Auxiliary diagnosis platform from mild cognitive impairment to Alzheimer disease

Abstract

Computer-implemented methods (CIM) and / or computer-implemented systems (CIS) that allow for early detection and / or prediction of the risk of developing a neurodegenerative disorder are disclosed. The CIM and / or CIS utilizes conventional test data of one or more biological fluids and a machine learning model to predict the risk of developing Alzheimer's disease in a subject exhibiting one or more signs or symptoms of mild cognitive impairment. Platforms of Mild Cognitive Impairment (MCI) to Alzheimer's Disease (AD) aided diagnosis utilizing these disclosed CIM and / or CIS, i.e., called MAP, have been developed. Thus, the platform analyzes the results of conventional test data from one or more biological fluids of a subject of interest to determine the presence of certain biomarkers identified herein as important in predicting progression from MCI to AD.

Description

Technical Field

[0001] The present invention is in the field of assessing an individual's risk of developing a neurodegenerative or non-neurodegenerative disorder, and in particular, using a discriminative artificial intelligence platform to predict an individual's risk of progressing from mild cognitive impairment to Alzheimer's disease by analyzing test results of one or more biological fluids (e.g., blood). Background Art

[0002] Alzheimer's disease (AD) is the most common cause of dementia in the elderly, characterized by a gradual decline in memory and cognitive abilities. As the aging population increases worldwide, AD is becoming a major global health problem. a By 2050, AD is projected to affect more than 100 million people, with approximately 4.6 million new cases diagnosed each year. a According to a 2020 report from the Alzheimer's Association, the total annual cost per AD subject is approximately $46,786, and the global cost of the disease is estimated to exceed $2 trillion by 2030. a Unfortunately, there are no effective drugs or interventions that can halt or reverse disease progression. This is in part because treatment is typically initiated after the onset of dementia symptoms, when AD pathology has already progressed. Therefore, early diagnosis and treatment of AD are crucial for successful prevention and treatment strategies.

[0003] Mild cognitive impairment (MCI) has attracted clinical and research interest in recent years, with the hypothesis that it represents a "transition zone" between normal cognition and dementia. a Subjects with MCI exhibit objective cognitive impairment but normal ability to perform daily activities. Subjects with MCI are at increased risk of developing AD, with an annual conversion rate of 10-30%, compared to 1-2% in unaffected individuals. a It is believed that intervention at the MCI stage will increase the likelihood of successful AD treatment. However, not all MCI subjects will progress to AD, and some cases may even revert to normal aging. a Therefore, it is crucial to identify those individuals with presymptomatic MCI who will eventually develop Alzheimer's disease to generate substantial treatment and health economic benefits.

[0004] As with AD diagnosis, current diagnostic methods for MCI are mainly based on neuropsychological tests and neuroimaging techniques, supplemented by biomarkers in cerebrospinal fluid (CSF) and blood (such as Aβ42, tau, and phosphorylated tau). Many predictive models for the conversion of MCI to AD have been developed using the results of these methods. a. However, these techniques have several limitations. First, clinical tests show low specificity and neuroimaging methods are expensive. Although CSF biomarkers have important diagnostic utility, detection requires an invasive lumbar puncture. In addition, the measurement of blood Aβ peptides and tau is challenging because their concentrations in the blood are extremely low, and their application in large-scale diagnostic and prognostic trials is limited by the need for sensitive and specific analytical methods.

[0005] Previous studies have demonstrated the potential of using routine peripheral blood parameters to predict AD risk, as many parameters are altered in AD subjects and some factors are associated with cognitive decline. Machine learning has been widely used in various fields, including assisting medical diagnosis. Recently, some efforts have been made to create AD risk prediction models using machine learning methods, which can obtain more significant and accurate results than traditional methods.

[0006] Given the easy availability and cost-effectiveness of routine blood tests, routine blood analytes could complement neuropsychological testing as a screening tool before proceeding with neuroimaging testing and measuring CSF / blood biomarkers to predict future cognitive decline in subjects with MCI. However, no studies have investigated the potential for predicting the risk of conversion from MCI to AD using routine laboratory analytes.

[0007] Therefore, it is an object of the present invention to construct and provide a discriminative artificial intelligence (AI) platform to achieve improved prediction of the risk of developing neurodegenerative disorders.

[0008] Another object of the present invention is to construct and provide a discriminative artificial intelligence (AI) platform to achieve improved prediction of risk of developing neurodegenerative disorders based on analysis of test results of one or more biological fluids (e.g., blood).

[0009] Another object of the present invention is to construct and provide a discriminative AI platform comprising an integration of machine learning models to achieve improved prediction of the risk of developing neurodegenerative disorders.

[0010] Another object of the present invention is to construct and provide a discriminative AI platform comprising an integration of machine learning models based on the analysis of test results of one or more biological fluids (e.g., blood) to achieve improved risk prediction for developing neurodegenerative disorders. Summary of the Invention

[0011] The subject matter of the present disclosure relates to a computer-implemented method (CIM) and / or a computer-implemented system (CIS) that allows for early detection and / or prediction of the risk of developing a neurodegenerative disorder (e.g., Alzheimer's disease). In particular, the CIM and / or CIS utilize conventional test data of one or more biological fluids and a machine learning model to predict the risk of a subject exhibiting one or more signs or symptoms of mild cognitive impairment developing Alzheimer's disease, i.e., predicting the risk of mild cognitive impairment progressing to Alzheimer's disease. Preferred biological fluids include blood and / or cerebrospinal fluid. More preferably, the biological fluid is blood.

[0012] A platform utilizing these disclosed CIMs and / or CISs has been developed, namely a platform for auxiliary diagnosis of mild cognitive impairment (MCI) to Alzheimer's disease (AD) called MAP. Thus, the platform analyzes the results of routine test data from one or more biological fluids (e.g., blood) of a subject of interest to determine the presence of certain biomarkers. Exemplary biomarkers include biomarkers associated with liver function, biomarkers associated with kidney function, minerals and proteins, blood cells, and the like. In some forms, the biomarkers analyzed using the platform are biomarkers associated with oxygen carrying capacity and immune function, such as hemoglobin, hematocrit, and red blood cells associated with oxygen carrying capacity; and neutrophils and leukocytes associated with immunity.

[0013] For example, biomarkers suitable for analysis using the platform include, but are not limited to, biomarkers associated with liver function (e.g., alanine aminotransferase (ALT), alkaline phosphatase (ALP), and bilirubin), biomarkers associated with kidney function (e.g., creatinine and urea), minerals and proteins (potassium, sodium, calcium, phosphate, albumin, globulin, and total protein), and blood cells (eosinophils (absolute), basophils (absolute), neutrophils (absolute), lymphocytes (absolute), monocytes (absolute), eosinophils (%), basophils (%), neutrophils (%), lymphocytes (%), monocytes (%), white blood cells (WBC), platelets, red blood cells (RBC), mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), hematocrit (HCT), hemoglobin, and red cell distribution width (RDW)), and combinations thereof.

[0014] In some forms, the platform analyzes results from routine test data for five biomarkers, including hemoglobin, hematocrit, neutrophils (absolute), red blood cells, and white blood cells.

[0015] In some forms, the platform analyzes results from routine test data for 19 biomarkers including ALT, ALP, bilirubin, potassium, calcium, phosphate, total protein, albumin, basophils (absolute), neutrophils (absolute), lymphocytes (absolute), basophils (%), neutrophils (%), lymphocytes (%), WBC, RBC, MCHC, HCT, and hemoglobin.

[0016] In some forms, the platform analyzes results from routine test data for 31 biomarkers including ALT, ALP, bilirubin, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, globulin, total protein, eosinophils (absolute), basophils (absolute), neutrophils (absolute), lymphocytes (absolute), monocytes (absolute), eosinophils (%), basophils (%), neutrophils (%), lymphocytes (%), monocytes (%), WBC, platelets, RBC, MCV, MCH, MCHC, HCT, hemoglobin, and RDW.

[0017] In some forms, the platform analyzes results from routine test data for 23 biomarkers including ALP, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, total protein, basophils (absolute), neutrophils (absolute), eosinophils (absolute), lymphocytes (absolute), monocytes (absolute), basophils (%), monocytes (%), WBC, platelets, MCV, MCH, MCHC, hemoglobin, and RDW (%).

[0018] Optionally, the specific biomarkers analyzed using the platform are selected based on the sex and age of the subject of interest. For example, for female subjects aged 65 to 74 years old, the 5-biomarker panel described above is analyzed using the platform. For female subjects aged 75 to 89 years old, the 19-biomarker panel described above is analyzed using the platform. For male subjects aged 65 to 74 years old, the 31-biomarker panel described above is analyzed using the platform. For male subjects aged 75 to 89 years old, the 23-biomarker panel described above is analyzed using the platform.

[0019] Optionally, the biomarkers analyzed using the platform are universal, regardless of the sex and age of the subject of interest. For example, a 5-biomarker panel including hemoglobin, hematocrit, neutrophils (absolute), red blood cells, and white blood cells is analyzed using the platform for female and male subjects aged 65 to 89 years.

[0020] The platform leverages the understanding of the differences between these biomarkers in subjects with mild cognitive impairment and Alzheimer's disease to more reliably predict and / or monitor the health of neural tissue (e.g., brain) from test results in one or more biofluids (e.g., blood) before the onset of Alzheimer's disease or advanced forms of the disease. Therefore, the platform has significant implications for early intervention and improved outcomes.

[0021] The platform provides information about the health of neural tissue based on the assessment of biomarkers (such as one or more biomarkers mentioned above) in one or more biological fluids, providing a cost-effective solution for assessing the risk of developing neurodegenerative disorders, which is currently missing in the field of neurodegenerative diagnosis. Therefore, complex neurodegenerative disorders such as Alzheimer's disease that are currently diagnosed very late (i.e., in the late symptomatic stage) can be diagnosed or otherwise identified before the onset of late symptoms. The platform also allows early testing of candidate drugs for clinical trials that previously failed due to poor patient selection, late intervention, and very high trial costs. These factors can be significantly improved using the platform.

[0022] The CIM and / or CIS of the present disclosure can be used to determine whether biomarkers in one or more biological fluids are indicative of a neurodegenerative disorder, such as Alzheimer's disease. The CIM and / or CIS can facilitate early diagnosis (e.g., prediction and / or detection) of a neurodegenerative disorder (e.g., Alzheimer's disease) much earlier than possible using previous methods and systems, such as many years before the development of symptoms associated with the disorder that would be detectable using previous methods and systems. The CIM and / or CIS system has accuracy (e.g., greater than 60% accuracy) in diagnosing a neurodegenerative disorder (e.g., Alzheimer's disease) as determined by one or more criteria described herein.

[0023] A CIM for predicting and / or detecting a neurodegenerative condition in a subject involves (i) employing one or more discriminative artificial intelligence (AI) platforms to perform an analysis of one or more test results of one or more biological fluids, preferably wherein the one or more discriminative AI platforms are operably connected to a graphical user interface (GUI) or a sound transmitter, and (ii) providing a prediction of a neurodegenerative condition on the GUI or the sound transmitter based on the analysis. At least one of the one or more discriminative AI platforms (i) has been trained on biomarker data classified by sex and / or age, wherein the biomarker data is obtained from an analysis of a biological fluid; and / or (ii) performs a prediction of a neurodegenerative condition taking into account the sex and / or age of the subject. Preferably, at least one of the one or more discriminative AI platforms has been trained by a human.

[0024] The CIS comprises one or more discriminative AI platforms comprising an integration of machine learning models. At least one of the one or more discriminative AI platforms (i) has been trained on biomarker data categorized by sex and / or age, wherein the biomarker data is obtained from analysis of biological fluids; and / or (ii) is capable of predicting a subject's risk of developing a neurodegenerative disorder, taking into account the subject's sex and / or age. Preferably, at least one of the one or more discriminative AI platforms has been trained by a human.

[0025] In some forms, the CIM and / or CIS comprises one or more discriminative AI platforms comprising an ensemble of machine learning models, the ensemble of machine learning models consisting of at least two or at least three machine learning models selected from logistic regression, Gaussian naive Bayes, random forest, gradient boosting, adaptive boosting, extreme random trees, linear discriminant analysis, and support vector machines. In preferred forms, the ensemble of machine learning models comprises: (i) logistic regression, Gaussian naive Bayes, and random forest; (ii) logistic regression, gradient boosting, and extreme random trees; (iii) linear discriminant analysis, support vector machine, extreme random trees; and / or (iv) random forest, adaptive boosting, extreme random trees.

[0026] Data show that an Alzheimer's disease prediction model trained with routine blood test data and machine learning algorithms can provide accurate predictions of the risk of conversion from mild cognitive impairment to Alzheimer's disease. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1A and 1B This is a flowchart showing the data processing steps involved in generating the data. Abbreviations: HADCL, Hospital Authority Data Collaborative Laboratory; MCI, mild cognitive impairment; AD, Alzheimer's disease; female 65-74, female 65-74 years; male 65-74, male 65-74 years; female 75-89, female 75-89 years; male 75-89, male 75-89 years; train, data distribution of the training set; test, independent test set used to validate model performance.

