Biomarker composition for machine learning-based diagnosis of neurodegenerative diseases, and use thereof

Abstract

The present invention relates to a biomarker composition for diagnosing neurodegenerative diseases, comprising at least one gene selected from the group consisting of NADPH oxidase 4 (NOX4), retinol binding protein 4 (RBP4) and C-X-C motif chemokine ligand 10 (CXCL10) or proteins encoded from the genes. It has been identified that the level of the biomarkers in a neurodegenerative disease animal model is lower than that of a normal group, and an optimal biomarker combination is identified using machine learning, and thus a biomarker having higher accuracy than conventional neurodegenerative disease diagnosis methods can be provided.

Description

Biomarker composition for diagnosing neurodegenerative diseases based on machine learning and use thereof

[0001] The present invention relates to a biomarker composition for diagnosing neurodegenerative diseases based on machine learning and its use, and more particularly, to a biomarker composition for diagnosing neurodegenerative diseases comprising at least one gene selected from the group consisting of NOX4 (NADPH oxidase 4), RBP4 (Retinol binding protein 4), and CXCL10 (CXC motif chemokine ligand 10) or a protein encoded by the gene, a composition for diagnosing neurodegenerative diseases comprising an agent capable of measuring the expression level of the gene or protein, a kit comprising the composition, and a method for providing information for diagnosing neurodegenerative diseases.

[0002] Cognitive Dysfunction Syndrome (CDS) is a type of neurodegenerative disease in animals, with a much higher incidence in dogs than in cats. It's estimated that 14% of companion dogs over the age of eight have CDS, but only 1.9% receive a proper diagnosis. The initial symptoms of CDS are mild, making accurate diagnosis difficult. However, early detection can slow the progression of the disease through diet management, exercise, and medication.

[0003] Existing diagnostic methods for cognitive dysfunction syndrome include the Canine Cognitive Dysfunction Rating (CCDR), a questionnaire-based self-assessment test, and imaging tests using MRI. However, the CCDR is subject to the subjective opinions of the owner and has limitations in detecting early cognitive changes in detail. Furthermore, MRI has the disadvantage of being relatively time-consuming, costly, and may not accurately reflect the complex neurological changes associated with early cognitive impairment.

[0004] Machine learning is a branch of artificial intelligence that empowers computers to learn without explicit programming. Machine learning has attracted significant attention in various fields, including new drug development, diagnostics, and biomarkers, as it can reduce costs, shorten time, and reduce the likelihood of side effects. In particular, machine learning can be used in biomarker research to rapidly sift through massive amounts of data and identify patterns and relationships that can indicate disease progression, enabling faster biomarker identification compared to conventional methods.

[0005] Against this backdrop, the inventors of the present invention compared the expression levels of NOX4 (NADPH oxidase 4), RBP4 (Retinol binding protein 4), and CXCL10 (CXC motif chemokine ligand 10) in blood samples of animal models of neurodegenerative diseases with those of blood samples of normal animals, and confirmed that the expression levels of NOX4, RBP4, and CXCL10 were reduced. By teaching this to a machine learning algorithm, the inventors derived an optimal combination of biomarkers capable of diagnosing neurodegenerative diseases, thereby completing the present invention.

[0006] [Prior Art Literature]

[0007] [Patent Document]

[0008] Republic of Korea Patent Publication No. 10-2569168 (August 17, 2023)

[0009] The purpose of the present invention is to provide a biomarker composition for diagnosing a neurodegenerative disease, a diagnostic composition, a kit including the composition, and a method for providing information for diagnosing a neurodegenerative disease.

[0010] To achieve the above purpose, the present invention provides a biomarker composition for diagnosing a neurodegenerative disease, comprising at least one gene selected from the group consisting of NOX4 (NADPH oxidase 4), RBP4 (Retinol binding protein 4), and CXCL10 (CXC motif chemokine ligand 10) or a protein encoded by the gene.

[0011] The present invention also provides a composition for diagnosing a neurodegenerative disease, comprising an agent for measuring the expression level of at least one mRNA selected from the group consisting of NOX4, RBP4, and CXCL10, or an agent for measuring the expression level of at least one protein selected from the group consisting of NADPH oxidase 4, Retinol binding protein 4, and CXC motif chemokine ligand 10.

[0012] The present invention also provides a kit for diagnosing a neurodegenerative disease comprising the composition.

[0013] The present invention also provides a method for providing information for diagnosing a neurodegenerative disease, comprising: (a) measuring the expression level of a gene consisting of NOX4 (NADPH oxidase 4), RBP4 (Retinol binding protein 4) and CXCL10 (CXC motif chemokine ligand 10) or a protein encoded by the gene in a biological sample isolated from an individual; and (b) comparing the expression level measured in step (a) with that of a biological sample isolated from a normal individual.

