A diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines

Abstract

The present invention relates to the field of biological diagnosis technology, and in particular to a diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines, including a classification joint module, which is constructed by integrating characteristic markers of platelet transcriptome, platelet proteome and plasma cytokine group, wherein the characteristic marker is selected from the platelet transcriptome, the platelet proteome and the plasma cytokine group in the biomarker detection data. The biomarker with a SHAP value greater than 0.01. The diagnostic kit provided by the present invention can effectively identify characteristic markers related to triple-negative ET, providing strong support for precision medicine and clinical diagnosis. At the same time, the diagnostic kit performs multiple tests on platelet RNA, protein and plasma cytokines through a single blood sample, which reduces the economic burden of patients and has broad clinical application prospects.

Description

Technical Field

[0001] The present invention relates to the field of biological diagnosis technology, and in particular to a diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines. Background Art

[0002] Essential thrombocytosis (ET) is a type of malignant blood system disease of pluripotent hematopoietic stem cells. Triple negative is a special type of essential thrombocythemia, which means that during the course of the disease, common JAK2, CALR and MPL mutations cannot be detected in its primitive and mature cells. The main common symptoms are headache, dizziness, bleeding, etc., which are easily confused with other common diseases (such as hypertension, diabetes, etc.). Due to its hidden onset and non-specific clinical characteristics, relying solely on clinical manifestations for diagnosis is prone to misdiagnosis or missed diagnosis. Although bone marrow puncture can evaluate the morphological changes of the bone marrow, for patients with triple negative ET, there may be no significant abnormalities in the bone marrow, or the morphological manifestations are similar to other blood diseases. Bone marrow puncture may not provide sufficient diagnostic information, especially for early cases or mild cases. In addition, the bone marrow morphology of ET may be similar to normal bone marrow changes, and the manifestations vary greatly among different patients, resulting in limited sensitivity and specificity as a diagnostic tool. Although genetic testing can detect common JAK2, CALR, and MPL mutations, existing detection methods may have low sensitivity for some uncommon gene mutations (such as TET2, ASXL1, etc.), and there is a lack of adequate standardized methods to evaluate the clinical significance of these variants. Some gene mutations occur less frequently in ET, and the relationship between these mutations and the clinical phenotype of the disease is not clear, which limits the diagnostic value of genetic testing in triple-negative ET.

[0003] Based on the above reasons, the current clinical diagnosis of triple-negative ET relies on a comprehensive evaluation of clinical symptoms, hematological characteristics, bone marrow morphology and even genetic testing. This requires not only the exclusion of other diseases during the diagnosis process, but also a high level of experience for the physician. At the same time, the need for multiple clinical tests to assist in diagnosis greatly increases the economic burden on clinical patients. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the related art. To this end, the object of the present invention is to provide a diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] A diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines, comprising a classification joint module, wherein the classification joint module is constructed by integrating characteristic markers of platelet transcriptome, platelet proteome and plasma cytokine group;

[0007] Wherein, the characteristic marker is selected from at least one of the platelet transcriptome, the platelet proteome and the plasma cytokine group;

[0008] The characteristic markers of the platelet transcriptome include: ADIPOR2, GFUS, TSPAN4, GANAB, KAT6A, ITGB1BP1, CKS1B, PRPF19, SP140, C4ORF48, AAMP, PEPD, GHDC, ARSG, METRN, SLC35C1, ME3, ADCK5, AANAT and ACAT2;

[0009] The characteristic markers of the platelet proteome include: CCDC88A, LCP1, SH3BGRL2, GGA2, RAB3D, SAR1B, RILPL2, ZNF330, DAPP1, RNASE, L TMEM63B, ANXA6, ANXA1, BDH1, CAT, RRM2B, TAP1 and GK2;

[0010] The characteristic markers of the plasma cytokine group include: LIF, IL-10, CXCL10, NRTN, MMP-1, CXCL6 and CXCL9.

[0011] Furthermore, the platelet transcriptome is derived from platelet RNA in a blood sample, the platelet proteome is derived from platelet protein in a blood sample, and the plasma cytokine group is derived from plasma in a blood sample.

[0012] Furthermore, platelet RNA was extracted using TRIzol reagent.

[0013] Furthermore, the platelet proteome in blood samples was quantified by TMT.

[0014] Furthermore, the biomarker data is selected from Olink data.

[0015] Furthermore, the classification joint module is trained using the XGBoost algorithm.

[0016] Furthermore, a five-fold crossover was used to verify the reliability of the diagnostic kit.