[0028] Figure 2A and 2BThis diagram shows the framework of a discriminative AI-based auxiliary diagnosis platform (MAP) for mild cognitive impairment (MCI) to Alzheimer's disease (AD). Abbreviations: MCI, mild cognitive impairment; AD, Alzheimer's disease; threshold to screen out biomarkers present in more than 90% of the 48,116 patients; IQR, interquartile range; T-test, Student's t-test; FFS, forward feature selection; MRMD, maximum relevance maximum distance; LR, logistic regression; GNB, Gaussian naive Bayes; RF, random forest; GB, gradient boosting; ET, extreme randomized trees; LDA, linear discriminant analysis; SVM, support vector machine; ADA, adaptive boosting; MAP, auxiliary diagnosis platform for mild cognitive impairment to Alzheimer's disease.

[0029] Figure 3 Depicts a bar graph showing the sensitivity and specificity of the model on cross-validation and independent test sets. Abbreviations: CV, cross-validation; SN, sensitivity; SP, specificity; F 65-74 ML model, machine learning model for females aged 65-74; M 65-74 ML model, machine learning model for males aged 65-74; F 75-89 ML model, machine learning model for females aged 75-89; M 75-89 ML model, machine learning model for males aged 75-89; TabNet, tabular attention network model.

[0030] Figure 4 A set of line graphs showing the area under the curve (AUC) performance on cross-validation (CV) and independent datasets are depicted. Abbreviations: ROC, receiver operating characteristic; ML model F 65-74, machine learning model for females aged 65-74; ML model M 65-74, machine learning model for males aged 65-74; ML model F 75-89, machine learning model for females aged 75-89; ML model M 75-89, machine learning model for males aged 75-89; TabNet, tabular attention network model. DETAILED DESCRIPTION

[0031] I. Definition

[0032] "Clinician" means a healthcare professional working under the auspices of a healthcare organization to provide care to subjects in need. As used herein, "clinician" includes surgeons, nurses, nurse practitioners, physician assistants, physical therapists, or other licensed or specially trained healthcare practitioners.

[0033] "Discriminative artificial intelligence" or "discriminative AI" refers to a set of machine learning techniques / algorithms that typically make predictions or inferences based on analysis of input data. Discriminative AI platforms are also known as predictive AI.

[0034] An "ensemble" in relation to machine learning models refers to the presence of two or more of these models. Examples of an ensemble may include 2, 3, 4, 5, 6, 7, 8, 9, or 10 of these models.

[0035] “Operably connected” refers to the connection of at least two components in the CIS via a technology including, but not limited to, Ethernet, Bluetooth, near field communication, WiFi, integrated circuits, or a combination thereof.

[0036] II. Computer-Implemented Methods and Systems and Biomarkers

[0037] Described herein are computer-implemented methods (CIMs) and / or computer-implemented systems (CISs) that allow early detection and / or prediction of the risk of developing neurodegenerative or non-neurodegenerative disorders. CIMs and / or CISs utilize conventional test data and machine learning models of one or more biofluids to predict the risk of developing neurodegenerative or non-neurodegenerative disorders. In the case of neurodegenerative disorders such as Alzheimer's disease involving cognitive decline, CIMs and / or CISs detect and / or predict the risk of developing neurodegenerative disorders in subjects who exhibit one or more signs or symptoms of mild cognitive impairment, i.e., predict the risk of mild cognitive impairment progressing to Alzheimer's disease.

[0038] Also disclosed is a platform utilizing these disclosed CIMs and / or CIS, namely a mild cognitive impairment (MCI) to Alzheimer's disease (AD) auxiliary diagnostic platform denoted as MAP, which is used to detect and / or predict the risk of developing neurodegenerative disorders. The platform analyzes the results of routine test data from one or more biological fluids to determine the presence and / or amount of certain biomarkers. Exemplary biomarkers generally include biomarkers associated with liver function, biomarkers associated with kidney function, minerals and proteins, blood cells, etc. In some forms, the biomarkers analyzed using the platform are biomarkers associated with oxygen carrying capacity and immune function, such as hemoglobin, hematocrit, and red blood cells associated with oxygen carrying capacity; and neutrophils and white blood cells associated with immunity.

[0039] For example, biomarkers suitable for analysis using the platform include, but are not limited to, biomarkers associated with liver function (e.g., alanine aminotransferase (ALT), alkaline phosphatase (ALP), and bilirubin), biomarkers associated with kidney function (e.g., creatinine and urea), minerals and proteins (potassium, sodium, calcium, phosphate, albumin, globulin, and total protein), and blood cells (eosinophils (absolute), basophils (absolute), neutrophils (absolute), lymphocytes (absolute), monocytes (absolute), eosinophils (%), basophils (%), neutrophils (%), lymphocytes (%), monocytes (%), white blood cells (WBC), platelets, red blood cells (RBC), mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), hematocrit (HCT), hemoglobin, and red cell distribution width (RDW)), and combinations thereof.

[0040] In some forms, the platform analyzes results from routine test data for five biomarkers, including hemoglobin, hematocrit, neutrophils (absolute), red blood cells, and white blood cells.

[0041] In some forms, the platform analyzes results from routine test data for 19 biomarkers including ALT, ALP, bilirubin, potassium, calcium, phosphate, total protein, albumin, basophils (absolute), neutrophils (absolute), lymphocytes (absolute), basophils (%), neutrophils (%), lymphocytes (%), WBC, RBC, MCHC, HCT, and hemoglobin.

[0042] In some forms, the platform analyzes results from routine test data for 31 biomarkers including ALT, ALP, bilirubin, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, globulin, total protein, eosinophils (absolute), basophils (absolute), neutrophils (absolute), lymphocytes (absolute), monocytes (absolute), eosinophils (%), basophils (%), neutrophils (%), lymphocytes (%), monocytes (%), WBC, platelets, RBC, MCV, MCH, MCHC, HCT, hemoglobin, and RDW.

[0043] In some forms, the platform analyzes results of routine test data for 23 biomarkers including ALP, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, total protein, basophils (absolute), neutrophils (absolute), eosinophils (absolute), lymphocytes (absolute), monocytes (absolute), basophils (%), monocytes (%), WBC, platelets, MCV, MCH, MCHC, hemoglobin, and RDW (%).

[0044] Optionally, the specific biomarkers analyzed using the platform are selected based on the sex and age of the subject of interest. For example, for female subjects aged 65 to 74 years old, the 5-biomarker panel described above is analyzed using the platform. For female subjects aged 75 to 89 years old, the 19-biomarker panel described above is analyzed using the platform. For male subjects aged 65 to 74 years old, the 31-biomarker panel described above is analyzed using the platform. For male subjects aged 75 to 89 years old, the 23-biomarker panel described above is analyzed using the platform.

[0045] Optionally, the biomarkers analyzed using the platform are universal, regardless of the sex and age of the subject of interest. For example, a 5-biomarker panel including hemoglobin, hematocrit, neutrophils (absolute), red blood cells, and white blood cells is analyzed using the platform for female and male subjects aged 65 to 89 years.

[0046] In cases where neurodegenerative disorders involve cognitive decline, the platform leverages understanding of the differences between biomarkers in subjects with mild cognitive impairment and neurodegenerative disorders to more reliably predict and / or monitor the health of neural tissue (e.g., brain) from test results of one or more biofluids before the onset of a neurological disorder. Therefore, the platform has significant implications for early intervention and improved outcomes.

[0047] The platform can provide detailed information about tissue health based on the assessment of biomarkers (such as one or more biomarkers mentioned above, such as hemoglobin, hematocrit, neutrophils (absolute), red blood cells and white blood cells) in one or more biofluids, providing a cost-effective solution for assessing the risk of developing neurodegenerative or non-neurodegenerative diseases. Therefore, neurodegenerative or non-neurodegenerative diseases currently undergoing very late diagnosis (that is, in the late symptomatic stage) can be diagnosed or otherwise identified before the onset of late symptoms. The platform also allows early testing of candidate drugs for clinical trials, which previously failed due to poor patient selection, late intervention and very high trial costs. The platform can also allow early intervention to improve neurodegenerative or non-neurodegenerative diseases. These factors can be significantly improved using the platform.

[0048] The machine learning program can relate to various supervised machine learning techniques, various semi-supervised machine learning techniques and / or various unsupervised machine learning techniques. For example, the machine learning program can utilize logistic regression, Gaussian Naive Bayes, random forest, gradient boosting, adaptive boosting, LPBoost, TotalBoost, BrownBoost, MadaBoost, LogitBoost, extreme random trees, linear discriminant analysis, support vector machine, decision tree, k- nearest neighbor, alternating decision tree (ADTree), decision tree stump, function tree (FT), logistic model tree (LMT), linear classifier, factor analysis, principal component analysis, neighborhood component analysis, sparse filtering, random neighbor embedding, autoencoder, stacked autoencoder, neural network, convolutional neural network, feedforward neural network, table attention network or any other machine learning algorithm or statistical algorithm. One or more models can be used together to generate an integrated method, wherein a machine learning integrated meta-algorithm such as boosting algorithm (boosting) (for example, AdaBoost, LPBoost, TotalBoost, BrownBoost, MadaBoost, LogitBoost etc.) can be used to optimize the integrated method to reduce bias and / or variance. For example, machine learning analysis can be performed using one or more of a variety of programming languages ​​and platforms (e.g., R, Weka, Python, and / or Matlab). Machine learning analysis can be performed using a machine learning platform (e.g., BigML).

[0049] CIM and / or CIS can be used to determine whether biomarkers (e.g., hemoglobin, hematocrit, neutrophils (absolute), red blood cells, and white blood cells) in one or more biological fluids are indicative of a neurodegenerative or non-neurodegenerative condition. CIM and / or CIS can facilitate early diagnosis (e.g., prediction and / or detection) of a neurodegenerative or non-neurodegenerative condition much earlier than is possible using previous methods and systems, such as many years before the onset of symptoms associated with the condition that would be detectable using previous methods and systems. The CIM and / or CIS system has accuracy (e.g., greater than 60% accuracy) in diagnosing a neurodegenerative or non-neurodegenerative condition as determined by one or more criteria described herein.

[0050] i. Computer-implemented methods

[0051] In some forms, a CIM for detecting and / or predicting a neurodegenerative or non-neurodegenerative condition in a subject comprises: (i) employing one or more discriminative AI platforms to perform analysis of one or more test results of one or more biological fluids, and (iii) providing a determination and / or prediction of the neurodegenerative or non-neurodegenerative condition on a graphical user interface (GUI) or sound transmitter based on the analysis. Preferably, the one or more discriminative AI platforms are operably connected to the GUI or sound transmitter. Preferably, the one or more test results are obtained from one or more tests ordered by a clinician. In some forms of the CIM, at least one of the one or more discriminative AI platforms: (i) has been trained on biomarker data categorized by sex and / or age, wherein the biomarker data is obtained from analysis of biological fluids; and / or (ii) performs predictions of the neurodegenerative or non-neurodegenerative condition while taking into account the sex and / or age of the subject. Preferably, at least one of the one or more discriminative AI platforms has been trained by a human.

[0052] Detection and / or prediction of neurodegenerative or non-neurodegenerative conditions can be achieved with an accuracy of greater than or equal to 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95% or 99%. In some forms, detection and / or prediction of neurodegenerative or non-neurodegenerative conditions can be based on analysis of one or more biological fluids from a single subject. In other forms, detection and / or prediction of neurodegenerative or non-neurodegenerative conditions can be based on analysis of one or more biological fluids from multiple subjects.

[0053] The condition can be a neurodegenerative condition. The condition can be selected from the group consisting of Alzheimer's disease, non-Alzheimer's dementia, Parkinson's disease, parkinsonism, motor neuron disease, Huntington's disease, Huntington's disease-like syndrome, transmissible spongiform encephalopathy, chronic traumatic encephalopathy, and tauopathy. The condition can be a non-neurodegenerative condition. The condition can be selected from the group consisting of primary tumors, metastatic tumors, epileptic seizures, epileptic seizures with focal cortical dysplasia, demyelinating conditions, non-neurodegenerative brain diseases, cerebrovascular diseases, and psychological conditions. In a preferred form, the condition is a neurodegenerative condition, and the CIM involves providing a prediction of the neurodegenerative condition on a GUI or sound transmitter.

[0054] The CIM can help predict a neurodegenerative or non-neurodegenerative disorder between one year and 10 years before the development of the neurodegenerative or non-neurodegenerative disorder. The CIM can also help predict a neurodegenerative or non-neurodegenerative disorder between one year and 10 years before the development of symptoms associated with the neurodegenerative or non-neurodegenerative disorder. The CIM can also help monitor a neurodegenerative or non-neurodegenerative disorder at multiple time points, the multiple time points being separated by multiple time intervals.