[0014] In the present invention, three biomarkers capable of early diagnosis of neurodegenerative diseases were selected and trained on a machine learning-based algorithm to derive an optimal biomarker combination. Using the biomarkers of the present invention, it is possible to detect neurodegenerative diseases, which are difficult to diagnose and lack precise criteria, early, and to continuously monitor and manage the disease status.

[0015] Figure 1A shows the proteome array of the immunoblot, where the spots on the membrane are composed of inflammatory cytokines, chemokines, and growth factors. The red squares indicate the expression levels of RBP4 and CXCL10 in APP / PS1 mice and MPTP mice. Figure 1B quantifies the levels of RBP4 and CXCL10 in serum of the control and neurodegenerative disease models (Student's two-tailed t-test vs. control). * ,p<0.05; ** ,p<0.01; *** ,p<0.001).

[0016] Figure 2A shows the results of immunoblot analysis for the expression of RPB4, CXCL10, and NOX4 in the normal and CCDS groups. Figure 2B shows the results of quantifying the levels of RPB4, CXCL10, and NOX4 in the plasma of the normal and CCDS groups (Student's two-tailed t-test compared to the control group).* , p<0.05).

[0017] Figure 3A shows the mean results of quantifying RPB4, CXCL10, and NOX4 levels in the normal group, MCI group, SCI group, and CCDS group including MCI and SCI. Figure 3B shows the distribution of biomarker levels in each group. The points in the graph represent the distribution of individual samples, and the violin-shaped lines represent the quartiles and median (one-way ANOVA). * , p<0.05; ** ,p<0.01; *** ,p<0.001; **** , p<0.0001, compared to the control group by Student's two-tailed t-test ### ,p<0.001; #### ,p<0.0001).

[0018] Figure 4 shows a correlation heatmap for the variables RBP4, CXCL10, NOX4, CCDR, and age.

[0019] Figure 5A shows the ROC curve using the SVM algorithm for the combination of RBP4 and NOX4 between the normal group and the MCI group. Figure 5B shows the ROC curve using the Extra Tree algorithm for the combination of RBP4, CXCL10, and NOX4 between the normal group and the SCI group. Figure 5C shows the ROC curve using the SVM algorithm for the combination of RBP4 and NOX4 between the normal group and the CCDS group.

[0020] Figure 6A shows the proteome array by immunoblot, where the spots on the membrane are composed of inflammatory cytokines, chemokines, and growth factors. The red squares indicate the expression levels of cystatin C, fetuin-A, myeloperoxidase (MPO), and osteopontin (OPN) in APP / PS1 mice and MPTP mice. Figure 6B quantifies the levels of cystatin C, fetuin-A, MPO, and OPN in serum of the control and neurodegenerative disease models (Student's two-tailed t-test vs. control). * , p<0.05; ** ,p<0.01; *** ,p<0.001).

[0021] Figure 7A shows the results of quantifying P-tau (Phospho-tau) and NfL (Neurofilament) levels in the normal and CCDS groups. Figure 7B shows the distribution of biomarker levels in each group. The points in the graph represent the distribution of individual samples, and the violin-shaped lines represent the quartiles and median (Student's two-tailed t-test).

[0022] Figure 8 shows the characteristics of the experimental animals used in this experiment.

[0023] Hereinafter, the present invention will be described in detail.

[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Generally, the nomenclature used herein is well known and commonly used in the art.

[0025] When the present invention says that a component or a step "includes", this does not mean that other components or other steps are excluded, but rather that other components or other steps may be further included, unless specifically stated otherwise.

[0026] In the present invention, the term "biomarker" means a molecule quantitatively or qualitatively related to the presence of a biological phenomenon, and may include all organic biomolecules such as proteins, nucleic acids, genes, lipids, glycolipids, glycoproteins, sugars, etc.

[0027] In the present invention, the term "biomarker composition" may include primers, probes, antisense oligonucleotides, etc. that can specifically bind to a biomarker and measure its expression level.

[0028] The above primers are fragments that recognize a target gene sequence, and include a pair of forward and reverse primers, but are preferably primer pairs that provide analysis results with specificity and sensitivity. When the nucleic acid sequence of the primer is a sequence that does not match the non-target sequence present in the sample, and thus the primer only amplifies the target gene sequence containing the complementary primer binding site and does not cause non-specific amplification, high specificity can be imparted.

[0029] The above probe refers to a substance that can specifically bind to a target substance to be detected in a sample, and refers to a substance that can specifically confirm the presence of the target substance in the sample through the binding. The type of the probe is not limited to a substance commonly used in the art, but may preferably be PNA (peptide nucleic acid), LNA (locked nucleic acid), peptide, polypeptide, protein, RNA, or DNA. More specifically, the probe includes a biomaterial derived from or similar to a living organism or manufactured in vitro, and may be, for example, an enzyme, a protein, an antibody, a microorganism, an animal or plant cell and organ, a nerve cell, DNA, and RNA. DNA includes cDNA, genomic DNA, and oligonucleotides, RNA includes genomic RNA, mRNA, and oligonucleotides, and examples of proteins may include antibodies, antigens, enzymes, peptides, etc.

