiration (FNA) and prophylactic lymph node dissection (LND) in patients. Therefore, there is an urgent need for a non-invasive, accurate, and efficient method to predict thyroid LNM preoperatively.

[0375] This invention aims to combine genomic data with advanced artificial intelligence (AI) algorithms to construct a LNM prediction model based on the gene expression profile of primary thyroid tumors. This approach may significantly improve the accuracy of preoperative assessment of LNM risk, thereby optimizing surgical planning, reducing unnecessary lymph node dissections, and lowering the likelihood of recurrence and secondary surgery, ultimately improving overall patient prognosis.

[0376] Predicting LNM in PTC is crucial for effective patient management, particularly in cases where ultrasound detection presents challenges in certain anatomical regions. Detecting central LNM by ultrasound is particularly challenging due to anatomical limitations and the lack of typical LNM features, such as microcalcifications and vascular proliferation. To address these limitations, machine learning (ML) and radiomics have shown considerable promise. For example, Liu et al. applied radiomics to predict LNM by extracting 614 features from B-ultrasound images and developed a support vector machine (SVM) model with an AUC of 0.78 on the training set and 0.73 on the test set. Similarly, Chang et al. developed a nomogram integrating deep learning, clinical features, and ultrasound characteristics, achieving AUC values ​​ranging from 0.809 to 0.829 in a validation cohort. A recent meta-analysis reported a pooled AUC of 0.84 for radiomics-based models, while an AUC of 0.81 for models combining radiomics with clinical data, indicating that incorporating clinical data generally does not significantly improve diagnostic performance. However, in a multicenter and cross-device setting, the performance of clinical and traditional radiomics models decreased significantly. Independent testing showed that the AUCs of the clinical-informed model and the traditional radiomics model were 0.67 and 0.56, respectively. These findings emphasize the need for more robust and reproducible diagnostic tools for LNM diagnosis.

[0377] Gene expression profiling has emerged as a powerful approach for diagnosing LNM, offering greater reproducibility by reducing reliance on imaging quality. For example, Cerruti et al. identified differentially expressed genes using serial analysis of gene expression (SAGE), achieving 80% diagnostic accuracy in detecting LNM. Another study using single-cell RNA sequencing developed an 11-gene signature and incorporated it into a nomogram for predicting LNM in patients with PTC, with AUC values ​​ranging from 0.812 to 0.864 across different datasets. Furthermore, a prediction model integrating clinical features such as age, sex, and tumor diameter with genetic markers such as RET fusion reported an AUC of 0.724, further highlighting the prognostic significance of BRAFV600E and RET mutations as indicators of metastasis. Compared to these previous studies, our three-gene expression model offers a more efficient and cost-effective approach. Its simplified application enhances accuracy, precision, and AUC values, making it a superior tool for diagnosing LNM in clinical practice.

[0378] The three-gene expression model of the present invention shows extremely high accuracy and stability in predicting lymph node metastasis (LNM) of papillary thyroid carcinoma (PTC). Compared with other methods, such as models based on radiomics or traditional clinical features, the model of the present invention has always maintained a high AUC value (0.91 and 0.95, respectively) in external validation of multiple clinical centers, which is significantly better than the previously reported model. In addition, traditional models often show inconsistent results when applied in multiple centers and across devices, while the model of the present invention is less affected by external factors because it is based on gene expression profiles, and has better cross-scenario applicability and repeatability. Therefore, the model of the present invention not only provides higher diagnostic accuracy, but also demonstrates its stability in different clinical settings, becoming a reliable tool for LNM diagnosis.

[0379] All documents mentioned in this application are incorporated herein by reference, just as if each document were incorporated herein by reference individually. It should also be understood that after reading the above teachings of the present invention, those skilled in the art may make various changes or modifications to the present invention, and that these equivalents also fall within the scope defined in the appended documents.

Claims

1. A kit, characterized in that: The kit contains a detection reagent, which is used to detect mRNA, cDNA, or a combination thereof of a thyroid cancer metastasis risk marker combination in a sample to be tested; Wherein, the combination of thyroid cancer metastasis risk markers is: (A1) RPS4Y1; (A2) PKHD1L1; and (A3) CRABP1; Wherein, the detection reagent includes a primer pair, and the primer pair includes: a marker primer pair for specifically amplifying the mRNA or cDNA of the thyroid cancer metastasis risk marker combination.

2. The kit according to claim 1, characterized in that The kit also includes: an internal reference primer pair for amplifying an internal reference gene.

3. The kit according to claim 2, characterized in that The marker primer pair comprises: (P1) Primer pair for amplifying RPS4Y1: SEQ ID NO: 1 and SEQ ID NO: 2; (P2) Primer pair for amplifying PKHD1L1: SEQ ID NO: 3 and SEQ ID NO: 4; (P3) a primer pair for amplifying CRABP1: SEQ ID NO: 5 and SEQ ID NO: 6; and / or The internal reference primer pair is SEQ ID NO:7 and SEQ ID NO:

8.

4. The kit according to claim 3, characterized in that The detection reagent also includes a probe used in conjunction with the primer pair. The probes include: probe 1: SEQ ID NO: 9 used in combination with a primer pair for amplifying RPS4Y1; probe 2: SEQ ID NO: 10 used in combination with a primer pair for amplifying PKHD1L1; probe 3: SEQ ID NO: 11 used in combination with a primer pair for amplifying CRABP1; and probe 4: SEQ ID NO: 12 used in combination with a primer pair for amplifying an internal reference gene.