[0055] The CIM comprises one or more discriminative AI platforms comprising an ensemble of machine learning models, a deep learning model, an ensemble of deep learning models, or a combination thereof. Preferably, the models in the ensemble or combination are weighted with a weight between 1 and 10 or between 1 and 5, for example, 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10.

[0056] In some forms of the CIM, one or more discriminative AI platforms comprise an ensemble of machine learning models comprising at least two or at least three machine learning models selected from the group consisting of logistic regression, Gaussian naive Bayes, random forest, gradient boosting, adaptive boosting, LPBoost, TotalBoost, BrownBoost, MadaBoost, LogitBoost, extreme random trees, linear discriminant analysis, support vector machine, decision tree, and k-nearest neighbor. In some forms, one or more discriminative AI platforms comprise an ensemble of machine learning models comprising at least two or at least three machine learning models selected from the group consisting of logistic regression, Gaussian naive Bayes, random forest, gradient boosting, adaptive boosting, extreme random trees, linear discriminant analysis, and support vector machine. In some forms, one or more discriminative AI platforms comprise an ensemble of machine learning models comprising: (i) logistic regression, Gaussian naive Bayes, and random forest; (ii) logistic regression, gradient boosting, and extreme random trees; (iii) linear discriminant analysis, support vector machine, and extreme random trees; and / or (iv) random forest, adaptive boosting, and extreme random trees.

[0057] In some forms of the CIM, one or more discriminative AI platforms comprise a deep learning model selected from a neural network, a convolutional neural network, a feedforward neural network, a tabular attention network, an autoencoder, or a stacked autoencoder.

[0058] As discussed herein, analysis of one or more test results involves detecting the presence and / or amount of biomarkers in one or more biological fluids. Optionally, analysis of one or more test results involves detecting the presence and / or amount of biomarkers, including biomarkers associated with liver function, biomarkers associated with kidney function, minerals and proteins, and blood cells, such as those associated with oxygen carrying capacity and immune function.

[0059] In some forms, analysis of one or more test results involves detecting the presence and / or amount of biomarkers including hemoglobin, hematocrit, neutrophils (absolute), red blood cells, and white blood cells.

[0060] In some forms, analysis of one or more test results involves detecting the presence and / or amount of biomarkers including ALT, ALP, bilirubin, potassium, calcium, phosphate, total protein, albumin, basophils (absolute), neutrophils (absolute), lymphocytes (absolute), basophils (%), neutrophils (%), lymphocytes (%), WBC, RBC, MCHC, HCT, and hemoglobin.

[0061] In some forms, analysis of one or more test results involves detecting the presence and / or amount of biomarkers comprising ALT, ALP, bilirubin, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, globulin, total protein, eosinophils (absolute), basophils (absolute), neutrophils (absolute), lymphocytes (absolute), monocytes (absolute), eosinophils (%), basophils (%), neutrophils (%), lymphocytes (%), monocytes (%), WBC, platelets, RBC, MCV, MCH, MCHC, HCT, hemoglobin, and RDW.

[0062] In some forms, analysis of one or more test results involves detecting the presence and / or amount of biomarkers including ALP, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, total protein, basophils (absolute), neutrophils (absolute), eosinophils (absolute), lymphocytes (absolute), monocytes (absolute), basophils (%), monocytes (%), WBC, platelets, MCV, MCH, MCHC, hemoglobin, and RDW (%).

[0063] In some forms, the analysis of the one or more test results involves monitoring changes in one or more biomarkers in one or more biological fluids of the subject of interest over a suitable time period. For example, one or more biomarkers in one or more biological fluids (e.g., blood) of the subject of interest are measured over a period of six months, one year, two years, three years, five years, or ten years, such as one or more of the biomarkers described above (e.g., hemoglobin, hematocrit, neutrophils (absolute), red blood cells, and white blood cells) once, twice, three times, five times, or ten times to obtain a trend in the test results for analysis. In the case of a neurodegenerative disorder involving cognitive decline, the prediction of a neurodegenerative disorder involves predicting the risk of mild cognitive impairment progressing to a neurodegenerative disease.

[0064] Preferably, the one or more biological fluids used to perform the one or more tests are extracellular fluids. Examples of extracellular fluids include, but are not limited to, blood, cerebrospinal fluid (CSF), serum, lymph, urine, interstitial fluid, amniotic fluid, peritoneal fluid, and combinations thereof.

[0065] ii. Computer-implemented systems

[0066] Also described herein is a CIS comprising one or more discriminant AI platforms comprising an integration of machine learning models, a deep learning model, an integration of deep learning models, or a combination thereof. Preferably, one or more discriminant AI platforms are operably connected to a graphical user interface (GUI) or a sound transmitter. Preferably, the models in the integration or combination are weighted with a weight between 1 and 10 or between 1 and 5, such as 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10. At least one of the one or more discriminant AI platforms: (i) has been trained on biomarker data classified by sex and / or age, wherein the biomarker data is obtained from an analysis of a biological fluid; and / or (ii) is capable of predicting the risk of a subject developing a neurodegenerative or non-neurodegenerative disorder, taking into account the sex and / or age of the subject. Preferably, at least one of the one or more discriminant AI platforms has been trained by a person.

[0067] The CIS facilitates analysis of one or more test results involving detection of the presence and / or amount of biomarkers in one or more biological fluids. Optionally, analysis of the one or more test results involves detection of the presence and / or amount of biomarkers including biomarkers associated with liver function, biomarkers associated with kidney function, minerals and proteins, and blood cells, such as those associated with oxygen carrying capacity and immune function.

[0068] In some forms, analysis of one or more test results involves detecting the presence and / or amount of biomarkers including hemoglobin, hematocrit, neutrophils (absolute), red blood cells, and white blood cells.

[0069] In some forms, analysis of one or more test results involves detecting the presence and / or amount of biomarkers including ALT, ALP, bilirubin, potassium, calcium, phosphate, total protein, albumin, basophils (absolute), neutrophils (absolute), lymphocytes (absolute), basophils (%), neutrophils (%), lymphocytes (%), WBC, RBC, MCHC, HCT, and hemoglobin.

[0070] In some forms, analysis of one or more test results involves detecting the presence and / or amount of biomarkers comprising ALT, ALP, bilirubin, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, globulin, total protein, eosinophils (absolute), basophils (absolute), neutrophils (absolute), lymphocytes (absolute), monocytes (absolute), eosinophils (%), basophils (%), neutrophils (%), lymphocytes (%), monocytes (%), WBC, platelets, RBC, MCV, MCH, MCHC, HCT, hemoglobin, and RDW.

[0071] In some forms, analysis of one or more test results involves detecting the presence and / or amount of biomarkers including ALP, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, total protein, basophils (absolute), neutrophils (absolute), eosinophils (absolute), lymphocytes (absolute), monocytes (absolute), basophils (%), monocytes (%), WBC, platelets, MCV, MCH, MCHC, hemoglobin, and RDW (%).

[0072] In some forms of the CIS, one or more discriminative AI platforms comprise an ensemble of machine learning models comprising at least two or at least three machine learning models selected from logistic regression, Gaussian naive Bayes, random forest, gradient boosting, adaptive boosting, LPBoost, TotalBoost, BrownBoost, MadaBoost, LogitBoost, extreme random trees, linear discriminant analysis, support vector machine, decision tree, and k-nearest neighbor. In some forms, one or more discriminative AI platforms comprise an ensemble of machine learning models comprising at least two or at least three machine learning models selected from logistic regression, Gaussian naive Bayes, random forest, gradient boosting, adaptive boosting, extreme random trees, linear discriminant analysis, and support vector machine. In some forms, one or more discriminative AI platforms comprise an ensemble of machine learning models comprising: (i) logistic regression, Gaussian naive Bayes, and random forest; (ii) logistic regression, gradient boosting, and extreme random trees; (iii) linear discriminant analysis, support vector machine, and extreme random trees; and / or (iv) random forest, adaptive boosting, and extreme random trees.

[0073] In some forms of the CIS, one or more discriminative AI platforms comprise a deep learning model selected from a neural network, a convolutional neural network, a feedforward neural network, a tabular attention network, an autoencoder, or a stacked autoencoder.

[0074] The one or more discriminant AI platforms are capable of detecting and / or predicting the risk of a subject developing a neurodegenerative or non-neurodegenerative condition, taking into account the sex and / or age of the subject, by analyzing one or more test results of one or more biological fluids. Preferably, the one or more biological fluids are extracellular fluids. Examples of extracellular fluids include, but are not limited to, blood, cerebrospinal fluid (CSF), serum, lymph, urine, interstitial fluid, amniotic fluid, peritoneal fluid, and combinations thereof.

[0075] Detection and / or prediction of neurodegenerative or non-neurodegenerative conditions can be achieved with an accuracy of greater than or equal to 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95% or 99%. In some forms, detection and / or prediction of neurodegenerative or non-neurodegenerative conditions can be based on analysis of one or more biological fluids from a single subject. In other forms, detection and / or prediction of neurodegenerative or non-neurodegenerative conditions can be based on analysis of one or more biological fluids from multiple subjects.

[0076] The disease can be a neurodegenerative disease. The disease can be selected from: Alzheimer's disease, non-Alzheimer's disease dementia, Parkinson's disease, Parkinson's syndrome disease, motor neuron disease, Huntington's disease, Huntington's disease-like syndrome, transmissible spongiform encephalopathy, chronic traumatic encephalopathy and tauopathy. The disease can be a non-neurodegenerative disease. The disease can be selected from: primary tumors, metastatic tumors, epileptic seizure disease, epileptic seizure disease with focal cortical dysplasia, demyelinating disease, non-neurodegenerative encephalopathy, cerebrovascular disease and psychological disorders. In a preferred form, the disease is a neurodegenerative disease, and the CIS provides a prediction of the neurodegenerative disease on the GUI or sound transmitter.

[0077] The CIS helps predict neurodegenerative or non-neurodegenerative disorders between 1 to 10 years before the development of the neurodegenerative or non-neurodegenerative disorder. The CIS can also help predict neurodegenerative or non-neurodegenerative disorders between 1 to 10 years before the development of symptoms associated with the neurodegenerative or non-neurodegenerative disorder. The CIS can also help monitor neurodegenerative or non-neurodegenerative disorders at multiple time points, the multiple time points being separated by multiple time intervals.

[0078] In the case of neurodegenerative disorders involving cognitive decline, prediction of neurodegenerative disorders involves predicting the risk of progression of mild cognitive impairment to a neurodegenerative disease.

[0079] IIII. How to use

[0080] Described CIM or CIS can be used for analyzing data, such as the result from the test to one or more types of biological fluid.More importantly, they can be used for detecting and / or assessing the risk of individual development neurodegeneration or non-neurodegenerative disease by analyzing the test result of one or more types of biological fluid.Especially, they can be used for predicting the risk of experimenter's progression to AD from MCI based on the assessment of these fluids.Preferably, CIM or CIS contribute to predicting neurodegenerative disease between 1 to 10 years before neurodegenerative disease development.CIS or CIM have general applicability, and are not limited to the test result data of the subject colony from specific geographic area of ​​the world.Preferably, data are from the test result of extracellular fluid, and described extracellular fluid includes but is not limited to blood, cerebrospinal fluid (CSF), serum, lymph, urine, interstitial fluid, amniotic fluid, peritoneal fluid and combination thereof.

[0081] The CIM and CIS of the present disclosure may be further understood through the following paragraphs or embodiments.

[0082] 1. A computer-implemented method (CIM) for predicting a neurodegenerative or non-neurodegenerative condition in a subject, the CIM comprising:

[0083] (i) employing one or more discriminative artificial intelligence (AI) platforms to perform analysis of one or more test results of one or more biological fluids, preferably wherein the one or more discriminative AI platforms are operably connected to a graphical user interface (GUI) or a sound transmitter, and

[0084] (iii) providing a prediction of the neurodegenerative or non-neurodegenerative condition on the GUI or the sound emitter based on the analysis.

[0085] 2. The CIM of paragraph 1, wherein at least one of the one or more discriminative AI platforms:

[0086] (i) has been trained on biomarker data disaggregated by sex and / or age, wherein the biomarker data is obtained from analysis of a biological fluid; and / or

[0087] (ii) prediction of a neurodegenerative or non-neurodegenerative disorder taking into account the sex and / or age of the subject.

[0088] 3. The CIM of paragraph 1 or 2, wherein the prediction of the neurodegenerative or non-neurodegenerative condition is achieved with an accuracy greater than or equal to 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95% or 99%.

[0089] 4. The CIM of any of paragraphs 1 to 3, wherein the prediction of the neurodegenerative or non-neurodegenerative condition is based on analysis of one or more biological fluids from a single subject.

[0090] 5. The CIM of any of paragraphs 1 to 3, wherein the prediction of the neurodegenerative or non-neurodegenerative condition is based on analysis of one or more biological fluids from a plurality of subjects.