[0030]

[0031] The present invention provides a biomarker composition for diagnosing a neurodegenerative disease, comprising at least one gene selected from the group consisting of NOX4 (NADPH oxidase 4), RBP4 (Retinol binding protein 4), and CXCL10 (CXC motif chemokine ligand 10) or a protein encoded by the gene.

[0032] The above NOX4 is a gene encoding NADPH oxidase 4. NADPH oxidase 4 is an enzyme included in the seven-member NADPH oxidase complex including Nox1-5 and Duox 1 and 2, and plays a role in reducing oxygen into reactive oxygen species (ROS).

[0033] The above RBP4 gene encodes retinol binding protein 4 (RBP4). RBP4 transports the fat-soluble vitamin retinol from the liver to peripheral tissues. Vitamin A is converted to retinol in the small intestinal epithelium, esterified, and circulates in the blood. Later, in liver cells, retinyl ester hydrase converts it back to retinol, where it binds to RBP synthesized in the liver and forms a complex with prealbumin, which then circulates.

[0034] CXCL10 is a gene encoding CXC motif chemokine ligand 10, which is involved in development, immune response, inflammatory response, tumor growth, and metastasis. CXCL10 can also be named interferon gamma-induced protein 10 or small-inducible cytokine B10.

[0035] The above NOX4 may be composed of an amino acid sequence represented by SEQ ID NO: 1 and may be encoded by a base sequence represented by SEQ ID NO: 2. The above RBP4 may be composed of an amino acid sequence represented by SEQ ID NO: 3 and may be encoded by a base sequence represented by SEQ ID NO: 4. The above CXCL10 may be composed of an amino acid sequence represented by SEQ ID NO: 5 (canine) or SEQ ID NO: 7 (human) and may be encoded by a base sequence represented by SEQ ID NO: 6 (canine) or SEQ ID NO: 8 (human). Those skilled in the art will appreciate that an amino acid sequence that maintains 80% or more, preferably 90% or more, more preferably 95% or more, and most preferably 98% or more homology to the above amino acid sequence through artificial or natural modification can be used as a marker of the present invention.

[0036] The above neurodegenerative diseases include Cognitive Dysfunction Syndrome (CDS), Parkinson's Disease (PD), Alzheimer's Disease (AD), Progressive Supranuclear Palsy (PSP), Multiple System Atrophy (MSA), Olivopontocerebellar Atrophy (OPCA), Shy-Drager Syndrome (SDS), Striatonigral Degeneration (SND), Huntington's Disease (HD), Amyotrophic Lateral Sclerosis (ALS), Essential Tremor (ET), and Cortical-Basal Degeneration (CBD). It may be at least one selected from the group consisting of CBD, Diffuse Lewy Body Disease (DLBD), Parkinson's-ALS-Dementia Complex, Niemann-Pick Disease, Pick's Disease, Cerebral ischemia, and Cerebral infarction, and preferably, it may be at least one selected from the group consisting of cognitive dysfunction syndrome, Parkinson's disease, or Alzheimer's disease, but is not limited thereto.

[0037] The above cognitive dysfunction syndrome is a cognitive dysfunction syndrome in animals, a disease similar to Alzheimer's disease in humans, and may include, but is not limited to, mild cognitive impairment (MCI) and severe cognitive impairment (SCI).

[0038] The mRNA of the above gene may be any one or more base sequences selected from the group consisting of sequence numbers 1, 3, and 5.

[0039] The above biomarker is characterized by being extracted from one or more samples selected from the group consisting of blood, plasma, serum, lymph, saliva, urine, tissue, cerebrospinal fluid, ascites, or a combination thereof, and may preferably be blood, plasma, or serum, but is not limited thereto.

[0040] If the expression level of the above biomarker is reduced compared to the normal control group, it may be judged that there is a possibility of developing a degenerative neurological disease.

[0041] According to one embodiment of the present invention, in blood samples from an Alzheimer's mouse model with a mutated APP / PS1 gene and a Parkinson's disease mouse model induced by MPTP, the expression of RBP4 and CXCL10 was commonly reduced compared to the normal group, and when the expression of NOX4, which has been reported to be related to degenerative neurological diseases, was also examined, it was confirmed that the expression of RBP4, CXCL10, and NOX4 was significantly lower in the CCDS group.

[0042]

[0043] The present invention also provides a composition for diagnosing a neurodegenerative disease, comprising an agent for measuring the expression level of any one or more mRNAs selected from the group consisting of NOX4, RBP4, and CXCL10, or an agent for measuring the expression level of any one or more proteins selected from the group consisting of NADPH oxidase 4, Retinol binding protein 4, and CXC motif chemokine ligand 10.