[0017] The above one or more technical solutions in the embodiments of the present invention have at least one of the following technical effects:

[0018] The present invention provides a diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines, including a classification joint module, which is constructed by integrating characteristic markers of platelet RNA transcriptome, platelet proteome and plasma cytokine group; by constructing the classification joint module, the diagnostic kit can accurately diagnose triple-negative ET in patients. At the same time, the diagnostic kit performs multiple tests on platelet RNA, protein and plasma cytokines through a single blood sample, which reduces the economic burden of patients and has broad clinical application prospects.

[0019] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a diagram of the differential expression analysis results of each group provided in the embodiments of the present invention.

[0021] Figure 2 It is a marker with a SHAP value greater than 0.01 in the platelet transcriptome obtained by screening the discovery set provided in the embodiment of the present invention.

[0022] Figure 3 It is a marker with a SHAP value greater than 0.01 in the platelet proteome obtained by screening the discovery set provided in the embodiment of the present invention.

[0023] Figure 4 It is a marker with a SHAP value greater than 0.01 in the plasma cytokine group obtained by screening the discovery set provided in the embodiment of the present invention.

[0024] Figure 5 This is a graph of five-fold cross-validation results within the discovery set provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present invention. The following embodiments are used to illustrate the present invention, but cannot be used to limit the scope of the present invention.

[0026] In the following examples, the experimental methods used are conventional methods unless otherwise specified, and the materials, reagents, etc. used are all available from commercial sources unless otherwise specified.

[0027] The samples of the present invention include discovery set samples and test set samples:

[0028] Among them, the discovery set contained 43 samples, including 9 healthy people, 10 triple-negative ETs, and 24 patients with non-triple-negative thrombocytosis; the test set contained 60 samples, including 16 healthy people, 10 triple-negative ETs, and 34 patients with non-triple-negative thrombocytosis.

[0029] A diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines, comprising a classification joint module, wherein the classification joint module is constructed by integrating characteristic markers of platelet transcriptome, platelet proteome and plasma cytokine group;

[0030] Wherein, the characteristic markers are selected from biomarkers having SHAP values ​​greater than 0.01 in the platelet transcriptome, the platelet proteome and the plasma cytokine group in the biomarker detection data.

[0031] According to a specific embodiment provided by the present invention, the construction process of a diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines is as follows:

[0032] 1. Transcriptomic analysis of platelets in healthy subjects and patients with different types of thrombocythemia.

[0033] The platelet RNA transcriptomic analysis process is as follows:

[0034] Platelet RNA was extracted from blood samples of different populations using TRIzol® reagent. RNA integrity was assessed using an Agilent 2100 bioanalyzer, and RNA concentration was determined using a NanoDrop spectrophotometer. Approximately 1 mg of high-quality RNA was selected for construction of sequencing libraries.

[0035] In the analysis of RNA-seq data, FastQC is used to assess the quality of the raw data and identify problems that need to be corrected. The quality control is then performed using the Trimmomatic processing tool, including removing low-quality sequences, cutting adapter sequences, and adjusting sequence length to ensure data quality.

[0036] The cleaned data were aligned to the reference genome using the Hierarchical Graph FM index for SequenceAlignment (HISAT2) to generate SAM files, which were then converted to more efficient BAM format files using SAMtools software. The StringTie data analysis software was used to assemble transcripts and generate count matrices to quantify gene expression levels, and the number of transcripts per million (TPM) was used to normalize gene expression and filter low-expression genes to ensure data accuracy and reliability. The biomaRt package in R language was used to obtain gene annotations from the Ensembl database to ensure accurate gene annotations, and principal component analysis (PCA) and limma differential analysis were performed on the data.

[0037] 2. Platelet proteomic analysis of healthy subjects and patients with different types of thrombocythemia.

[0038] Platelet proteomics was labeled by tandem mass tag (TMT), and the analysis process was as follows:

[0039] The experiment was designed for clinical cohort samples, taking into account the different clinical characteristics of the population, so that the samples were dispersed as evenly as possible to minimize batch effects. The present invention prepared quality control (QC) samples while processing the samples. The QC samples were mixed samples of equal volume of the experimental samples, and the system stability was evaluated during the entire experiment.

[0040] 1×10 8 Platelets were lysed in a solution containing 8 M urea, 50 mM 2-hydroxyethylpiperazineethanesulfonic acid (HEPES), 0.5% sodium deoxycholate, and 1× phoSTop TM The protein concentration was determined by a micro-BCA protein quantification kit in the lysate. After the lysate was diluted four times with 50 mM HEPES, Lys-C was added at a ratio of 1:100 for digestion for 3 hours, followed by trypsin at a ratio of 1:30 for digestion for 3 hours, and continued to digest overnight at the same ratio. The protein was reduced with 1 mM dithiothreitol (DTT), alkylated with 10 mM iodoacetamide (IA), and terminated with 30 mM DTT.