[0091] 6. The CIM of any of paragraphs 1 to 5, comprising providing a prediction of a neurodegenerative disorder on a GUI or a sound emitter.

[0092] 7. The CIM of any of paragraphs 1 to 6, wherein the CIM helps predict a neurodegenerative disorder between 1 and 10 years before the development of the neurodegenerative disorder.

[0093] 8. The CIM of any of paragraphs 1 to 7, wherein the neurodegenerative disorder is selected from Alzheimer's disease, a non-Alzheimer's dementia disorder, Parkinson's disease, a parkinsonian syndrome disorder, a motor neuron disease (such as amyotrophic lateral sclerosis), Huntington's disease, a Huntington's disease-like syndrome, a transmissible spongiform encephalopathy, a chronic traumatic encephalopathy, a tauopathy (such as Pick's disease, corticobasal degeneration, progressive supranuclear palsy or Niemann-Pick disease), or any other neurodegenerative disorder.

[0094] 9. The CIM of any of paragraphs 1 to 8, wherein at least one of the one or more discriminative AI platforms comprises an ensemble of machine learning models, wherein the models in the ensemble are weighted with a weight between 1 and 10 or between 1 and 5, e.g., 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10.

[0095] 10. The CIM of any of paragraphs 1 to 9, wherein the one or more discriminative AI platforms comprise an ensemble of machine learning models, the ensemble of machine learning models comprising at least two machine learning models or at least three machine learning models selected from logistic regression, Gaussian Naive Bayes, Random Forest, Gradient Boosting, Adaptive Boosting, LPBoost, TotalBoost, BrownBoost, MadaBoost, LogitBoost, Extreme Random Trees, Linear Discriminant Analysis, Support Vector Machine, Decision Tree, and k-Nearest Neighbor.

[0096] 11. The CIM of any of paragraphs 1 to 9, wherein the one or more discriminative AI platforms comprise an ensemble of machine learning models, wherein the ensemble of machine learning models comprises at least two machine learning models or at least three machine learning models selected from logistic regression, Gaussian naive Bayes, random forest, gradient boosting, adaptive boosting, extreme random trees, linear discriminant analysis, and support vector machine.

[0097] 12. The CIM of any of paragraphs 1 to 11, wherein the one or more discriminative AI platforms comprise an ensemble of machine learning models comprising:

[0098] (i) Logistic regression, Gaussian naive Bayes, and random forest;

[0099] (ii) Logistic regression, gradient boosting, and extreme randomized trees;

[0100] (iii) Linear Discriminant Analysis, Support Vector Machines, Extremely Randomized Trees; and / or

[0101] (iv) Random Forest, Adaptive Boosting, Extremely Randomized Trees.

[0102] 13. The CIM of any of paragraphs 1 to 12, wherein the analysis of the one or more test results involves detecting the presence and / or amount of a biomarker in the one or more biological fluids.

[0103] 14. The CIM of any of paragraphs 1 to 13, wherein the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers associated with liver function, biomarkers associated with kidney function, minerals and proteins, or blood cells, or a combination thereof, optionally wherein the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers associated with oxygen carrying capacity and immune function.

[0104] 15. The CIM of any of paragraphs 1 to 14, wherein the analysis of the one or more test results involves detecting the presence and / or amount of a biomarker comprising alanine aminotransferase (ALT), alkaline phosphatase (ALP), bilirubin, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, globulin, total protein, eosinophils (absolute), basophils (absolute), neutrophils (absolute), lymphocytes (absolute), monocytes (absolute), eosinophils (%), basophils (%), neutrophils (%), lymphocytes (%), monocytes (%), white blood cells (WBC), platelets, red blood cells (RBC), mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), hematocrit (HCT), hemoglobin, or red cell distribution width (RDW), or a combination thereof; and

[0105] Optionally, wherein the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers comprising hemoglobin, hematocrit, neutrophils (absolute), red blood cells, and white blood cells; or

[0106] Optionally, wherein the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers comprising ALT, ALP, bilirubin, potassium, calcium, phosphate, total protein, albumin, basophils (absolute), neutrophils (absolute), lymphocytes (absolute), basophils (%), neutrophils (%), lymphocytes (%), WBC, RBC, MCHC, HCT, and hemoglobin; or

[0107] Optionally, wherein the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers comprising ALT, ALP, bilirubin, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, globulin, total protein, eosinophils (absolute), basophils (absolute), neutrophils (absolute), lymphocytes (absolute), monocytes (absolute), eosinophils (%), basophils (%), neutrophils (%), lymphocytes (%), monocytes (%), WBC, platelets, RBC, MCV, MCH, MCHC, HCT, hemoglobin, and RDW; or

[0108] Optionally, the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers comprising ALP, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, total protein, basophils (absolute), neutrophils (absolute), eosinophils (absolute), lymphocytes (absolute), monocytes (absolute), basophils (%), monocytes (%), WBC, platelets, MCV, MCH, MCHC, hemoglobin, and RDW (%).

[0109] 16. The CIM of any of paragraphs 1 to 15, wherein the prediction of the neurodegenerative condition relates to predicting the risk of progression of mild cognitive impairment to the neurodegenerative disease.

[0110] 17. The CIM of any of paragraphs 1 to 16, wherein the one or more test results are indicated by a clinician.

[0111] 18. The CIM of any of paragraphs 1 to 17, wherein the one or more biological fluids is an extracellular fluid.

[0112] 19. The CIM of any of paragraphs 1 to 18, wherein the one or more biological fluids are selected from the group consisting of blood, cerebrospinal fluid (CSF), serum, lymph fluid, urine, interstitial fluid, amniotic fluid, peritoneal fluid, and combinations thereof.

[0113] 20. A computer-implemented system (CIS) comprising one or more discriminative artificial intelligence (AI) platforms, the one or more discriminative artificial intelligence (AI) platforms comprising an ensemble of machine learning models, wherein at least one of the one or more discriminative AI platforms:

[0114] (i) has been trained on biomarker data disaggregated by sex and / or age, wherein the biomarker data is obtained from analysis of a biological fluid; and / or

[0115] (ii) being able to predict a subject's risk of developing a neurodegenerative or non-neurodegenerative disorder taking into account the subject's sex and / or age.

[0116] 21. The CIS of paragraph 20, wherein the one or more discriminative AI platforms are capable of predicting a subject's risk of developing a neurodegenerative or non-neurodegenerative condition by analyzing one or more test results of one or more biological fluids, taking into account the subject's sex and / or age.

[0117] 22. The CIS of paragraph 20 or 21, wherein the models in the ensemble or combination are weighted with a weight between 1 and 10 or between 1 and 5, such as 1, 2, 3, 4, 5, 6, 7, 8, 9 and 10.

[0118] 23. The CIS of any of paragraphs 20 to 22, wherein the one or more discriminative AI platforms comprise an ensemble of machine learning models, the ensemble of machine learning models comprising at least two machine learning models or at least three machine learning models selected from logistic regression, Gaussian Naive Bayes, random forest, gradient boosting, adaptive boosting, LPBoost, TotalBoost, BrownBoost, MadaBoost, LogitBoost, extreme random trees, linear discriminant analysis, support vector machine, decision tree, and k-nearest neighbor.

[0119] 24. The CIS of any of paragraphs 20 to 23, wherein the one or more discriminative AI platforms comprise an ensemble of machine learning models comprising at least two machine learning models or at least three machine learning models selected from logistic regression, Gaussian naive Bayes, random forest, gradient boosting, adaptive boosting, extreme random trees, linear discriminant analysis, and support vector machines.

[0120] 25. The CIS of any of paragraphs 20 to 24, wherein the one or more discriminative AI platforms comprise an ensemble of machine learning models comprising:

[0121] (i) Logistic regression, Gaussian naive Bayes, and random forest;

[0122] (ii) Logistic regression, gradient boosting, and extreme randomized trees;

[0123] (iii) Linear Discriminant Analysis, Support Vector Machines, and Extremely Randomized Trees; and / or

[0124] (iv) Random Forest, Adaptive Boosting, and Extremely Randomized Trees.

[0125] 26. The CIS of any of paragraphs 20 to 25, wherein the one or more discriminative AI platforms are operably connected to a graphical user interface (GUI) or a sound transmitter.

[0126] 27. The CIS of any of paragraphs 21 to 25, wherein the analysis of the one or more test results involves detecting the presence and / or amount of a biomarker in the one or more biological fluids.

[0127] 28. The CIS of any of paragraphs 21 to 27, wherein the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers associated with liver function, biomarkers associated with kidney function, minerals and proteins, or blood cells, or a combination thereof, optionally wherein the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers associated with oxygen carrying capacity and immune function.

[0128] 29. The CIS of any of paragraphs 21 to 28, wherein the analysis of the one or more test results involves detecting the presence and / or amount of a biomarker comprising alanine aminotransferase (ALT), alkaline phosphatase (ALP), bilirubin, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, globulin, total protein, eosinophils (absolute), basophils (absolute), neutrophils (absolute), lymphocytes (absolute), monocytes (absolute), eosinophils (%), basophils (%), neutrophils (%), lymphocytes (%), monocytes (%), white blood cells (WBC), platelets, red blood cells (RBC), mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), hematocrit (HCT), hemoglobin, or red cell distribution width (RDW), or a combination thereof; and

[0129] Optionally, wherein the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers comprising hemoglobin, hematocrit, neutrophils (absolute), red blood cells, and white blood cells; or

[0130] Optionally, wherein the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers comprising ALT, ALP, bilirubin, potassium, calcium, phosphate, total protein, albumin, basophils (absolute), neutrophils (absolute), lymphocytes (absolute), basophils (%), neutrophils (%), lymphocytes (%), WBC, RBC, MCHC, HCT, and hemoglobin; or

[0131] Optionally, wherein the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers comprising ALT, ALP, bilirubin, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, globulin, total protein, eosinophils (absolute), basophils (absolute), neutrophils (absolute), lymphocytes (absolute), monocytes (absolute), eosinophils (%), basophils (%), neutrophils (%), lymphocytes (%), monocytes (%), WBC, platelets, RBC, MCV, MCH, MCHC, HCT, hemoglobin, and RDW; or

[0132] Optionally, the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers comprising ALP, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, total protein, basophils (absolute), neutrophils (absolute), eosinophils (absolute), lymphocytes (absolute), monocytes (absolute), basophils (%), monocytes (%), WBC, platelets, MCV, MCH, MCHC, hemoglobin, and RDW (%).

[0133] 30. The CIS of any of paragraphs 20 to 29, which is capable of predicting a subject's risk of developing a neurodegenerative disorder, taking into account the subject's sex and / or age.

[0134] 31. The CIS of any of paragraphs 20 to 30, wherein predicting the risk of a subject developing a neurodegenerative condition involves predicting the risk of mild cognitive impairment progressing to a neurodegenerative disease.

[0135] 32. The CIS of any one of paragraphs 20 to 31, wherein the neurodegenerative disorder is selected from Alzheimer's disease, a non-Alzheimer's dementia disorder, Parkinson's disease, a parkinsonian syndrome disorder, a motor neuron disease (such as amyotrophic lateral sclerosis), Huntington's disease, a Huntington's disease-like syndrome, a transmissible spongiform encephalopathy, a chronic traumatic encephalopathy, a tauopathy (such as Pick's disease, corticobasal degeneration, progressive supranuclear palsy or Niemann-Pick disease), or any other neurodegenerative disorder.

[0136] 33. The CIS of any of paragraphs 20 to 32, wherein predicting the risk of a subject developing a neurodegenerative or non-neurodegenerative condition is achieved with an accuracy of greater than or equal to 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95% or 99%.

[0137] 34. The CIS of any of paragraphs 21 to 333, wherein the one or more biological fluids is an extracellular fluid.

[0138] 35. The CIS of any of paragraphs 21 to 34, wherein the one or more biological fluids are selected from blood, cerebrospinal fluid (CSF), serum, lymph, urine, interstitial fluid, amniotic fluid, peritoneal fluid, and combinations thereof.

[0139] 36. The CIM of any of paragraphs 1 to 19 or the CIS of any of paragraphs 20 to 35, wherein at least one of the one or more discriminative AI platforms has been trained by a human.

[0140] Example

[0141] Example 1: Companion Diagnostic Platform for Detecting the Transition from Mild Cognitive Impairment to Alzheimer's Disease, Starting with Data from 48,116 Individuals

[0142] This non-limiting example establishes an AD prediction model in MCI subjects using routine blood test data and machine learning methods. Using a large population of 43,981 AD patients and 4,537 MCI subjects accumulated from 2000 to 2019 in Hong Kong, China, and 210 original features collected from routine blood tests, a prediction model for AD conversion was generated by using a variable machine learning algorithm. In addition, the analysis was further grouped by sex (male and female) and age (65-74 years old, 75-89 years old) to obtain more accurate predictions, with an AUC of 0.70. These findings highlight the use of routine laboratory test data and sophisticated machine learning methods for accurate AD risk modeling in MCI subjects and provide a useful method for early screening of AD.