[0044] The agent for measuring the mRNA expression level may be, but is not limited to, an antisense oligonucleotide, a primer pair, or a probe that specifically binds to the mRNA.

[0045] The agent for measuring the above protein expression level may be, but is not limited to, an antibody or an antigen-binding fragment thereof, a ligand, a PNA (Peptide nucleic acid) or an aptamer that specifically binds to the above protein.

[0046] The present invention also provides a kit for diagnosing a neurodegenerative disease comprising the composition.

[0047] The above kit may be, but is not limited to, an RT-PCR kit, a competitive RT-PCR kit, a real-time RT-PCR kit, a DNA chip kit, or a protein chip kit.

[0048]

[0049] The present invention also provides a method for providing information for diagnosing a neurodegenerative disease, comprising the following steps:

[0050] (a) a step of measuring the expression level of a gene consisting of NOX4 (NADPH oxidase 4), RBP4 (Retinol binding protein 4) and CXCL10 (CXC motif chemokine ligand 10) or a protein encoded by the gene in a biological sample isolated from an object; and

[0051] (b) A step of comparing the expression level measured in step (a) with a biological sample isolated from a normal individual.

[0052] The above sample may be extracted from one or more samples selected from the group consisting of blood, plasma, serum, lymph, saliva, urine, tissue, cerebrospinal fluid, ascites, or a combination thereof, and may preferably be blood, plasma, or serum, but is not limited thereto.

[0053] The above measurements include electrophoresis, immunoblotting, enzyme-linked immunosorbent assay (ELISA), polymerase chain reaction (PCR), Northern blotting, protein chip, immunoprecipitation, microarray, electron microscopy, western blotting, radioimmunoassay (RIA), radial immunodiffusion, Ouchterlony immunodiffusion, rocket immunoelectrophoresis, immunohistochemical staining, immunoprecipitation assay, complement fixation assay, immunofluorescence, It may be performed by, but is not limited to, immunochromatography, fluorescence activated cell sorter analysis (FACS analysis), or a combination thereof.

[0054] After the above step (b), if the expression level of the gene is reduced compared to a normal individual, a step of determining that there is a possibility of developing a degenerative neurological disease may be additionally included.

[0055] After the above step (b), a step of generating a model for predicting the onset of a neurodegenerative disease by teaching a machine learning algorithm model the expression level of the gene may be additionally included.

[0056] The above algorithm model may be one or more algorithm models selected from the group consisting of Random Forest, Support Vector Machine (SVM), Adaboost, Bagging, Gradient Boost, Extra Tree, and XgBoost, but is not limited thereto.

[0057] According to one embodiment of the present invention, when the expression level of a gene consisting of NOX4 (NADPH oxidase 4), RBP4 (Retinol binding protein 4) and CXCL10 (CXC motif chemokine ligand 10) or a protein encoded from the gene is trained on a machine learning algorithm model, among the groups combining two biomarkers, when RBP4 and NOX4 are combined and trained on an SVM algorithm model, MCI, SCI and CCDS can be predicted at the highest level, and when three biomarkers are combined, when RBP4, CXCL10 and NOX4 are combined and trained on an SVM algorithm model, MCI can be predicted at the highest level, and when trained on an Extra Tree algorithm model, SCI and CCDS can be predicted at the highest level, but is not limited thereto.

[0058]

[0059] The description of the diagnostic composition, kit and information providing method, its effects and all related descriptions are the same as those described above, and therefore, the description thereof is omitted to avoid excessive complexity of this specification due to duplicate description.

[0060]

[0061] Hereinafter, the present invention will be described in more detail through examples. These examples are intended solely to illustrate the present invention more specifically, and it will be apparent to those skilled in the art that the scope of the present invention is not limited by these examples, in accordance with the gist of the present invention.

[0062]

[0063] Example 1. Experimental materials and methods

[0064]

[0065] 1-1. Animal research

[0066]

[0067] [APP / PS1 Alzheimer's Mouse Model]

[0068] Serum samples from the APP (human amyloid precursor protein) / PS1 (mutant human presenilin 1) Alzheimer's mouse model were obtained from the Laboratory Animal Resources Bank (LAREB) of the National Institute of Food and Drug Safety Evaluation in Daegu, Korea. All experiments were conducted with the approval of the Daegu-Gyeongbuk Medical Innovation Foundation (Approval Number: DGMIF 21111602-00).

[0069]

[0070] [MPTP-induced Parkinson's disease mouse model]

[0071] C57BL / 6J mice (male, 7–8 weeks old, n=20) were used for the experiment. 1-Methyl-4-phenyl-1,2,3,6-tetrahydropyridine (S47312; Selleck Chemical, Houston, TX, USA; 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine; MPTP, n=10), a toxic substance that induces parkinsonism, was administered intraperitoneally at a dose of 30 mg / kg daily in 0.9% saline for 30 days. All experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of Soonchunhyang University (Approval Number: SCH23-0043).