[0041] The digested peptides were purified using a reversed-phase Sep-Pak C18 column and eluted with a solution containing 60% acetonitrile and 0.1% trifluoroacetic acid (TFA). The peptides were then dried under vacuum and redissolved in 50 mM HEPES. The peptides were labeled using TMTpro 18plex labeling reagent, and different sample batches were labeled separately. Each batch of peptides was separated into 64 fractions and combined into 30 fractions. The fractions were further dried and redissolved in 0.1% formic acid. All fractions were analyzed using an Orbitrap Exploris 480 mass spectrometer, which was connected to a nanoliter Ultra Performance Liquid Chromatography (UPLC) system U3000 via an NSI source.

[0042] Peptides were separated on a reversed-phase analytical column (PepMap RSLC C18, 2 μm, 100 Å, 75 μm × 15 cm) using a 60-min linear gradient between 5% and 40% solvent B (65% acetonitrile, 0.2% formic acid, 3% dimethyl sulfoxide) and solvent A (0.2% formic acid, 3% dimethyl sulfoxide) at a flow rate of 200 nl / min.

[0043] Data acquisition was performed in data dependent acquisition (DDA) mode and set to Top 20 MS / MS scan. MS1 scan was in the range of m / z 410-1600, the orbitrap resolution was 60000, the automatic gain control (AGC) target value was set to 300%, and the maximum ion injection time was 50ms. MS / MS scan used high energy collision dissociation, NCE was set to 35%, the ion isolation width was m / z 1.0, and single charge and charge state higher than S ions were excluded.

[0044] The mass spectrometry raw data were processed using JUMP software (v2.4.0.305) and the human protein database (Universal Protein, UniProt). The parameters were set as follows: the specificity of the enzyme cleavage was limited to complete trypsinization, and a maximum of two missed cleavages were allowed. The fixed modification of cysteine ​​was alkylation (+57.02146 Da), the N-terminus and lysine residues were set to static TMT modification (+304.2071453 Da), and the dynamic oxidation modification of methionine (+15.99492 Da) was considered, with a maximum of three dynamic modification sites allowed. The precursor ion mass error tolerance was set to 15 ppm, and the product ion tolerance was set to 10 ppm.

[0045] The peptide-spectrum matches (PSM) standard was a target false discovery rate (FDR) of 1%. To ensure high-quality PSM, the precursor peak intensity was required to be at least 50%, and the minimum and median intensities of the TMT channel were set to 2000 and 10000, respectively. The data were aligned across batches using a sample pool labeled TMTpro-126 as an internal reference. More than 7,000 proteins in platelets were identified and quantified, and PCA dimensionality reduction analysis of the platelet proteome data revealed no batch effect between different batches.

[0046] 3. Proteomic analysis of plasma targeted cytokines related to healthy subjects and patients with different types of thrombocythemia.

[0047] The process of plasma targeted cytokine-related proteome analysis is as follows:

[0048] Sample randomization: Use the RAND function in the Excel table to output the numerical value corresponding to each sample, and then arrange the values ​​from small to large so that the samples are randomly distributed in the 96-well plate.

[0049] Sample sampling: According to the randomized sample distribution table, 10-40 µL of each sample was added to a 96-well plate, 5 µL of mixed plasma sample was added to wells A12 and B12, 5 µL of negative control group (NC) was added to wells C12, D12, and E12, and 5 µL of interplate control group (IPC) was added to wells F12, G12, and H12.

[0050] Hybridization incubation: Prepare hybridization mixture first, take out the reagent and thaw it at room temperature, vortex and centrifuge, take a nuclease-free 1.5ml centrifuge tube, add 280 µL of incubation solution, 40 µL of incubation stabilizer, 40 µL of A-probes, and 40 µL of B-probes in sequence, vortex and centrifuge. Take out an 8-tube strip and add 47 µL of hybridization mixture to each well. Take out another 96-well plate, mark it as an incubation plate, use an 8-channel pipette to add 3 µL of hybridization mixture to each well, and then add 1 µL of sample to each well one by one. After applying the sealing film, centrifuge at 400g for 1 minute, put it in the PCR instrument, set it to 4℃ infinity, record the time of putting it into the PCR instrument, and the incubation time must meet 16 hours.

[0051] Extension and amplification: Set the PCR program to 50℃ for 20 minutes; 95℃ for 5 minutes; 95℃ for 30 seconds, 54℃ for 1 minute, and 60℃ for 1 minute for 17 cycles; maintain at 10℃. After the settings are completed, run the PCR instrument and press the pause button to maintain the temperature of the PCR instrument at 50℃. To prepare the extension amplification mixture, take a 15 ml centrifuge tube, add 9385µL of high-purity water, 1100 µL of proximity extension assay (PEA) solution, 55µL of PEA enzyme, and 22µL of PCR polymerase in sequence, vortex and centrifuge. Take out the incubation plate from the PCR instrument and centrifuge. Use an 8-channel pipette to add 96 µL of the extension amplification mixture to each well, seal the film, vortex and centrifuge, place it on the PCR instrument preheated at 50℃, and then resume the program.