[0143] Materials and methods

[0144] (i) Participants

[0145] The researchers investigated routine laboratory data of patients with AD or MCI registered with the Hospital Authority Data Collaboration Laboratory (HADCL) in Hong Kong, China. HADCL was officially launched in December 2019 with the aim of collecting health data from 43 public hospitals and institutions, 49 specialist outpatient clinics and 74 general outpatient clinics in Hong Kong to establish a big data analysis platform to facilitate biotechnology research and help improve clinical and medical services. AD and MCI diagnosis results were calculated using the International Statistical Classification of Diseases and Related Health Problems (ICD-10). Diagnostic codes G30.0, G30.1, G30.8, G30.9, and G31.84 were extracted from the primary diagnosis field of outpatient or inpatient records from 2000 to 2019 and supplemented with the diagnostic definitions of mental and behavioral disorders (MBDs) proposed in ICD10-2010. A total of 48,116 registered patients aged 65 years or older at the time of diagnosis of AD or MCI were identified in the platform data. These subjects were divided into three groups: AD subjects (43,579; 30,041 women; 13,538 men), who were diagnosed with AD at the time of first diagnosis; MCI subjects (4135; 2157 women; 1978 men), who were diagnosed with MCI and did not convert to AD before 2019; and MCI-to-AD subjects (402; 237 women; 165 men), who were diagnosed with MCI at the initial state and then converted to AD during follow-up. Common laboratory data collected for each patient included chemical pathology, hematology, and immunology analytes, totaling 210 analytes.

[0146] (ii) Data processing

[0147] Although 402 subjects converted from MCI to AD, only one male and one female subject had at least three years of follow-up. Furthermore, the exact timing of the conversion of MCI subjects to AD is unknown. Therefore, the MCI to AD group was excluded due to limited available data.

[0148] Considering that common laboratory data often contain missing values, analytes common to more than 90% of subjects were used as quality filters for samples. These biomarkers included analytes related to liver function (alanine aminotransferase (ALT), alkaline phosphatase (ALP), bilirubin), variables related to renal function (creatinine, urea), minerals and proteins (potassium, sodium, calcium, phosphate, albumin, globulin, total protein), and blood cells (eosinophils (absolute), basophils (absolute), neutrophils (absolute), lymphocytes (absolute), monocytes (absolute), basophils (%), basophils (%), neutrophils (%), lymphocytes (%), monocytes (%), white blood cells (WBC), platelets, red blood cells (RBC), mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), hematocrit (HCT), hemoglobin, red cell distribution width (RDW)). During this process, each biomarker was used to identify patients with a positive value for that biomarker. After 31 rounds of screening, data from all these identified patients were collected for subsequent analysis. Ultimately, 30,211 samples were removed, resulting in 15,900 subjects with AD and 1,603 subjects with MCI.

[0149] To exclude the influence of sex and age on the prediction results, the patients were further divided into four groups: female subjects aged 65 to 74 years (female 65-74; MCI, n=262; AD, n=1,622), male subjects aged 65 to 74 years (male 65-74; MCI, n=307; AD, n=1,331), female subjects aged 75 to 89 years (female 75-89; MCI, n=517; AD, n=8,760), and male subjects aged 75 to 89 years (male 75-89; MCI, n=517; AD, n=4,187).

[0150] In short, in the data processing procedure, patients with less than 2 (including 2) missing values ​​in 31 biomarkers were retained. Initially, patients with more than 2 missing values ​​were removed, resulting in 1,501 female patients aged 65-74 (MCI, 216; AD, 1,285), 1,289 male patients aged 65-74 (MCI, 249; AD, 1,040), 7,460 female patients aged 75-89 (MCI, 420; AD, 7,040), and 3,758 male patients aged 75-89 (MCI, 420; AD, 3,338). The samples were then balanced and divided into training sets and independent test sets. However, the performance of the model using data from these samples was unstable, showing a sensitivity close to 0.90 and a specificity close to 0.50 (data not shown). Next, patients with more than 1 missing value were excluded, and only data containing one or no missing values ​​were retained. This screening procedure resulted in 1,235 female patients aged 65 to 74 years (MCI, 179; AD, 1,056), 1,048 male patients aged 65 to 74 years (MCI, 206; AD, 842), 6,064 female patients aged 75 to 89 years (MCI, 356; AD, 5,708), and 2,984 male patients aged 75 to 89 years (MCI, 334; AD, 2,650). Again, these samples were balanced and divided into training sets and independent test sets for further analysis. The results showed that models trained with data from patients containing fewer than one missing value exhibited poorer stability than those with fewer than two missing values.

[0151] Therefore, these samples are removed and only samples without any missing values ​​are retained ( Figure 1ATo ensure the reliability of the results, the interquartile range (IQR) was used to remove outliers in the sample. This is widely used to find outliers in the data by dividing the dataset into quartiles and using the distance between the third quartile and the first quartile to determine the IQR. Samples with moderate outliers (K = 1.5) were removed. Finally, 10,344 patients were excluded, leaving 777 female patients aged 65 to 74 years (MCI, n = 131; AD, n = 646), 642 male patients aged 65 to 74 years (MCI, n = 147; AD, n = 495), 3,821 female patients aged 75 to 89 years (MCI, n = 249; AD, n = 3,572), and 1,919 male patients aged 75 to 89 years (MCI, n = 222; AD, n = 1,697).

[0152] (iii) Individual biomarker analysis

[0153] Student's t-test 28 Differences in 31 biomarkers between MCI and AD in different groups were assessed (data without missing values ​​were used). P < 0.05 was considered statistically significant.

[0154] (iv) Modeling with machine learning and deep learning

[0155] The number of AD subjects in these data significantly outnumbered those with MCI. Given that class imbalance can lead to high false positive rates and poor generalization during modeling, we randomly reduced the number of AD subjects to the same number as the MCI subjects to minimize bias. The final number of subjects in each balanced group was 262 female patients aged 65 to 74 years (MCI, n = 131; AD, n = 131), 294 male patients aged 65 to 74 years (MCI, n = 147; AD, n = 147), 498 female patients aged 75 to 89 years (MCI, n = 249; AD, n = 249), and 444 male patients aged 75 to 89 years (MCI, n = 222; AD, n = 222).

[0156] In order to train a more stable model, this work adopts ensemble learning . Ensemble learning combines multiple weakly supervised models to create a more comprehensive model. The algorithm assigns higher weights to weak classifiers that are more important, and if a weak classifier incorrectly predicts a patient, the other weak classifiers can correct it. To build the best classification algorithm, 85% of each dataset (female 65-74, male 65-74, female 75-89, and male 75-89) was randomly selected as a training set, and the remaining 15% was used as an independent test set. Details of the final data distribution are as follows Figure 1B As shown in Table 1. Using 5-fold cross validation (CV) Weak classifier models are trained, and extensive screening and grid search are used to determine the optimal algorithm and the best performing weights, respectively.

[0157]

[0158] (a) Machine Learning

[0159] Figure 2A and 2B The overall framework of the MAP in the embodiment is shown. Based on the t-test results, five important biomarkers were identified in the four groups, including hemoglobin, HCT, neutrophils, RBC, and WBC. The results of the model trained with these five biomarkers showed that only the model with females aged 65 to 74 had an AUC greater than 0.65. In females aged 65 to 74, the above five biomarkers were combined with logistic regression (LR) , Gaussian Naive Bayes (GNB) and Random Forest (RF) The basic classifiers were combined to generate the model for women aged 65 to 74. These basic classifiers were used for ensemble learning, with the weights set to LR:GNB:RF = 1:3:1. In the other three groups, the performance of the model using the feature matrix formed by the five significant biomarkers could not achieve relatively high accuracy (AUC below 0.65). Therefore, forward feature selection (FFS) was used. The biomarkers in Table 2 were added one by one to optimize the model. Forward feature stepping (FFS) is a feature selection method used to select the best feature subset to train a machine learning model. On the training set of women 75-89, starting from one feature vector, features were gradually added, one feature at a time. Training was stopped when the AUC of the CV test was greater than 0.65. Therefore, an improved model for women 75-89 was developed, which was generated by a classifier that combined 19 biomarkers (analytes related to liver function (ALT, ALP, bilirubin), minerals and proteins (potassium, calcium, phosphate, total protein, albumin) and blood cells (basophils (absolute), neutrophils (absolute), lymphocytes (absolute), basophils (%), neutrophils (%), lymphocytes (%), WBC, RBC, MCHC, HCT, hemoglobin)) with LR, gradient boosting (GB) and Extremely Randomized Trees (ExtraTree, ET) Algorithm integration. Similarly, the female 75-89 model was constructed using ensemble learning with the weights set to LR: GB: ET = 1: 3: 4.

[0160] Since the model development using FFS did not work well for male data, the strategy was changed to train the model by using all 31 biomarkers. Another improved prediction model for males aged 65-74 was developed by linear discriminant analysis (LDA) , Support Vector Machine (SVM) 40 The 31 biomarkers of the ET algorithm were constructed. The model for males aged 65-74 was constructed using ensemble learning, with the basic classifier weights set as LDA: SVM: ET = 1: 2: 4. For males aged 75-89, the maximum correlation-maximum distance (MRMD) was first performed. Feature selection was performed to screen out 23 biomarkers (analytes related to liver function (ALP), variables related to renal function (creatinine, urea), minerals and proteins (potassium, sodium, calcium, phosphate, albumin, total protein), and blood cells (basophils (absolute), neutrophils (absolute), eosinophils (absolute), lymphocytes (absolute), monocytes (absolute), basophils (%), monocytes (%), WBC, platelets, MCV, MCH, MCHC, hemoglobin, RDW (%)), which were then combined with RF, AdaBoost classifier (ADA), and ET algorithms. These basic classifiers were used for ensemble learning to generate a model for male 75-89 patients with weights set to RF:ADA:ET = 1:2:3. MRMD b It is a feature selection algorithm based on PageRank, which obtains a graph by calculating the correlation distance matrix (CDM) and the Euclidean matrix (EDM), and uses the eigenvalues ​​as the nodes of the graph and the EDM as the edge weights. The PageRank value of each function is calculated and the results are sorted in descending order. The feature subset with the highest score is the result of MRMD.

[0161] Although logistic regression (LR) b Although it's called a regression, it's actually a classification model. Logistic regression is simple, parallelizable, and interpretable. The logistic distribution is a continuous distribution defined by location and scale parameters. Its shape is similar to the normal distribution, except that it has a longer tail. Samples are conceptualized as points in space. By treating samples as points in space, logistic regression not only finds a curve in space that separates the two classes, but also finds a direct relationship between the classification probability and the input vector.

[0162] The sample space is 31-dimensional, support vector machine (SVM) bA 30-dimensional hyperplane is used to separate the points in space. The goal of classification is to find the optimal hyperplane. A reasonable choice of hyperplane is one that separates the two points with the largest margin. SVM selects the hyperplane that maximizes the distance to the hyperplane for the data points closest to the hyperplane. In these models, SVM uses a Gaussian kernel function. Gaussian Naive Bayes (GNB) b It handles continuous variables and assumes that each feature for each category is normally distributed. The Gaussian Naive Bayes algorithm internally uses the probability density function of the normal distribution to calculate the probability.

[0163] There are dependencies between weak classifiers in boosting algorithms, but there are no dependencies in bagging. The ensemble strategy of bagging is usually a simple voting algorithm, and the category with the most votes is the final model output. The bagging algorithm trains the model by sampling each time, and has strong generalization ability. This helps to reduce the variance of the model, but the disadvantage is that the training set is not well fitted. Random Forest (RF) b An improved version of bagging, Extremely Randomized Trees (ERTs) is a generalized version of RF. RF uses boosting to select samples from the training set for each decision tree, whereas random sampling is not typically used in Extraly Randomized Trees. After selecting a split feature, RF's decision tree selects an optimal split point based on the Gini coefficient and mean squared error. However, Extraly Randomized Trees randomly selects a feature value to split the decision tree. As a result, compared to RF, Extremely Randomized Trees further reduces variance, but also increases bias. Boosting algorithms work by combining poorly performing models in a specific way to obtain a better-performing model. Boosting algorithms are a class of algorithms that perform gradient descent in a function space with a convex loss function. Gradient boosting and adaptive boosting (Adaboost) are based on this principle. Adaboost (ADA) was the first successfully developed boosting algorithm for binary classification. After the first tree is created, the tree's performance on each training instance is used to weight the next tree. Training data that is difficult to predict is given more weight, while instances that are easy to predict are given less weight. The updated weights will affect the learning of the next tree. After all tree models are established, predictions are made on new samples, and the performance of each tree is weighted by the accuracy of the training data. Compared with bagging and RF algorithms, ADA can fully consider the weight of each classifier, which is the main reason why ADA is time-consuming. Gradient boosting (GB) bIt is a generalized version of ADA. Gradient boosting optimizes the objective by repeatedly selecting a function that points in the direction of the negative gradient. In a sense, gradient boosting = gradient descent + boosting. Gradient boosting selects the average performance mode and adjusts it based on the performance of the previous model. ADA discovers model deficiencies by increasing the weight of the error score data point. Gradient boosting calculates the gradient to discover model deficiencies. Extreme Trees (ET) b It is a variant of random forest, designed to further increase the randomness and diversity of the model to improve model performance and generalization. The ET algorithm randomly selects biomarkers from 31 biomarkers to find the best split. This randomness makes each decision tree node more independent. ET uses a voting method to ensemble, and each decision tree produces a prediction result. The column with the most votes is the result. Linear Discriminant Analysis (LDA) b It attempts to maximize the distance between different categories through linear projection (finding the projection axis) while minimizing the distance within the same category. LDA uses the selected features for projection to map the data into a low-level space.