[0072]

[0073] [Blood sample collection]

[0074] Blood samples were collected from 85 dogs at a veterinary hospital affiliated with iamdt Co., Ltd., with the consent of their guardians. Blood samples from healthy dogs and dogs with documented cognitive impairment using CCDR scores were placed in EDTA anticoagulant tubes and separated into plasma and blood cells by centrifugation. Within 1 hour of collection, plasma was transferred to cryogenic tubes and stored frozen at -80°C until transport to Soonchunhyang University. All blood samples used were reviewed and approved by the Soonchunhyang University IACUC (Approval Number: SCH23-0069). In addition, blood samples from three male beagle dogs (two 5-month-old dogs and one 96-month-old dog) that had not undergone any surgical or chemical treatment were obtained from the National Institute of Food and Drug Safety Evaluation (LAREB) in Daegu. The above experiment was approved by the Daegu-Gyeongbuk Medical Innovation Foundation (approval number for 5-month-old beagle dog: DGMIF-21071203-00, approval number for 96-month-old dog: KMEDI-22071501-00).

[0075]

[0076] 1-2. Dog experimental group

[0077]

[0078] Animals were classified into four groups: normal, mild cognitive impairment (MCI), severe cognitive impairment (SCI), and canine cognitive dysfunction syndrome (CCDS). MCI, SCI, and CCDS categories were categorized based on the CCDR. Initially, the scale, consisting of 13 behavioral items, was refined into 10 items after excluding three questions that were generally difficult to answer (Table 1). The cumulative score ranges from 0 to 60, with scores between 25 and 35 indicating MCI and scores ≥36 indicating SCI. Additionally, a CCDS group encompassing both MCI and SCI was included in the classification.

[0079]

[0080]

[0081]

[0082] 1-3. Proteome Profiler Array

[0083]

[0084] Cytokines, chemokines, and growth factors present in the mouse serum of the control and CCDS groups were semiquantitatively evaluated using the Proteome Profiler Mouse XL Cytokine Array Kit (ARY028; R&D Systems, Minneapolis, MN, USA). Serum samples were incubated on nitrocellulose membranes strictly according to the manufacturer's instructions and prestained with capture antibodies. Next, each membrane was treated with a biotinylated antibody cocktail, followed by streptavidin conjugation to horseradish peroxidase. HRP luminescence was expressed, and positive signals were captured using a Chemi Reagent Mix on light-sensitive X-ray film at an exposure time of 1 to 10 minutes. To analyze the expression level of each point on each membrane, quantification of FC (fold change) was performed using HLImage++ software (v25.0.0r, Western Vision Software, Salt Lake City, UT, USA), and the intensity was compared with the average value of the control sample.

[0085]

[0086] 1-4. Antibodies

[0087]

[0088] The following antibodies were used in this experiment: monoclonal mouse anti-RBP4 antibody (orb751184, Biorbyt, Berkeley, CA, USA); polyclonal rabbit anti-CXCL10 antibody (abx104024, Abbexa, Cambridge, UK); polyclonal rabbit anti-NOX4 antibody (NB110-58849, Novusbio, Centennial, CO, USA); polyclonal rabbit anti-transferrin antibody (NBP1-97472, Novusbio, Centennial, CO, USA); Peroxidase labeled horse anti-mouse IgG (H + L), 7076P2, Cell Signaling Technology, Danvers, MA, USA; peroxidase labeled goat anti-rabbit IgG (H + L), PI-1000, Vector Laboratories, Burlingame, CA, USA).

[0089]

[0090] 1-5. Immunoblot

[0091]

[0092] Dog plasma was homogenously mixed in RIPA buffer (R0278, Sigma-Aldrich, St. Louis, MO, USA) and supplemented with a phosphatase inhibitor (P3200, GenDEPOT, Barker, TX, USA). To accurately analyze plasma components, albumin was removed using Minute™ albumin depletion reagent. The effect of albumin removal was confirmed using SDS-PAGE and Coomassie Brilliant Blue staining (CR2006, Biosesang, Yongin, Korea). Protein concentration in the albumin-removed plasma was quantified using a BCA assay kit (21071, iNtRON Biotechnology, Seongnam, Korea). After SDS-PAGE on a 10-15% Tris-glycine gel, proteins were transferred to PVDF membranes (10600023, GE Healthcare, Freiburg, Germany). Nonspecific binding was blocked with 5% bovine serum albumin (BSA, SM-BOV, GeneAll Biotechnology, Seoul, Korea) in 1X TBS-T (TR2007, Biosesang, Yongin, Korea) containing 10X TBS and Tween 20. Primary antibodies, specifically monoclonal mouse anti-RBP4, polyclonal rabbit anti-CXCL10, and polyclonal rabbit anti-NOX4 antibodies, were incubated in 1% BSA in TBS-T overnight at 4°C. Subsequently, HRP-conjugated secondary antibodies, horse anti-mouse and goat anti-rabbit antibodies, were treated for 2 h at room temperature. Target proteins were detected using ECL Western blot detection reagents, and signals were visualized using a chemiluminescence bioimaging instrument (CELLGENTEK, Daejeon, Korea). Analysis was performed using ImageJ software v1.It was performed using 52t (Bethesda, MD, USA).