[0052] Chip pretreatment: Unpack and take out the chip, check whether the chip and syringe are damaged. Take out the syringe, press down the piston hole, and push the liquid in the syringe. Click the instrument Target 96 option, the instrument pushes out the sample stage, puts in the chip and cover, and automatically returns to the sample stage. Click Start, and the chip pretreatment begins.

[0053] Machine operation: Take out a 1.5 ml centrifuge tube, add 550 µL of detection buffer (Detection Solution), 230 µL of high purity water (High Purity Water), 7.8 µL of detection enzyme (Detection Enzyme), and 3.1 µL of PCR polymerase in sequence, vortex and centrifuge. Take out another strip tube and dispense it into 95 µL per well. Take out a 96-well plate, add 7.2 µL of mixed solution to each well, and after the extension amplification program is completed, add 2.8 µL of product, vortex and mix well for use. Prepare the diluted sample plate, primer plate and pre-treated chip, add 5µL of primer to the left side of the chip, and add 5µL of diluted product to the right side in sequence to avoid bubbles. After the sample is added, put it into the sample stage, set the run name, and click Run.

[0054] Data analysis and export: After the instrument is shut down, the run name file is exported, and the file suffix is ​​"XXX.q100". Open Olink NPX Signature 1.7.1.0 software, import "XXX.q100", set the reagent panel information, reagent version number, sample name, click "OK", and the software will automatically analyze and output NPX Excel document and QC AR PDF document.

[0055] 4. Integrate the characteristic markers of platelet transcriptome, platelet proteome and plasma cytokine group in the discovery set samples.

[0056] Through the above steps 1 to 3, the Olink data of platelet RNA transcriptome, platelet proteome and plasma cytokine group of 43 samples in the discovery set were collected, and differential expression analysis was performed on the data. The analysis results are as follows Figure 1 shown.

[0057] In order to reduce model overfitting and improve generalization ability, in the process of selecting feature markers, the present invention uses SHAP values ​​to analyze the importance of each feature to the model prediction results. Since this is a multi-classification task, the present invention calculates the SHAP value of each category and measures the contribution of each feature by taking the average absolute value. A larger SHAP value indicates that the feature has a greater impact on model prediction. Biomarkers with a contribution greater than 0.01 in the Olink data of the platelet transcriptome, platelet proteome, and plasma cytokine group were included and used for the construction of subsequent joint models. Biomarkers with SHAP values ​​greater than 0.01, such as Figure 2 , Figure 3 and Figure 4 As shown, the platelet RNA transcriptome includes the following markers: ADIPOR2, GFUS, TSPAN4, GANAB, KAT6A, ITGB1BP1, CKS1B, PRPF19, SP140, C4ORF48, AAMP, PEPD, GHDC, ARSG, METRN, SLC35C1, ME3, ADCK5, AANAT and ACAT2; the platelet proteome includes the following proteins: CCDC88A, LCP1, SH3BGRL2, GGA2, RAB3D, SAR1B, RILPL2, ZNF330, DAPP1, RNASE, L TMEM63B, ANXA6, ANXA1, BDH1, CAT, RRM2B, TAP1 and GK2; the plasma cytokine group includes the following cytokines: LIF, IL-10, CXCL10, NRTN, MMP-1, CXCL6 and CXCL9.

[0058] After standardization and preprocessing, the above biomarkers are input into the model as feature vectors for training and prediction by the XGBoost algorithm. The output of the model is a probability value that can predict whether the patient is a healthy person, a triple-negative ET patient, or a non-triple-negative ET patient.

[0059] The XGBoost algorithm and SHAP model were used to select characteristic markers for differentially expressed genes from different data sources, and a classification joint module was constructed by integrating these multi-omics characteristic markers. The data set was divided into three categories according to the sample type characteristics: normal healthy people, triple-negative ET, and non-triple-negative thrombocythemia, and the classification modules were trained separately on the training set using the XGBoost algorithm.

[0060] XGBoost is an additive model based on decision trees. It can be expressed as a weighted sum of a set of basis functions, which is as follows:

[0061] ;

[0062] in:

[0063] T is the number of trees in the model;

[0064] Indicates t The prediction results of the tree, X is the input feature vector;

[0065] F is the space of the tree model.