[0164] (b) Deep Learning

[0165] Attention-based Tabular Network (TabNet) b It is a deep learning model that combines an attention mechanism and a decision tree. The core idea of ​​TabNet is to gradually select important biomarkers for model training and prediction through a self-attention mechanism. TabNet calculates an attention score for each biomarker. The higher the attention score, the more important the biomarker. Biomarkers with high attention scores are selected to better capture information. TabNet's predictions are made step by step, using a decision tree structure that starts from the root node and branches left or right in the decision tree based on the attention score of the current feature. TabNet uses a loss function to incentivize the attention mechanism to select important biomarkers and gradually bring the model closer to the patient label (MCI or AD). The combination of the attention mechanism and the decision tree structure achieves feature selection and model prediction. This combination allows the model to maintain interpretability while effectively analyzing data to improve model performance.

[0166] (v) Performance evaluation

[0167] Cross-validation bA validation technique used in machine learning. Its advantage is that all data can be used to train the model. In 5-fold cross-validation, the computer extracts 4 / 5 of the total training set as the true training set and 1 / 5 as the test set. This process is repeated five times, training five models. The cross-validation performance is the average of the five models.

[0168] True positives (TP) are actual AD subjects classified as AD; true negatives (TN) are MCI subjects that are not classified as AD subjects; false positives (FP) are MCI subjects classified as AD; and false negatives (FN) are AD subjects that are only classified as MCI subjects. These variables are included in Eq. (1) to Eq. (4). Precision indicates how many of the patients predicted as AD by the model are always AD subjects. Sensitivity (SN, Eq. (1)) represents the proportion of patients predicted as AD among the always AD subjects. Specificity (SP, also known as recall. Eq. (2)) aims to calculate the proportion of patients predicted as MCI to subjects who are always MCI. Accuracy (ACC, Eq. (3)) is a reliable indicator in balanced data. ACC refers to the percentage of the total number of patients correctly predicted by the classifier. F-score is a comprehensive evaluation indicator of precision and recall. F-score (Eq. (4)) can well measure the relationship between precision and recall, making better judgments for the classifier. AUC refers to the area under the ROC curve. The larger the AUC, the better the performance of the model.

[0169]

[0170] result

[0171] (i) Student's t-test

[0172] The t-test results are summarized in Table 2, which identified 24 biomarkers that showed significant differences between MCI subjects and AD subjects in at least one group. The results also revealed that AD subjects had significantly decreased hemoglobin, HCT, and RBC, but increased neutrophils and WBC in all four groups.

[0173]

[0174] (ii) Model with an AUC of 0.70 on the cross-validation test set

[0175] The prediction accuracy and AUC values ​​of each model in cross validation (CV) are shown in Table 3 and Figure 3 shown.

[0176]

[0177] Specifically, using an ensemble approach combining LR, GNB, and RF, a 5-biomarker algorithm for the female 65-74 model produced an accuracy (ACC) of 0.63, an AUC of 0.70, a sensitivity (SN) of 0.61, and a specificity (SP) of 0.65. For the female 75-89 model, a 19-biomarker algorithm ensembled with LR, GNB, and ET achieved similar performance, with an ACC of 0.63, an AUC of 0.67, an SN of 0.69, and an SP of 0.58. Similarly, a 31-biomarker algorithm constructed using LDA, SVM, and ET for the male 65-74 model provided an ACC of 0.63, an AUC of 0.66, an SN of 0.62, and an SP of 0.64, while a 23-biomarker algorithm generated using RF, ADA, and ET for the male 75-89 model achieved an ACC of 0.66, an AUC of 0.68, an SN of 0.66, and an SP of 0.65.

[0178] To demonstrate the impact of age and gender on the prediction results, the model was trained using the same biomarkers and machine learning methods, but the patients were not divided into four different age and gender groups. The results showed that the prediction model without age and gender grouping performed worse than the model grouped by age and gender. Figure 3 It can be observed that after excluding age and sex, the model showed a higher SN but lower SP, resulting in a bias in identifying MCI subjects who did not convert to AD. This is likely because the model over-focused on biomarkers of AD patients and ignored biomarkers of MCI subjects.

[0179] (iii) Model with an AUC of 0.76 on an independent test set

[0180] Next, the model performance is evaluated on an independent test set. The prediction results are summarized in Table 4 and Figure 3 , and the detailed ROC curve is depicted in Figure 4 In line with the results on the CV test set, the performance of each model on the independent test set was relatively stable, with an average ACC of 0.6425, AUC of 0.685, SN of 0.6275, and SP of 0.6575. The consistency of performance on the CV test set and the independent test set suggests that our model is resistant to noise, outliers, or slight changes in the dataset, and can therefore provide reliable predictions for unknown samples. Consistent with the findings on the CV test set, the model that did not stratify patients into different age and gender groups performed worse than the model that stratified patients by age and gender, providing an average ACC of 0.56, AUC of 0.605, SN of 0.7175, and SP of 0.4075 (Tables 4 and Figure 3 ).

[0181]

[0182] (iv) MAP outperforms the Tabular Attention Network (TabNet) model

[0183] To compare the performance of MAP with models trained using deep learning methods, we used TabNet 42 (a deep learning method widely used to train interpretable models) trained 5- / 19- / 23- / 31-biomarker sets to construct TabNet models with or without age and gender grouping. The results showed that the TabNet model performed poorly on the CV test set and the independent test set (Tables 3 and 4, and Figure 3 and 4 When grouped by age and gender, TabNet models have similar AUC to MAP on the CV test set, but their ACC is lower, and the difference between SNs and SPs is larger than that of MAP (Tables 3 and 4, and Figure 3 and 4 ). Regarding the independent test set, the TabNet model achieved lower ACC, AUC, and SN, but the difference between SN and SP was larger than that between MAP (Table 4 and Figure 3 and 4 ). Consistent with previous data, the TabNet model without age and gender grouping performed worse on the CV test set and the independent test set (Tables 3, 4, and Figure 3 and 4 ).

[0184] The work described here represents (to the investigators' knowledge) the first report to predict the risk of conversion from MCI to AD using routine laboratory blood test data and machine learning techniques. Here, 31 analytes shared by >90% of patients were first screened, and samples with missing values ​​for any of these variables were excluded. Next, the resulting samples were divided into four groups (females aged 65-74 years, males aged 65-74 years, females aged 75-89 years, and males aged 75-89 years) to minimize biases due to sex and age. Significant differences in five biomarkers (hemoglobin, hematocrit, neutrophils, red blood cells, and white blood cells) were observed between MCI and AD patients across the four groups (Table 2). Subsequently, models were trained using these five biomarkers in conjunction with various machine learning algorithms, and the CV test set (AUC = 0.70) and independent test set (AUC = 0.76) demonstrated relatively good accuracy in predicting AD conversion in the female group aged 65-74 years. Furthermore, different feature selection methods and modeling approaches were used to improve prediction accuracy in the other three groups, ultimately generating a relatively accurate prediction model with an AUC approaching 0.70. The findings suggest that AD risk in MCI subjects can be predicted by building a predictive model using routine laboratory blood test data and sophisticated machine learning methods.

[0185] Although 210 raw variables were available from routine laboratory tests, most of them were excluded because they were not common to more than 90% of patients. Ultimately, only 31 variables were included for modeling, and these variables could be roughly divided into four groups: analytes related to liver function (ALT, ALP, bilirubin), variables related to renal function (creatinine, urea), minerals and proteins (potassium, sodium, calcium, phosphate, albumin, globulin, total protein), and blood cells (eosinophils (absolute), basophils (absolute), neutrophils (absolute), lymphocytes (absolute), monocytes (absolute), eosinophils (%), basophils (%), neutrophils (%), lymphocytes (%), monocytes (%), WBC, platelets, RBC, MCV, MCH, MCHC, HCT, hemoglobin, RDW). All of these parameters have been reported to be associated with cognitive function or AD. For example, plasma ALP was significantly higher in AD patients than in healthy controls and was negatively correlated with cognitive function. The urea level in AD brain is significantly increased, while the potassium level is significantly decreased. Compared with the normal control group, AD patients have decreased lymphocytes and basophils, and increased MCH and MCV. b .

[0186] Consistent with previous studies, AD patients have lower hemoglobin, HCT, and RBC levels than MCI subjects, but higher neutrophil and WBC levels. Lower hemoglobin, HCT, and RBC values ​​indicate impaired oxygen supply or hypoxia in the brain, which may promote Aβ production and neurodegeneration. Reduced hemoglobin levels have been reported to promote neuroinflammation and oxidative stress-mediated Aβ production, accelerating neurodegeneration. Furthermore, Aβ impairs RBC morphology and function. Therefore, decreased hemoglobin, HCT, and RBC levels are closely associated with AD, and restoring RBC quantity and quality may be beneficial for AD prevention. The significant increase in neutrophil and WBC counts in AD patients suggests that neutrophil and leukocyte proliferation may be affected by AD-related pathological factors, such as neuroinflammation and oxidative stress. Indeed, several inflammatory cytokines, such as TNF-α and IL-9, which are involved in AD pathogenesis, can increase neutrophil populations. Elevated neutrophil counts, in turn, promote neutrophil activation and TNF-α release by stimulating T cells. Moreover, activated neutrophils destroy the blood-brain barrier, leading to increased infiltration of inflammatory cells and cytokines in the brain, exacerbating neurodegeneration. 54-56 .

[0187] To the researchers' knowledge, this is the first study to predict the risk of MCI converting to AD based on routine laboratory blood biomarkers. The main strengths of this study are the large sample size, the data stratified by sex and age, and the easy access to routine blood analytes. Although the subjects were aged between 65 and 89 years, the researchers believe that the prediction model can still be generalized to the entire population. Some considerations include the fact that the participants provided insufficient useful information. For example, the educational level 57 The participants' cognitive performance, such as the Mini-Mental State Examination (MMSE), is important for AD but cannot be accessed due to privacy protection policies. 58 and Montreal Cognitive Assessment (MoCA) 59 Scores are also unavailable.

[0188] In summary, this work identified five important conventional blood biomarkers (hemoglobin, HCT, neutrophils (absolute), RBC, and WBC) that showed significant differences between MCI subjects and AD patients. Using conventional blood test data and sophisticated machine learning methods, an accurate prediction model for the conversion of MCI to AD was developed. In addition, an auxiliary diagnostic platform (MAP, http: / / lab.malab.cn / ~lijing / MAP.html ), which uses routine blood test results to predict the risk of progression to AD in subjects with MCI. These findings provide a non-invasive, cost-effective, and convenient method for early screening of AD and help doctors focus on high-risk patients.

[0189] References

[0190] 1a. Tiwari S, Atluri V, Kaushik A, Yndart A, Nair M. Alzheimer'sdisease: pathogenesis, diagnostics, and therapeutics. Int J Nanomed. 2019;14:5541-5554. 1b. Wan X, Wang W, Liu J, Tong T. Estimating the sample mean and standard deviation from the sample size, median, range and / or interquartilerange. BMC medical research methodology. 2014;14:1-13.

[0191] 2a. Silva MVF, Loures CDG, Alves LCV, de Souza LC, Borges KBG,Carvalho MD. Alzheimer's disease: risk factors and potentially protectivemeasures. J Biomed Sci. 2019;26. 2b. Wright RE. Logistic regression. 1995.

[0192] 3a. Barnes DE, Yaffe K. The projected effect of risk factor reductionon Alzheimer's disease prevalence. Lancet Neurol. 2011;10(9):819-828. 3b.Hearst MA, Dumais ST, Osuna E, Platt J, Scholkopf B. Support vector machines.IEEE Intelligent Systems and their applications. 1998;13(4):18-28.

[0193] 4a. Jia JP, Wei CB, Chen SQ, et al. The cost of Alzheimer's diseasein China and re-estimation of costs worldwide. Alzheimers Dement. 2018;14(4):483-491. 4b. Ontivero-Ortega M, Lage-Castellanos A, Valente G, Goebel R,Valdes-Sosa M. Fast Gaussian Naïve Bayes for searchlight classificationanalysis. Neuroimage. 2017;163:471-479.