[0093]

[0094] 1-6. Enzyme-linked immunosorbent assay (ELISA) for diagnosis

[0095]

[0096] Blood samples were assayed for RBP4 (MBS739348), CXCL10 (MBS747479), NOX4 (MBS737351), P-Tau (MBS7230007), and NfL (MBS7231454) using ELISA kits from MyBioSource (San Diego, CA, USA). Each well of a plate precoated with the specific detection antibodies of each kit was injected with canine plasma containing RBP4-, CXCL10-, NOX4-, P-Tau-, and NfL-HRP-conjugated antibodies. The reaction was performed at 37°C for 1 h. After incubation, the plasma and antibodies were discarded, and each well was thoroughly washed with wash buffer. Subsequently, substrate solution was injected into each well, and the plates were incubated again for 15 min at 37°C under light-protected conditions. Upon addition of the stop solution, a colorimetric change from blue to yellow was observed, and the optical density of the solution was measured at 450 nm using a microplate reader.

[0097]

[0098] 1-7. Machine Learning for CCDS Analysis

[0099]

[0100] We analyzed the CCDS classification using the proposed biomarkers for machine learning. For this purpose, we utilized scikit-learn (v0.24.2) in Python (v3.6.13). In the machine learning classification task, three comparisons were performed: normal vs. MCI, normal vs. SCI, and normal vs. MCI and SCI. The number of samples used for classification was 50, 46, and 85, respectively. The machine learning model was trained and evaluated through 10 replicate experiments. Seventy percent of the total data was used as training data, and the remaining 30% as test data. Various algorithms were utilized for the machine learning models, including Support Vector Machine (SVM), Extra Tree, Random Forest, Gradient Boost, Bagging, Adaboost, and XGBoost, all of which are provided in Scikit-Learn. Performance was measured using metrics such as Area Under the Curve (AUC), accuracy, and F1 score.

[0101]

[0102] 1-8. Statistics

[0103]

[0104] All statistical analyses were performed using Prism 10 (GraphPad Software Inc., San Diego, CA, USA). Data are presented as mean ± standard error of the mean (Mean ± SEM). Statistical analyses were performed using Student's two-tailed t-test to compare two groups. For multiple group comparisons, analysis of variance (ANOVA) with post hoc comparisons using Tukey's multiple comparison test was performed. P values ​​( * p<0.05; **p<0.01; *** p<0.001; **** p<0.0001) was considered statistically significant.

[0105]

[0106] Example 2. Screening of novel biomarkers in Alzheimer's and Parkinson's disease models.

[0107]

[0108] We screened novel biomarkers within the proteomic profiles of blood samples from mouse models of Alzheimer's disease (AD) and Parkinson's disease (PD), as well as from normal animals (Figs. 1A and 6A). As a result, among 111 inflammatory cytokine markers analyzed, the expression of RBP4 and CXCL10 showed significant differences in both AD and PD models (Fig. 1A). Additionally, among the 111 inflammatory cytokine markers analyzed, the expression of Cystatin C, Fetuin-A, MPO, and OPN showed significant differences in the AD model, while Cystatin C, Fetuin-A, and MPO showed significant differences in the PD model (Fig. 6A). These results suggest that RBP4 and CXCL10 in the blood can serve as reliable indicators of neurodegenerative diseases (Figs. 1B and 6B).

[0109]

[0110] Example 3. Decreased expression of RBP4, CXCL10, and NOX4 in the plasma of dogs with neurological diseases.

[0111]

[0112] 3-1. Information on experimental animals

[0113] General details of the animals used in the present invention are shown in Table 2 and Figure 8.

[0114]

[0115]

[0116]

[0117] 3-2. Validation of Selected Biomarkers

[0118]

[0119] To verify the observations on RBP4 and CXCL10 in animal models of neurodegenerative diseases, Western blot analysis was performed (Fig. 2A). Additionally, the correlation between the expression of NOX4, which has been reported to be associated with neurodegenerative diseases (Sun Shin Yi et al, Redox Biol. 2021 May:41:101947), and the expression of RBP4 and CXCL10 was investigated. The expression of RBP4, CXCL10, and NOX4 was significantly lower in the plasma of CCDS dogs than in normal dogs, which was consistent with the results obtained in mice induced with APP / PS1 and MPTP (Fig. 2B).