[0066] In multi-classification tasks, the softmax function is used to map the raw score of each category to the range [0, 1], and the sum of the probabilities of all categories is 1.

[0067] Use XGBoost's multi model to build a predictive classification model, specify the number of categories as 3, use MulticlassLogarithmic Loss as the model loss function, and the final probability conversion model is as follows:

[0068] in:

[0069] y=1 represents triple-negative ET; y=2 represents non-triple-negative ET; y=3 represents healthy people;

[0070] P ( y =1| X ) represents the probability of triple-negative ET after inputting the biomarker characteristics of the sample, such as RNA, protein, and cytokines, into the prediction model; P ( y =2| X ) represents the probability of non-triple-negative ET after inputting the biomarker characteristics of the sample, such as RNA, protein, and cytokines, into the prediction model; P ( y =3| X ) represents the probability of a healthy person after inputting the biomarker characteristics of the sample such as RNA, protein and cytokine into the prediction model;

[0071] Among them, e is a natural constant;

[0072] z1 is the cumulative score of the biomarker signature of the sample calculated by the model under the triple-negative ET category;

[0073] z2 is the cumulative score of the biomarker signature of the sample calculated by the model under the non-triple-negative ET category;

[0074] z3 is the cumulative score of the biomarker characteristics of the sample calculated by the model under the category of healthy people;

[0075] It is the sum of the three category scores calculated by the model after exponential transformation under a natural constant, which is used to normalize the probability to 1.

[0076] The discriminant model based on probability is as follows, and its function is expressed as follows:

[0077] ;

[0078] in:

[0079] C is the predicted category of the model;

[0080] arg max is a commonly used function, which means selecting the category with the maximum prediction probability as the prediction category.

[0081] Based on the above results, we built an XGBoost model with 50 decision trees to diagnose triple-negative thrombocythemia by modeling and classifying data from biomarkers such as RNA, protein, and cytokines. Each tree makes conditional judgments based on specific biomarkers, and finally outputs the classification results through the weighted prediction values ​​of multiple trees. The structure of the 50 decision trees output by the XGBoost model established by the model is as follows:

[0082] Decision tree 1: Node 0, judging whether the RNA feature PEPD is less than 2.619; if not, it reaches leaf node 1, and the output value is -0.206; if yes, it enters node 2 and continues to judge whether the protein feature CCDC88A is less than 349871; if yes, it reaches leaf node 3, and the output value is 0.351; if no, it reaches leaf node 4, and the output value is -0.129;

[0083] Decision tree 2: Node 0, judge whether the RNA feature TSPAN4 is less than 3.024; if not, it reaches leaf node 1, and the output value is 0.402; if yes, it enters node 2 and continues to judge whether the RNA feature ADIPOR2 is less than 6.827; if yes, it reaches leaf node 3, and the output value is -0.194; if no, it reaches leaf node 4, and the output value is -7.663×10 -9 ;

[0084] Decision tree 3: Node 0, judging whether the RNA feature ADIPOR2 is less than 5.680; if yes, it reaches leaf node 1, and the output value is 0.351; if no, it reaches leaf node 2, and the output value is -0.208;

[0085] Decision tree 4: Node 0, judge whether the RNA feature PEPD is less than 2.619; if not, it reaches leaf node 1, and the output value is -0.184; if yes, it enters node 2 and continues to judge whether the protein feature CCDC88A is less than 349871; if yes, it reaches leaf node 3, and the output value is 0.261; if no, it reaches leaf node 4, and the output value is -0.119;

[0086] Decision tree 5: Node 0, judge whether the RNA feature SP140 is less than 2.502; if yes, it reaches leaf node 1, and the output value is 0.287; if no, it enters leaf node 2 and continues to judge whether the RNA feature ARSG is less than 2.335; if yes, it reaches leaf node 3, and the output value is -0.176; if no, it reaches leaf node 4, and the output value is -0.002;

[0087] Decision tree 6: Node 0, judging whether the RNA feature ADIPOR2 is less than 5.680; if yes, it reaches leaf node 1, and the output value is 0.259; if no, it reaches leaf node 2, and the output value is -0.186;

[0088] Decision tree 7: Node 0, determine whether the RNA feature PEPD is less than 2.619; if not, reach leaf node 1, and the output value is -0.168; if yes, enter node 2, and continue to determine whether the protein feature CCDC88A is less than 349871; if yes, reach leaf node 3, and the output value is 0.210; if no, reach leaf node 4, and the output value is -0.105;

[0089] Decision tree 8: Node 0, judge whether the RNA feature SP140 is less than 2.502; if yes, it reaches leaf node 1, and the output value is 0.230; if no, it enters node 2 and continues to judge whether the protein feature RNASEL is less than 222250; if yes, it reaches leaf node 3, and the output value is -0.161; if no, it reaches leaf node 4, and the output value is 0.007;