[0194] 5a. Grassi M, Rouleaux N, Caldirola D, et al. A Novel Ensemble-BasedMachine Learning Algorithm to Predict the Conversion From Mild CognitiveImpairment to Alzheimer's Disease Using Socio-Demographic Characteristics,Clinical Information, and Neuropsychological Measures. Front Neurol. 2019;10.5b. Biau G, Scornet E. A random forest guided tour. Test. 2016;25:197-227

[0195] 6a. Petersen RC, Doody R, Kurz A, et al. Current concepts in mildcognitive impairment. Arch Neurol. 2001;58(12):1985-1992. 6b. Friedman JH.Stochastic gradient boosting. Computational statistics & data analysis. 2002;38(4):367-378

[0196] 7a. Gomersall T, Smith SK, Blewett C, Astell A. "It's definitelyAlzheimer's': Perceived benefits and drawbacks of a mild cognitive impairmentdiagnosis. Br J Health Psychol. 2017;22(4):786-804. 7b. Margineantu DD,Dietterich TG. Pruning adaptive boosting. Paper presented at: ICML1997

[0197] 8a. Moradi E, Pepe A, Gaser C, Huttunen H, Tohka J, Initi AsDN.Machine learning framework for early MRI-based Alzheimer's conversionprediction in MCI subjects. NeuroImage. 2015;104:398-412. 8b. Natekin A,Knoll A. Gradient boosting machines, a tutorial. Frontiers in neurorobotics.2013;7:21.

[0198] 9a. Zhang TT, Liao Q, Zhang DM, et al. Predicting MCI to ADConversation Using Integrated sMRI and rs-fMRI Machine Learning and GraphTheory Approach. Front Aging Neurosci. 2021;13. 9b. Geurts P, Ernst D,Wehenkel L. Extremely randomized trees. Machine learning. 2006;63:3-42.

[0199] 10a. Syaifullah AH, Shiino A, Kitahara H, Ito R, Ishida M, TanigakiK. Machine Learning for Diagnosis of AD and Prediction of MCI ProgressionFrom Brain MRI Using Brain Anatomical Analysis Using DiffeomorphicDeformation. Front Neurol. 2021;11. 10b. Balakrishnama S, Ganapathiraju A.Linear discriminant analysis-a brief tutorial. Institute for Signal andinformation Processing. 1998;18(1998):1-8.

[0200] 11a. Minhas S, Khanum A, Alvi A, et al. Early MCI-to-AD ConversionPrediction Using Future Value Forecasting of Multimodal Features. ComputIntell Neurosci. 2021;2021. 11b. Shah C, Du Q, Xu Y. Enhanced TabNet:Attentive Interpretable Tabular Learning for Hyperspectral ImageClassification. Remote Sensing. 2022;14(3):716.

[0201] 12a. Chen YR, Qian XL, Zhang YY, et al. Prediction Models forConversion From Mild Cognitive Impairment to Alzheimer's Disease: ASystematic Review and Meta-Analysis. Front Aging Neurosci. 2022;14. 12b. HeS, Guo F, Zou Q. MRMD2. 0: a python tool for machine learning with featureranking and reduction. Current Bioinformatics. 2020;15(10):1213-1221

[0202] 13a. Mapstone M, Cheema AK, Fiandaca MS, et al. Plasma phospholipidsidentify antecedent memory impairment in older adults. Nat Med. 2014;20(4):415-+. 13b. Xing W, Ghorbani A. Weighted pagerank algorithm. Paper presentedat: Proceedings. Second Annual Conference on Communication Networks andServices Research, 2004.2004.

[0203] 14a. Chen SH, Bu XL, Jin WS, et al. Altered peripheral profile ofblood cells in Alzheimer disease A hospital-based case-control study.Medicine. 2017;96(21).

[0204] 14b. Ververidis D, Kotropoulos C. Fast and accurate sequentialfloating forward feature selection with the Bayes classifier applied tospeech emotion recognition. Signal processing. 2008;88(12):2956-2970.

[0205] 15a. Proitsi P, Kim M, Whiley L, et al. Association of blood lipidswith Alzheimer's disease: A comprehensive lipidomics analysis. AlzheimersDement. 2017;13(2):140-151. 15b. Browne MW. Cross-validation methods. Journalof mathematical psychology. 2000;44(1):108-132.

[0206] 16. Boccardi V, Bubba V, Murasecco I, et al. Serum alkalinephosphatase is elevated and inversely correlated with cognitive functions insubjective cognitive decline: results from the ReGAl 2.0 project. Aging ClinExp Res. 2021;33(3):603-609.

[0207] 17. Wu KY, Xu CY, Qiu GZ, et al. Association of lower liver functionwith cognitive impairment in the Shenzhen ageing-related disorder cohort inChina. Front Aging Neurosci. 2022;14.

[0208] 18. Baldini A, Greco A, Lomi M, et al. Blood Analytes as Biomarkersof Mechanisms Involved in Alzheimer's Disease Progression. Int J Mol Sci.2022;23(21).

[0209] 19. Gomes KB, Pereira RG, Braga AA, et al. Machine Learning-BasedRoutine Laboratory Tests Predict One-Year Cognitive and Functional Decline ina Population Aged 75+ Years. Brain Sci. 2023;13(4).

[0210] 20. Kononenko I. Machine learning for medical diagnosis: history,state of the art and perspective. Artificial Intelligence in medicine. 2001;23(1):89-109.

[0211] 21. Zhang F, Li Z, Zhang B, Du H, Wang B, Zhang X. Multi-modal deeplearning model for auxiliary diagnosis of Alzheimer’s disease.Neurocomputing. 2019;361:185-195.

[0212] 22. Yu K, Tan L, Lin L, Cheng X, Yi Z, Sato T. Deep-learning-empowered breast cancer auxiliary diagnosis for 5GB remote E-health. IEEEWireless Communications. 2021;28(3):54-61.

[0213] 23. Inoue M, Suzuki H, Meno K, Liu S, Korenaga T, Uchida K.Identification of Plasma Proteins as Biomarkers for Mild Cognitive Impairmentand Alzheimer’s Disease Using Liquid Chromatography–Tandem Mass Spectrometry.International Journal of Molecular Sciences. 2023;24(17):13064.

[0214] 24. Tanveer M, Richhariya B, Khan RU, et al. Machine learningtechniques for the diagnosis of Alzheimer’s disease: A review. ACMTransactions on Multimedia Computing, Communications, and Applications(TOMM). 2020;16(1s):1-35.

[0215] 25. Falahati F, Westman E, Simmons A. Multivariate data analysis andmachine learning in Alzheimer's disease with a focus on structural magneticresonance imaging. Journal of Alzheimer's disease. 2014;41(3):685-708.

[0216] 26. Organization WH. International Statistical Classification ofDiseases and related health problems: Alphabetical index. Vol 3: World HealthOrganization; 2004.

[0217] 27. Wan X, Wang W, Liu J, Tong T. Estimating the sample mean andstandard deviation from the sample size, median, range and / or interquartilerange. BMC medical research methodology. 2014;14:1-13.

[0218] 28. Mishra P, Singh U, Pandey CM, Mishra P, Pandey G. Application ofstudent's t-test, analysis of variance, and covariance. Annals of cardiacanaesthesia. 2019;22(4):407.

[0219] 29. Branco P, Torgo L, Ribeiro RP. A survey of predictive modeling onimbalanced domains. ACM computing surveys (CSUR). 2016;49(2):1-50.

[0220] 30. Haixiang G, Yijing L, Shang J, Mingyun G, Yuanyue H, Bing G.Learning from class-imbalanced data: Review of methods and applications.Expert systems with applications. 2017;73:220-239.

[0221] 31. Sagi O, Rokach L. Ensemble learning: A survey. WileyInterdisciplinary Reviews: Data Mining and Knowledge Discovery. 2018;8(4):e1249.

[0222] 32. Browne MW. Cross-validation methods. Journal of mathematicalpsychology. 2000;44(1):108-132.

[0223] 33. Wright RE. Logistic regression. 1995.

[0224] 34. Ontivero-Ortega M, Lage-Castellanos A, Valente G, Goebel R,Valdes-Sosa M. Fast Gaussian Naïve Bayes for searchlight classificationanalysis. Neuroimage. 2017;163:471-479.

[0225] 35. Rigatti SJ. Random forest. Journal of Insurance Medicine. 2017;47(1):31-39.

[0226] 36. Meyer H, Reudenbach C, Hengl T, Katurji M, Nauss T. Improvingperformance of spatio-temporal machine learning models using forward featureselection and target-oriented validation. Environmental Modelling & Software.2018;101:1-9.

[0227] 37. Friedman JH. Stochastic gradient boosting. Computationalstatistics & data analysis. 2002;38(4):367-378.

[0228] 38. Sharaff A, Gupta H. Extra-tree classifier with metaheuristicsapproach for email classification. Paper presented at: Advances in ComputerCommunication and Computational Sciences: Proceedings of IC4S 20182019.

[0229] 39. Balakrishnama S, Ganapathiraju A. Linear discriminant analysis-abrief tutorial. Institute for Signal and information Processing. 1998;18(1998):1-8.

[0230] 40. Hearst MA, Dumais ST, Osuna E, Platt J, Scholkopf B. Supportvector machines. IEEE Intelligent Systems and their applications. 1998;13(4):18-28.

[0231] 41. He S, Guo F, Zou Q. MRMD2. 0: a python tool for machine learningwith feature ranking and reduction. Current Bioinformatics. 2020;15(10):1213-1221.

[0232] 42. Shah C, Du Q, Xu Y. Enhanced TabNet: Attentive InterpretableTabular Learning for Hyperspectral Image Classification. Remote Sensing.2022;14(3):716.

[0233] 43. Kellett KAB, Williams J, Vardy ERLC, Smith AD, Hooper NM. Plasmaalkaline phosphatase is elevated in Alzheimer's disease and inverselycorrelates with cognitive function. Int J Mol Epidemiol Genet. 2011;2(2):114-121.

[0234] 44. Xu JS, Begley P, Church SJ, et al. Graded perturbations ofmetabolism in multiple regions of human brain in Alzheimer's disease:Snapshot of a pervasive metabolic disorder. Biochim Biophys Acta Mol BasisDis. 2016;1862(6):1084-1092.

[0235] 45.Roberts BR, Doecke JD, Rembach A, et al. Rubidium and potassiumlevels are altered in Alzheimer's disease brain and blood but not incerebrospinal fluid. Acta Neuropathol Com. 2016;4.

[0236] 46. Peers C, Dallas ML, Boycott HE, Scragg JL, Pearson HA, Boyle JP.Hypoxia and Neurodegeneration. Hypoxia and Consequences from Molecule toMalady. 2009;1177:169-177.

[0237] 47. Salminen A, Kauppinen M, Kaamiranta K. Hypoxia / ischemia activateprocessing of Amyloid Precursor Protein: impact of vascular dysfunction inthe pathogenesis of Alzheimer's disease. J Neurochem. 2017;140(4):536-549.

[0238] 48. Jayakumar R, Kusiak JW, Chrest FJ, et al. Red cell perturbationsby amyloid β-protein. Bba-Gen Subjects. 2003;1622(1):20-28.

[0239] 49. Mohanty JG, Eckley DM, Williamson JD, Launer LJ, Rifkind JM. Dored blood cell β-amyloid interactions alter oxygen delivery in Alzheimer'sdisease? Oxygen Transport to Tissue Xxix. 2008;614:29-35.

[0240] 50. Pluta R, Ulamek-Koziol M, Januszewski S, Czuczwar SJ. Platelets,lymphocytes and erythrocytes from Alzheimer's disease patients: the quest forblood cell-based biomarkers. Folia Neuropathol. 2018;56(1):14-20.

[0241] 51. Cowburn AS, Deighton J, Walmsey SR, Chilvers ER. The survivaleffect of TNF-α in human neutrophils is mediated via NF-κB-dependent IL-8release. Eur J Immunol. 2004;34(6):1733-1743.

[0242] 52. Holmes C, Cunningham C, Zotova E, Culliford D, Perry VH.Proinflammatory cytokines, sickness behavior, and Alzheimer disease.Neurology. 2011;77(3):212-218.

[0243] 53. Dong XY, Nao JF, Shi JL, Zheng DM. Predictive Value of RoutinePeripheral Blood Biomarkers in Alzheimer's Disease. Front Aging Neurosci.2019;11.

[0244] 54. Baik SH, Cha MY, Hyun YM, et al. Migration of neutrophilstargeting amyloid plaques in Alzheimer's disease mouse model. NeurobiolAging. 2014;35(6):1286-1292.

[0245] 55. Pietronigro EC, Della Bianca V, Zenaro E, Constantin G. NETosisin Alzheimer's Disease. Front Immunol. 2017;8.

[0246] 56. Sayed A, Bahbah EI, Kamel S, Barreto GE, Ashraf GM, Elfil M. Theneutrophil-to-lymphocyte ratio in Alzheimer's disease: Current understandingand potential applications. J Neuroimmunol. 2020;349.