[0120]

[0121] 3-3. Measurement of biomarker expression in additionally collected blood

[0122]

[0123] The utility of the biomarkers was confirmed in additional samples collected through ELISA (Figs. 3 and 7). Compared to the normal group, the expression of RBP4, CXCL10, and NOX4 was decreased in the plasma of the CCDS group. Although the p-value for RBP4 was not significant, a trend toward a decrease in RBP4 expression was observed at concentrations between the normal and CCDS dogs. In particular, CXCL10 and NOX4 showed significant decreases compared to the normal, MCI, and SCI groups, and the decrease in expression was more pronounced between the normal and CCDS groups (Fig. 3). The bar graphs quantify the RBP4, CXCL10, and NOX4 ELISA levels in each group, and the data are expressed as the mean ± SEM (Fig. 3A). The violin plots represent the distribution of biomarker levels in each group, with dots representing the distribution of each sample and lines representing the quartiles and medians (Fig. 3B). Compared to the normal group, the expression of P-Tau and NfL, known biomarkers of cognitive dysfunction, was not significantly different in the plasma of the CCDS group (Fig. 7). These results are consistent with those in mouse models of neurodegenerative diseases.

[0124]

[0125] Example 4. Machine Learning for CCDS Classification

[0126]

[0127] A machine learning algorithm was applied to evaluate the combination of biomarkers that could most effectively determine the disease status (normal, MCI, SCI, CCDS). Changing the number of samples used in machine learning can significantly affect the results. Therefore, to ensure an equal number of normal samples in the analysis group, the analysis was performed on an equal number of randomly selected samples. Ten randomly repeated experiments were performed using seven classification algorithms, each using 70% of the total data set as training data.

[0128]

[0129] Example 5. Checking correlations between variables

[0130]

[0131] Prior to developing a machine learning model, a correlation heatmap was generated using all 85 samples (Figure 4) to evaluate the relationships between variables. The analyzed variables included RBP4, CXCL10, NOX4, CCDR, and age. Higher and lower correlations are indicated in red and blue, respectively. In particular, the correlation coefficient between CCDR and age was 0.66, indicating an increasing trend in CCDR scores with increasing age. A significant correlation (r=0.7) was also observed between CXCL10 and NOX4. Furthermore, the results of RBP4, CXCL10, and NOX4 were inversely proportional to the CCDR score, and the levels decreased as the CCDR score increased.

[0132]

[0133] Example 6. Predicting MCI using Machine Learning

[0134]

[0135] To evaluate the differences between the normal and MCI groups, 25 MCI samples were matched and an equal number of normal samples were randomly selected for analysis. The F1 score is a machine learning performance evaluation index used in classification models, and a value of 0.7 or higher is interpreted as having excellent results. Among the biomarker combinations listed in Table 3, the combination of RBP4 and NOX4 achieved the highest prediction result, with an F1 score of 0.84. The remaining three combinations showed F1 scores of 0.77 (RBP4 + CXCL10 and CXCL10 + NOX4) or 0.73 (RBP4 + CXCL10 + NOX4). The results of the comparative analysis by algorithm between the normal and MCI groups (Table 4) and the SVM algorithm analysis between the normal and MCI groups (Table 3) show that the genetic combination of RBP4 and NOX4 is the most effective among all combinations. To predict outcomes in the normal and MCI groups, receiver operating characteristic (ROC) curves were utilized based on animal plasma analysis (Fig. 5a).

[0136]

[0137]

[0138]

[0139]

[0140] Example 7. SCI Prediction Using Machine Learning

[0141]

[0142] Comparing the results of the three types of combinations that formed combinations of two genes, the results of the comparative analysis by algorithm between the normal group and the SCI group (Table 6) and the SVM algorithm analysis between the normal group and the SCI group (Table 6) showed that the combination of RBP4 and NOX4 showed the highest result (F1 score 0.73), and the combination of RBP4 and CXCL10 showed the lowest result (F1 score 0.59) (Table 5). In addition, when all three genes were used, Extra Tree showed the best analysis result among the seven algorithms (F1 score 0.75) (Table 5). Despite the small sample size, using the combination of three genes proved to be the most effective set for distinguishing between the normal group and the SCI group. To predict the results in the normal group and the SCI group, the receiver operating characteristic (ROC) curve was utilized based on animal plasma analysis (Figure 5B).

[0143]

[0144]

[0145]

[0146]

[0147] Example 8. CCDS Prediction Using Machine Learning

[0148]

[0149] The combination of biomarkers to distinguish between the normal group and the CCDS group was analyzed by algorithm comparison between the normal group and the CCDS group (Table 8) and the SVM algorithm analysis between the normal group and the CCDS group (Table 8). The combination of RBP4 and NOX4 showed the highest F1 score of 0.81. On the other hand, the combination of CXCL10 and NOX4 showed the lowest F1 score of 0.74 (Table 7). When all three genes were utilized, the F1 score was 0.79 (Table 7). The comparative analysis of samples from the normal group and the CCDS group showed that the combination of RBP4 and NOX4 was the most effective combination for distinguishing between the two conditions. To predict the results in the normal group and the CCDS group, a receiver operating characteristic (ROC) curve was utilized based on animal plasma analysis (Fig. 5C).