[0090] Decision tree 9: Node 0, judging whether the RNA feature ADIPOR2 is less than 5.680; if yes, it reaches leaf node 1, and the output value is 0.212; if no, it reaches leaf node 2, and the output value is -0.170;

[0091] Decision tree 10: Node 0, judging whether the RNA feature PEPD is less than 2.61; if not, it reaches leaf node 2, and the output value is -0.156; if yes, it enters node 1 and continues to judge whether the protein feature CCDC88A is less than 348826; if yes, it reaches leaf node 3, and the output value is 0.175; if no, it reaches leaf node 4, and the output value is -0.035;

[0092] Decision tree 11: Node 0, judging whether the RNA feature TSPAN4 is less than 3.024; if not, it reaches leaf node 1, and the output value is 0.197; if yes, it enters node 2 and continues to judge whether the protein feature ERH is less than 35921; if yes, it reaches leaf node 3, and the output value is 0.013; if no, it reaches leaf node 2, and the output value is -0.148;

[0093] Decision tree 12: Node 0, judging whether the RNA feature ADIPOR2 is less than 5.680; if yes, it reaches leaf node 1, and the output value is 0.182; if no, it reaches leaf node 2, and the output value is -0.158;

[0094] Decision tree 13: Node 0, determine whether the RNA feature PEPD is less than 2.619; if not, reach leaf node 1, and the output value is -0.146; if yes, enter node 2, and continue to determine whether the protein feature SAR1B is less than 1522819; if yes, reach leaf node 3, and the output value is -0.040; if no, reach leaf node 4, and the output value is 0.155;

[0095] Decision tree 14: Node 0, judging whether the RNA feature SP140 is less than 2.502; if yes, it reaches leaf node 2, and the output value is 0.177; if no, it enters node 2 and continues to judge whether the RNA feature C4orf48 is less than 0.175; if yes, it reaches leaf node 3, and the output value is -0.136; if no, it reaches leaf node 4, and the output value is -0.009;

[0096] Decision tree 15: Node 0, judging whether the RNA feature ADIPOR2 is less than 5.680; if yes, it reaches leaf node 1, and the output value is 0.161; if no, it reaches leaf node 2, and the output value is -0.148;

[0097] Decision tree 16: Node 0, determine whether the RNA feature SLC35C1 is less than 0.685; if not, reach leaf node 1, and the output value is -0.037; if yes, enter node 2, and continue to determine whether the RNA feature SP140 is less than 2.525; if yes, reach leaf node 3, and the output value is -0.137; if no, reach leaf node 4, and the output value is 0.139;

[0098] Decision tree 17: Node 0, judging whether the RNA feature SP140 is less than 2.502; if yes, it reaches leaf node 1, and the output value is 0.161; if no, it enters node 2 and continues to judge whether the protein feature ANXA1 is less than 240667; if yes, it reaches leaf node 3, and the output value is -0.005; if no, it reaches leaf node 4, and the output value is -0.126;

[0099] Decision tree 18: Node 0, judging whether the RNA feature ADIPOR2 is less than 5.680; if yes, it reaches leaf node 1, and the output value is 0.146; if no, it reaches leaf node 2, and the output value is -0.140;

[0100] Decision tree 19: Node 0, judging whether the RNA feature PRPF19 is less than 3.678; if yes, it reaches leaf node 1, and the output value is 0.121; if no, it enters node 2 and continues to judge whether the RNA feature AANAT is less than -2.075; if yes, it reaches leaf node 3, and the output value is 0.023; if no, it reaches leaf node 4, and the output value is -0.132;

[0101] Decision tree 20: Node 0, judging whether the RNA feature ADIPOR2 is less than 6.827; if yes, it reaches leaf node 1, and the output value is -0.120; if no, it enters node 2 and continues to judge whether the RNA feature AANAT is less than -1.840; if yes, it reaches leaf node 3, and the output value is 0.032; if no, it reaches leaf node 4, and the output value is 0.150;

[0102] Decision tree 21: Node 0, judging whether the RNA feature ADIPOR2 is less than 5.680; if yes, it reaches leaf node 1, and the output value is 0.132; if no, it reaches leaf node 2, and the output value is -0.132;

[0103] Decision tree 22: Node 0, determine whether the RNA feature SLC35C1 is less than 0.685; if not, go to leaf node 1, the output value is -0.121; if yes, go to node 2, and continue to determine whether the RNA feature AANAT is less than -2.276; if yes, go to leaf node 3, the output value is 0.113; if not, go to node 4, the output value is -0.004;