5. The kit according to claim 1, characterized in that The kit also includes a label or instructions, which indicates that the kit is used for (a) determining the risk of thyroid cancer metastasis, and / or (b) evaluating the therapeutic effect of thyroid cancer metastasis.

6. Use of mRNA, cDNA, or a detection reagent thereof for a combination of thyroid cancer metastasis risk markers, characterized in that: Used for preparing a diagnostic reagent or a kit, wherein the diagnostic reagent or the kit is used for judging whether a thyroid cancer patient has thyroid cancer metastasis; Wherein, the combination of thyroid cancer metastasis risk markers is: (A1) RPS4Y1; (A2) PKHD1L1; and (A3) CRABP1; Wherein, the detection reagent comprises a primer pair, and the primer pair comprises: for specifically amplifying the The present invention relates to a marker primer pair for mRNA or cDNA of a combination of prostate cancer metastasis risk markers.

7. The use according to claim 6, characterized in that The kit also includes: an internal reference primer pair for amplifying an internal reference gene.

8. The use according to claim 7, characterized in that The marker primer pair comprises: (P1) Primer pair for amplifying RPS4Y1: SEQ ID NO: 1 and SEQ ID NO: 2; (P2) Primer pair for amplifying PKHD1L1: SEQ ID NO: 3 and SEQ ID NO: 4; (P3) a primer pair for amplifying CRABP1: SEQ ID NO: 5 and SEQ ID NO: 6; and / or The internal reference primer pair is SEQ ID NO:7 and SEQ ID NO:

8.

9. The use according to claim 7, characterized in that The detection reagent also includes a probe used in conjunction with the primer pair. The probes include: probe 1: SEQ ID NO: 9 used in combination with a primer pair for amplifying RPS4Y1; probe 2: SEQ ID NO: 10 used in combination with a primer pair for amplifying PKHD1L1; probe 3: SEQ ID NO: 11 used in combination with a primer pair for amplifying CRABP1; and probe 4: SEQ ID NO: 12 used in combination with a primer pair for amplifying an internal reference gene.

10. The use according to claim 6, characterized in that The diagnostic reagent or kit is used to detect thyroid cancer metastasis risk markers in thyroid proto-cancer tissue.

11. The use according to claim 6, characterized in that The thyroid cancer includes papillary thyroid carcinoma.

12. The use according to claim 1, characterized in that The diagnostic reagent or kit obtains an assessment result of the risk of thyroid cancer metastasis by detecting the difference ΔCt between the Ct value of each marker in qPCR amplification and the Ct value of the internal reference gene in thyroid protocarcinoma tissue.

13. A device for early screening of thyroid cancer metastasis, characterized in that: The device comprises: (a) an input module, wherein the input module is used to input expression data of a characteristic gene combination of a thyroid cancer patient; Wherein, the characteristic gene combination includes: (A1) RPS4Y1; (A2) PKHD1L1; and (A3) CRABP1; The expression data is the ΔCt value of each characteristic gene obtained by PCR amplification using the primer pair, and the ΔCt value is the difference between the Ct value of the characteristic gene and the Ct value of the internal reference signal; Wherein, the primer pair includes: a marker primer pair for specifically amplifying the mRNA or cDNA of the thyroid cancer metastasis risk marker combination; and an internal reference primer pair for amplifying an internal reference gene; (b) a processing module or a risk assessment module, wherein the module is configured to: analyze and calculate the input characteristic genes to obtain a risk assessment; wherein when (A1) RPS4Y1, (A2) PKHD1L1 and (A3) CRABP1 are significantly upregulated, it indicates that the risk of thyroid cancer metastasis of the thyroid cancer patient is high; otherwise, it indicates that the risk of thyroid cancer metastasis of the thyroid cancer patient is low; and (c) An output module, which is used to output the risk assessment result.

14. The device according to claim 13, characterized in that The risk assessment module is configured to: when performing risk assessment, use a random forest model to perform risk assessment on the input expression data.

15. The device according to claim 14, characterized in that The risk assessment module is configured to use the random forest model formula shown in the following formula (I) to perform risk assessment: The random forest model is a trained and optimized model, T is the number of decision trees, and x1, x2 and x3 correspond to the ΔCt values ​​of the characteristic genes (A1) RPS4Y1; (A2) PKHD1L1; and (A3) CRABP1, respectively.

16. The device according to claim 13, characterized in that The expression data described were obtained using the detection reagents: Wherein, the detection reagent comprises: (P1) Primer pair for amplifying RPS4Y1: SEQ ID NO: 1 and SEQ ID NO: 2 and probe 1 used in combination: SEQ ID NO: 9; (P2) Primer pair for amplifying PKHD1L1: SEQ ID NO: 3 and SEQ ID NO: 4 and probe 2 used in combination: SEQ ID NO: 10; (P3) a primer pair for amplifying CRABP1: SEQ ID NO: 5 and SEQ ID NO: 6 and a probe 3 used in combination: SEQ ID NO: 11; and The internal reference primer pair is SEQ ID NO: 7 and SEQ ID NO: 8 and the probe 4 used in conjunction with it is SEQ ID NO:

12.

17. The device according to claim 13, characterized in that The device also includes a detection module, which is used to detect the mRNA level of the risk marker.

18. The device according to claim 17, characterized in that The detection module includes a PCR detector.

Images