[0150]

[0151]

[0152]

Claims

1. A biomarker composition for diagnosing a neurodegenerative disease, comprising at least one gene selected from the group consisting of NOX4 (NADPH oxidase 4), RBP4 (Retinol binding protein 4), and CXCL10 (CXC motif chemokine ligand 10) or a protein encoded by the gene.

2. A biomarker composition according to claim 1, characterized in that the neurodegenerative disease is at least one selected from the group consisting of cognitive dysfunction syndrome, Parkinson's disease, Alzheimer's disease, progressive supranuclear palsy, multiple system atrophy, olivary-pontine-cerebellar atrophy, Shy-Drager syndrome, striatonigral dystrophy, Huntington's disease, amyotrophic lateral sclerosis, essential tremor, corticobasal degeneration, diffuse Lewy body disease, Parkinson's-ALS-dementia complex, Niemann-Pick disease, Pick's disease, cerebral ischemia, and cerebral infarction.

3. A biomarker composition according to claim 1, characterized in that the mRNA of the gene is composed of at least one base sequence selected from the group consisting of sequence numbers 1, 3, and 5.

4. A biomarker composition according to claim 1, characterized in that the biomarker is extracted from at least one sample selected from the group consisting of blood, plasma, serum, lymph, saliva, urine, tissue, cerebrospinal fluid, ascites, or a combination thereof.

5. A biomarker composition, characterized in that, in the first paragraph, if the expression level of the biomarker is reduced compared to the normal control group, it is determined that there is a possibility of onset of a degenerative neurological disease.

6. A composition for diagnosing a neurodegenerative disease, comprising a preparation for measuring the expression level of any one or more mRNAs selected from the group consisting of NOX4, RBP4, and CXCL10, or a preparation for measuring the expression level of any one or more proteins selected from the group consisting of NADPH oxidase 4, Retinol binding protein 4, and CXC motif chemokine ligand 10.

7. A diagnostic composition according to claim 6, characterized in that the agent for measuring the mRNA expression level is an antisense oligonucleotide, primer pair, or probe that specifically binds to the mRNA.

8. A diagnostic composition, characterized in that the agent for measuring the protein expression level in paragraph 6 is an antibody or an antigen-binding fragment thereof, a ligand, a PNA (Peptide nucleic acid) or an aptamer that specifically binds to the protein.

9. A kit for diagnosing a degenerative neurological disease comprising the composition of paragraph 1 or paragraph 6.

10. A kit according to claim 9, characterized in that the kit is an RT-PCR kit, a competitive RT-PCR kit, a real-time RT-PCR kit, a DNA chip kit, or a protein chip kit.

11. A method for providing information for diagnosing a neurodegenerative disease, comprising the following steps: (a) a step of measuring the expression level of a gene consisting of NOX4 (NADPH oxidase 4), RBP4 (Retinol binding protein 4) and CXCL10 (CXC motif chemokine ligand 10) or a protein encoded by the gene in a biological sample isolated from an object; and (b) A step of comparing the expression level measured in step (a) with a biological sample isolated from a normal individual.

12. A method for providing information, characterized in that in paragraph 11, the sample is at least one selected from the group consisting of blood, plasma, serum, lymph, saliva, urine, tissue, cerebrospinal fluid, ascites, or a combination thereof.

13. A method for providing information, characterized in that the measurement in claim 11 is performed by electrophoresis, immunoblotting, enzyme-linked immunosorbent assay, polymerase chain reaction, Northern blotting, protein chip, immunoprecipitation, microarray, electron microscopy, Western blot, radioimmunoassay, radioimmunodiffusion, oakteroniimmunodiffusion, rocket immunoelectrophoresis, immunohistochemical staining, immunoprecipitation assay, complement fixation assay, immunofluorescence, immunochromatography, FACS analysis, or a combination thereof.

14. A method for providing information, wherein the method further comprises a step of determining that there is a possibility of developing a degenerative neurological disease if the expression level of the gene is reduced compared to a normal individual after step (b).

15. A method for providing information, comprising, in paragraph 11, a step of generating a model for predicting the onset of a neurodegenerative disease by teaching a machine learning algorithm model the expression level of the gene after step (b).

16. An information providing method according to claim 15, wherein the algorithm model is at least one algorithm model selected from the group consisting of Random Forest, Support Vector Machine (SVM), Adaboost, Bagging, Gradient Boost, Extra Tree, and XgBoost.

Images