[0104] Decision tree 23: Node 0, judging whether the RNA feature SP140 is less than 2.502; if yes, it reaches leaf node 1, and the output value is 0.139; if no, it enters node 2 and continues to judge whether the protein feature RAB3D is less than 9932204; if yes, it reaches leaf node 3, and the output value is -0.107; if no, it reaches leaf node 2, and the output value is -0.018;

[0105] Decision tree 24: Node 0, judging whether the RNA feature ADIPOR2 is less than 5.680; if yes, it reaches leaf node 1, and the output value is 0.120; if no, it reaches leaf node 2, and the output value is -0.124;

[0106] Decision tree 25: Node 0, determine whether the RNA feature PEPD is less than 2.619; if not, reach leaf node 2, the output value is -0.113; if yes, enter node 1, and continue to determine whether the protein feature CCDC88A is less than 347642; if yes, reach leaf node 3, the output value is 0.104; if no, reach leaf node 4, the output value is -0.003;

[0107] Decision tree 26: Node 0, determine whether the RNA feature TSPAN4 is less than 3.024; if not, reach leaf node 1, and the output value is 0.130; if yes, enter node 2, and continue to determine whether the RNA feature PRPF19 is less than 3.864; if yes, reach leaf node 3, and the output value is -0.097; if no, reach leaf node 2, and the output value is -0.024;

[0108] Decision tree 27: Node 0, judging whether the RNA feature ADIPOR2 is less than 5.680; if yes, it reaches leaf node 1, and the output value is 0.110; if no, it reaches leaf node 2, and the output value is -0.117;

[0109] Decision tree 28: Node 0, determine whether the RNA feature SLC35C1 is less than 0.685; if not, reach leaf node 1, and the output value is -0.106; if yes, enter node 2, and continue to determine whether the cytokine feature CXCL10 is less than 7.624; if yes, reach leaf node 3, and the output value is 0.011; if no, reach leaf node 4, and the output value is 0.090;

[0110] Decision tree 29: Node 0, judging whether the RNA feature SP140 is less than 2.502; if yes, it reaches leaf node 1, and the output value is 0.122; if no, it reaches leaf node 2, and the output value is -0.079;

[0111] Decision tree 30: Node 0, judging whether the RNA feature ADIPOR2 is less than 5.680; if yes, it reaches leaf node 1, and the output value is 0.102; if no, it reaches leaf node 2, and the output value is -0.110;

[0112] Decision tree 31: Node 0, judging whether the RNA feature ITGB1BP1 is less than 2.748; if yes, it reaches leaf node 1, and the output value is 0.061; if no, it reaches leaf node 2, and the output value is -0.099;

[0113] Decision tree 32: Node 0, judging whether the RNA feature ADIPOR2 is less than 6.827; if yes, it reaches leaf node 1, and the output value is -0.093; if no, it reaches leaf node 2, and the output value is 0.102;

[0114] Decision tree 33: Node 0, judging whether the RNA feature ADIPOR2 is less than 5.680; if yes, it reaches leaf node 1, and the output value is 0.094; if no, it reaches leaf node 2, and the output value is -0.103;

[0115] Decision tree 34: Node 0, judging whether the RNA feature AANAT is less than -1.840; if yes, it reaches leaf node 1, and the output value is 0.067; if no, it reaches leaf node 2, and the output value is -0.084;

[0116] Decision tree 35: Node 0, judging whether the RNA feature SP140 is less than 2.502; if yes, it reaches leaf node 1, and the output value is 0.108; if no, it reaches leaf node 1, and the output value is -0.070;

[0117] Decision tree 36: Node 0, judging whether the RNA feature ADIPOR2 is less than 5.680; if yes, it reaches leaf node 1, and the output value is 0.086; if no, it reaches leaf node 2, and the output value is -0.096;

[0118] Decision tree 37: Node 0, judging whether the RNA feature SLC35C1 is less than 0.685; if yes, it reaches leaf node 1, and the output value is 0.056; if no, it reaches leaf node 2, and the output value is -0.087;

[0119] Decision tree 38: Node 0, judging whether the protein feature TMEM63B is less than 231548; if yes, it reaches leaf node 1, and the output value is -0.076; if no, it reaches leaf node 2, and the output value is 0.092;

[0120] Decision tree 39: Node 0, judging whether the RNA feature ADIPOR2 is less than 6.248; if yes, it reaches leaf node 1, and the output value is 0.072; if no, it reaches leaf node 2, and the output value is -0.088;

[0121] Decision tree 40: Node 0, judging whether the RNA feature PEPD is less than 2.619; if yes, it reaches leaf node 1, and the output value is 0.052; if no, it reaches leaf node 2, and the output value is -0.081;