[0247] 57. Stern Y, Gurland B, Tatemichi TK, Tang MX, Wilder D, Mayeux R.Influence of education and occupation on the incidence of Alzheimer'sdisease. Jama. 1994;271(13):1004-1010.

[0248] 58. Hoops S, Nazem S, Siderowf A, et al. Validity of the MoCA andMMSE in the detection of MCI and dementia in Parkinson disease. Neurology.2009;73(21):1738-1745.

[0249] 59. Nasreddine ZS, Phillips NA, Bédirian V, et al. The MontrealCognitive Assessment, MoCA: a brief screening tool for mild cognitiveimpairment. Journal of the American Geriatrics Society. 2005;53(4):695-699.

[0250] 60. Zetterberg H, Wahlund L-O, Blennow K. Cerebrospinal fluid markersfor prediction of Alzheimer's disease. Neuroscience letters. 2003;352(1):67-69.

[0251] 61. Iqbal K, Liu F, Gong CX, Grundke-Iqbal I. Tau in Alzheimerdisease and related tauopathies. Current Alzheimer Research. 2010;7(8):656-664.

[0252] 62. Fossati S, Cejudo JR, Debure L, et al. Plasma tau complements CSFtau and P-tau in the diagnosis of Alzheimer's disease. Alzheimer's &Dementia: Diagnosis, Assessment & Disease Monitoring. 2019;11:483-492.

[0253] 63. Pinto TC, Machado L, Bulgacov TM, et al. Is the MontrealCognitive Assessment (MoCA) screening superior to the Mini-Mental State Examination (MMSE) in the detection of mild cognitive impairment (MCI) and Alzheimer's Disease (AD) in the elderly? International psychogeriatrics. 2019;31(4):491-504.

[0254] Those skilled in the art will recognize, or be able to ascertain using no more than routine experimentation, many equivalents to the specific embodiments of the invention described herein. Such equivalents are intended to be encompassed by the following claims.

Claims

1. A computer-implemented method (CIM) for predicting a neurodegenerative disorder or a non-neurodegenerative disorder in a subject, the CIM comprising: (i) employing one or more discriminative artificial intelligence (AI) platforms to perform analysis of one or more test results of one or more biological fluids, preferably wherein the one or more discriminative artificial intelligence platforms are operably connected to a graphical user interface (GUI) or a sound transmitter; and (ii) providing a prediction of the neurodegenerative disorder or non-neurodegenerative disorder on the GUI or the sound emitter based on the analysis.

2. The CIM of claim 1, wherein at least one of the one or more discriminative AI platforms: (i) has been trained on biomarker data disaggregated by sex and / or age, wherein the biomarker data is obtained from analysis of a biological fluid; and / or (ii) predicting the neurodegenerative disorder or non-neurodegenerative disorder taking into account the sex and / or age of the subject.

3. The CIM of claim 1 or 2, wherein the prediction of the neurodegenerative disorder or non-neurodegenerative disorder is achieved with an accuracy greater than or equal to 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95% or 99%.

4. The CIM of any one of claims 1 to 3, wherein the prediction of the neurodegenerative disorder or non-neurodegenerative disorder is based on analysis of one or more biological fluids from a single subject.

5. The CIM of any one of claims 1 to 3, wherein the prediction of the neurodegenerative disorder or non-neurodegenerative disorder is based on analysis of one or more biological fluids from a plurality of subjects.

6. The CIM of any one of claims 1 to 5, wherein a prediction of a neurodegenerative disorder is provided on the GUI or the sound emitter.

7. The CIM of any one of claims 1 to 6, wherein the CIM helps predict a neurodegenerative disorder between 1 to 10 years before the neurodegenerative disorder develops.

8. The CIM of any one of claims 1 to 7, wherein the neurodegenerative disorder is Alzheimer's disease, a non-Alzheimer's dementia disorder, Parkinson's disease, a parkinsonian disorder, a motor neuron disease (such as amyotrophic lateral sclerosis), Huntington's disease, a Huntington's disease-like syndrome, a transmissible spongiform encephalopathy, a chronic traumatic encephalopathy, a tauopathy (such as Pick's disease, corticobasal degeneration, progressive supranuclear palsy, or Niemann-Pick disease), or any other neurodegenerative disorder.

9. The CIM of any one of claims 1 to 8, wherein at least one of the one or more discriminative AI platforms comprises an ensemble of machine learning models, wherein the models in the ensemble are weighted with a weight between 1 and 10 or between 1 and 5, e.g., 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10.

10. The CIM according to any one of claims 1 to 9, wherein The one or more discriminative AI platforms include an integration of machine learning models, which includes at least two machine learning models or at least three machine learning models selected from logistic regression, Gaussian naive Bayes, random forest, gradient boosting, adaptive boosting, LPBoost, TotalBoost, BrownBoost, MadaBoost, LogitBoost, extreme random trees, linear discriminant analysis, support vector machine, decision tree and k-nearest neighbor.

11. The CIM of any one of claims 1 to 9, wherein the one or more discriminative AI platforms comprise an ensemble of machine learning models comprising at least two machine learning models or at least three machine learning models selected from logistic regression, Gaussian Naive Bayes, random forest, gradient boosting, adaptive boosting, extreme random trees, linear discriminant analysis, and support vector machines.

12. The CIM of any one of claims 1 to 11, wherein the one or more discriminative AI platforms comprise an integration of machine learning models comprising: (i) Logistic regression, Gaussian naive Bayes, and random forest; (ii) Logistic regression, gradient boosting, and extreme randomized trees; (iii) Linear Discriminant Analysis, Support Vector Machines, Extremely Randomized Trees; and / or (iv) Random Forest, Adaptive Boosting, Extremely Randomized Trees.

13. The CIM of any one of claims 1 to 12, wherein analysis of the one or more test results comprises detecting the presence and / or amount of a biomarker in the one or more biological fluids.

14. The CIM of any one of claims 1 to 13, wherein the analysis of the one or more test results comprises detecting the presence and / or amount of: biomarkers associated with liver function, biomarkers associated with kidney function, minerals and proteins, or blood cells, or a combination thereof, optionally wherein the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers associated with oxygen carrying capacity and immune function.

15. The CIM of any one of claims 1 to 14, wherein the analysis of the one or more test results comprises detecting the presence and / or amount of a biomarker comprising alanine aminotransferase (ALT), alkaline phosphatase (ALP), bilirubin, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, globulin, total protein, eosinophils (absolute), basophils (absolute), neutrophils (absolute), lymphocytes (absolute), monocytes (absolute), eosinophils (%), basophils (%), neutrophils (%), lymphocytes (%), monocytes (%), white blood cells (WBC), platelets, red blood cells (RBC), mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), hematocrit (HCT), hemoglobin, or red cell distribution width (RDW), or a combination thereof; and Optionally, wherein the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers comprising hemoglobin, hematocrit, neutrophils (absolute), red blood cells, and white blood cells; or Optionally, wherein the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers comprising ALT, ALP, bilirubin, potassium, calcium, phosphate, total protein, albumin, basophils (absolute), neutrophils (absolute), lymphocytes (absolute), basophils (%), neutrophils (%), lymphocytes (%), WBC, RBC, MCHC, HCT, and hemoglobin; or Optionally, wherein the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers comprising ALT, ALP, bilirubin, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, globulin, total protein, eosinophils (absolute), basophils (absolute), neutrophils (absolute), lymphocytes (absolute), monocytes (absolute), eosinophils (%), basophils (%), neutrophils (%), lymphocytes (%), monocytes (%), WBC, platelets, RBC, MCV, MCH, MCHC, HCT, hemoglobin, and RDW; or Optionally, the analysis of the one or more test results involves detecting the presence and / or amount of biomarkers comprising ALP, creatinine, urea, potassium, sodium, calcium, phosphate, albumin, total protein, basophils (absolute), neutrophils (absolute), eosinophils (absolute), lymphocytes (absolute), monocytes (absolute), basophils (%), monocytes (%), WBC, platelets, MCV, MCH, MCHC, hemoglobin, and RDW (%).

16. The CIM of any one of claims 1 to 15, wherein the prediction of the neurodegenerative disorder comprises predicting the risk of progression of mild cognitive impairment to the neurodegenerative disorder.

17. The CIM of any one of claims 1 to 16, wherein the one or more test results are indicated by a clinician.

18. The CIM of any one of claims 1 to 17, wherein the one or more biological fluids is an extracellular fluid.

19. The CIM of any one of claims 1 to 18, wherein the one or more biological fluids are selected from the group consisting of blood, cerebrospinal fluid (CSF), serum, lymph fluid, urine, interstitial fluid, amniotic fluid, peritoneal fluid, and combinations thereof.

20. A computer-implemented system (CIS) comprising one or more discriminative artificial intelligence (AI) platforms, the one or more discriminative artificial intelligence (AI) platforms comprising an ensemble of machine learning models, wherein at least one of the one or more discriminative AI platforms: (i) has been trained on biomarker data disaggregated by sex and / or age, wherein the biomarker data is obtained from analysis of a biological fluid; and / or (ii) being able to predict the risk of a subject developing a neurodegenerative disorder or a non-neurodegenerative disorder taking into account the subject's sex and / or age.

21. The CIS of claim 20, wherein the one or more discriminative AI platforms are capable of predicting a subject's risk of developing a neurodegenerative disorder or a non-neurodegenerative disorder, taking into account the subject's sex and / or age, by analyzing one or more test results of one or more biological fluids.

22. The CIS of claim 20 or 21, wherein the models in the ensemble are weighted with a weight between 1 and 10 or between 1 and 5, such as 1, 2, 3, 4, 5, 6, 7, 8, 9 and 10.

23. The CIS according to any one of claims 20 to 22, wherein The one or more discriminative AI platforms include an integration of machine learning models, which includes at least two machine learning models or at least three machine learning models selected from logistic regression, Gaussian naive Bayes, random forest, gradient boosting, adaptive boosting, LPBoost, TotalBoost, BrownBoost, MadaBoost, LogitBoost, extreme random trees, linear discriminant analysis, support vector machine, decision tree and k-nearest neighbor.

24. The CIS of any one of claims 20 to 23, wherein the one or more discriminative AI platforms comprise an ensemble of machine learning models comprising at least two machine learning models or at least three machine learning models selected from logistic regression, Gaussian naive Bayes, random forest, gradient boosting, adaptive boosting, extreme random trees, linear discriminant analysis, and support vector machines.

25. The CIS of any one of claims 20 to 24, wherein the one or more discriminative AI platforms comprise an ensemble of machine learning models comprising: (i) Logistic regression, Gaussian naive Bayes, and random forest; (ii) Logistic regression, gradient boosting, and extreme randomized trees; (iii) Linear Discriminant Analysis, Support Vector Machines, and Extremely Randomized Trees; and / or (iv) Random Forest, Adaptive Boosting, and Extremely Randomized Trees.

26. The CIS of any one of claims 20 to 25, wherein the one or more discriminative AI platforms are operably connected to a graphical user interface (GUI) or a sound transmitter.

27. The CIS of any one of claims 20 to 26, which is capable of predicting the risk of a subject developing a neurodegenerative disorder taking into account the subject's sex and / or age.

28. The CIS of any one of claims 20 to 27, wherein predicting the risk of a subject developing a neurodegenerative disorder comprises predicting the risk of mild cognitive impairment progressing to the neurodegenerative disorder.

29. The CIS of any one of claims 20 to 28, wherein the neurodegenerative disorder is selected from the group consisting of Alzheimer's disease, a non-Alzheimer's dementia disorder, Parkinson's disease, a parkinsonian syndrome disorder, a motor neuron disease (such as amyotrophic lateral sclerosis), Huntington's disease, a Huntington's disease-like syndrome, a transmissible spongiform encephalopathy, a chronic traumatic encephalopathy, a tauopathy (such as Pick's disease, corticobasal degeneration, progressive supranuclear palsy, or Niemann-Pick disease), or any other neurodegenerative disorder.

30. The CIS of any one of claims 20 to 29, wherein predicting a subject's risk of developing a neurodegenerative disorder or a non-neurodegenerative disorder is achieved with an accuracy greater than or equal to 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95% or 99%.

31. The CIS of any one of claims 21 to 30, wherein the one or more biological fluids is an extracellular fluid.

32. The CIS of any one of claims 21 to 31, wherein the one or more biological fluids are selected from the group consisting of blood, cerebrospinal fluid (CSF), serum, lymph fluid, urine, interstitial fluid, amniotic fluid, peritoneal fluid, and combinations thereof.

33. The CIS of any one of claims 21 to 32, wherein the CIS is configured to implement the CIM of any one of claims 1 to 19.

34. The CIM of any one of claims 1 to 19 or the CIS of any one of claims 20 to 33, wherein at least one of the one or more discriminative AI platforms has been trained by a human.

Images