[0122] Decision tree 41: Node 0, judging whether the RNA feature SP140 is less than 2.502; if yes, it reaches leaf node 1, and the output value is 0.096; if no, it reaches leaf node 2, and the output value is -0.062;

[0123] Decision tree 42: Node 0, judging whether the protein feature LCP1 is less than 649070; if yes, it reaches leaf node 1, and the output value is -0.080; if no, it reaches leaf node 2, and the output value is 0.060;

[0124] Decision tree 43: Node 0, judging whether the RNA feature AANAT is less than -1.840; if yes, it reaches leaf node 1, and the output value is 0.056; if no, it reaches leaf node 2, and the output value is -0.068;

[0125] Decision tree 44: Node 0, judging whether the RNA feature SP140 is less than 2.502; if yes, it reaches leaf node 1, and the output value is 0.090; if no, it reaches leaf node 2, and the output value is -0.057;

[0126] Decision tree 45: Node 0, reaches leaf node 1, output value is -0.014;

[0127] Decision tree 46: Node 0, judging whether the RNA feature KAT6A is less than 5.957; if yes, it reaches leaf node 1, and the output value is -0.062; if no, it reaches leaf node 2, and the output value is 0.051;

[0128] Decision tree 47: Node 0, judging whether the RNA feature TSPAN4 is less than 3.024; if yes, it reaches leaf node 1, and the output value is -0.055; if no, it reaches leaf node 2, and the output value is 0.086;

[0129] Decision tree 48: Node 0, reaches leaf node 1, output value is -0.013;

[0130] Decision tree 49: Node 0, judging whether the RNA feature PEPD is less than 2.425; if yes, it reaches leaf node 1, and the output value is 0.041; if no, it reaches leaf node 2, and the output value is -0.056;

[0131] Decision tree 50: Node 0, judging whether the RNA feature SP140 is less than 2.502; if yes, it reaches leaf node 1, and the output value is 0.082; if no, it reaches leaf node 2, and the output value is -0.053;

[0132] The above decision tree structure shows how to construct a decision path through specific biomarkers, such as RNA expression levels and protein concentrations.

[0133] In the classification joint module training stage, the present invention uses five-fold cross validation (StratifiedKFold) within the discovery set, and the results are as follows Figure 5As shown in the figure, the classification joint model is further evaluated by combining five evaluation indicators: accuracy, F1-Score, precision, recall and area under the curve (AUC). The evaluation results are shown in Table 1.

[0134] Table 1 Evaluation results of each group and its classification joint model on the discovery set

[0135]

[0136] 5. Validation of the diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines.

[0137] Finally, the present invention used a test set of an independent sample cohort to verify the trained XGBoost classification model. The test set had a total of 60 samples, including 16 healthy people, 10 triple-negative ET, and 34 non-triple-negative thrombocytosis patients. The verification results are shown in Table 2.

[0138] Table 2 Independent test set verification results

[0139]

[0140] The present invention provides a diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines, which fully utilizes multi-omics data and machine learning methods from data preprocessing, feature selection to model training and evaluation, and establishes a reliable diagnostic kit. The diagnostic kit can effectively identify characteristic markers related to triple-negative ET, providing strong support for precision medicine and clinical diagnosis.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines, characterized in that: The method comprises a classification joint module, wherein the classification joint module is constructed by integrating characteristic markers of a platelet transcriptome, a platelet proteome, and a plasma cytokine group; Wherein, the characteristic marker is selected from at least one of the platelet transcriptome and the platelet proteome; The characteristic markers of the platelet transcriptome include: KAT6A, ITGB1BP1, PRPF19, PEPD, GHDC, ARSG, SLC35C1, ADCK5 and AANAT; The characteristic markers of the platelet proteome include: GGA2, RAB3D, SAR1B, ZNF330, RNASEL, ATPAF1, BDH1, RRM2B, TAP1 and GK2.

2. The diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines according to claim 1, characterized in that: The platelet transcriptome is derived from platelet RNA in a blood sample, the platelet proteome is derived from platelet protein in a blood sample, and the plasma cytokine group is derived from plasma in a blood sample.

3. The diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines according to claim 2, characterized in that: Platelet RNA was extracted using TRIzol reagent.

4. The diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines according to claim 2, characterized in that: Quantification of platelet proteome in blood samples by TMT.

5. The diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines according to claim 1, characterized in that: Biomarker data were selected from Olink data.

6. The diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines according to any one of claims 1 to 5, characterized in that: The classification joint module is trained using the XGBoost algorithm.

7. The diagnostic kit for triple-negative thrombocythemia based on platelet RNA, protein and plasma cytokines according to any one of claims 1 to 5, characterized in that: A five-fold crossover test was used to verify the reliability of the diagnostic kit.

Images