Diagnostic method of determining the presence of a label-free antibody-antigen complex of interest in a biological sample

Abstract

Diagnostic method (100) of determining the presence of a label-free antibody-antigen complex of interest in a biological sample: Providing (101) a trained model based on a training data set processed by multivariate statistics and a machine learning algorithm, wherein the training data set comprises: immunoassay training data of the labeled antibody-antigen complex obtained from at least one immunoassay technique; spectral training data comprising one or more training vibrational spectra of the labeled and / or label-free antibody-antigen complex obtained from at least one vibrational spectroscopy technique; Recording (102) at least one sample vibrational spectrum of a biological sample using the at least one vibrational spectroscopy technique; and Applying (103) the trained model onto the at least one sample vibrational spectrum and determine whether or not the label-free antibody-antigen complex is present in the biological sample.

Description

[0001] Diagnostic method of determining the presence of a label-free antibody-antigen complex of interest in a biological sample

[0002] Field of the Invention

[0003] The present invention relates to a diagnostic method of determining the presence of a label- free / unlabeled antibody-antigen complex of interest in a biological sample and a diagnostic method of determining the presence of a property of interest of a label-free / unlabeled molecule of a biological sample.

[0004] Background of the Invention

[0005] Diagnostic methods relate to sample analysis, which may include determining whether a parameter which may be associated and / or indicative of a disease and / or (health) condition of a patient, such as a biomolecule and / or a molecule and / or a chemical element, is present in the sample. Further, the sample analysis may include determining the concentration of the parameter and / or a property of the parameter. The property of the parameter may include a conformational state, a binding state, a binding strength, a binding type, a charging state, an aggregation state, the state and / or content of the local environment of the parameter or the like. The property of the parameter may specifically comprise the binding state of an antibody-antigen complex wherein the parameter corresponds to at least one of the antibody or the antigen or a molecule that is associated and / or binds with the antibody or the antigen. Typically, the parameter corresponds to an analyte, specifically an antibody that is present and / or produced by an organism that is tested.

[0006] Antibody-antigen interactions are of paramount importance to the life science industry. These antibody-antigen interactions involve a combination of electrostatic interactions, hydrogen bonds, Van der Waals forces and hydrophobic interactions. Understanding these interactions can be challenging owing to the complex nature of these molecules. Even powerful spectroscopic techniques at times do not provide clear evidence of the binding interactions. For example, in cases in which the backbone of both the antigen and antibody is a protein the spectral features of the antigen, the antibody and the antigen-antibody complex are indistinguishable. Commonly used methods such as surface plasmon resonance, fluorescence, ELISA etc. require modifications to the antibody or the antigen molecule. In a wide variety of diagnostic tests relying on the antibody-antigen interactions, the parameter, which may correspond to the antibody, is typically labeled to make its presence detectable. These structural modifications, including the labeling, are cost intensive, time consuming and tedious. Thus, there is an unmet need for a method that can characterize and / or detect these interactions bypassing the labeling and / or modification steps.

[0007] Vibrational spectroscopy, such as Raman spectroscopy has the potential to be used in clinical applications, such as the early detection of diseases, monitoring of treatment response, and identification of biomarkers for personalized medicine. However, this potential is limited as Raman spectra of bio fluids are complex due to the presence of a wide variety of molecules, each with its characteristic Raman bands. Further, the Raman bands are often broad and have significant overlap with other constituents. Moreover, intermolecular interactions between different analytes and / or parameters in bio fluids can profoundly influence the vibrational bands of the individual components. These challenges limit the use of this otherwise powerful technique for clinical diagnostics.

[0008] Summary of the Invention

[0009] It is therefore desirable to provide methods and / or concepts, which address the above challenges. Specifically, it is desirable to identify a property of a molecule of a biological sample. Moreover, it is desirable to identify the binding state of an antibody and an antigen and whether an antibodyantigen complex is formed. It is further of interest to identify the property of a molecule and / or the binding state of an antibody and an antigen without the need to process the sample and / or add labels to the molecules. It is desirable to provide a reliable and / or efficient method and / or concept of determining the presence of a label-free antibody-antigen complex of interest in a biological sample and / or the presence of a property of interest of an label-free / unlabeled molecule of a biological sample.

[0010] At least one of these objects is overcome by the subject matter of the independent claims. Further embodiments with optional features are subjected to the dependent claims. The invention according to aspects and related embodiments thereof are described as follows, in more detail.

[0011] According to a first aspect, a diagnostic method of determining the presence of a label-free antibodyantigen complex of interest in a biological sample comprises: providing a trained model based on a training data set processed by multivariate statistics and a machine learning algorithm, wherein the training data set comprises: immunoassay training data of a corresponding labeled antibody-antigen complex (i.e. the labeled version of the label-free antibody-antigen complex) obtained from at least one immunoassay technique; spectral training data comprising one or more training vibrational spectra of the labeled and / or the label-free antibody-antigen complex obtained from at least one vibrational spectroscopy technique; recording at least one sample vibrational spectrum of a biological sample using the at least one vibrational spectroscopy technique; and applying the trained model onto the at least one sample vibrational spectrum and determine whether (or not) the label-free antibodyantigen complex is present in the biological sample.

[0012] The method allows identifying the binding state of an antibody and an antigen and finding out whether an antibody-antigen complex is formed. Specifically, the method allows identifying the binding state of an antibody and an antigen without the need to process the sample and / or add labels to the molecules, i.e. the antibody and the antigen in the sample. A reliable and efficient method of determining the presence of a label-free antibody-antigen complex of interest in a biological sample is therefore provided. The method allows integration of vibrational spectroscopy, specifically Raman spectroscopy for identifying antibody-antigen complexes using multivariate statistics and machine learning for the characterization of such antibody-antigen interactions.

[0013] An antibody-antigen complex, also known as an immune complex, forms when an antibody binds to its specific antigen. Immune complexes are involved in various immune responses and can be measured to diagnose certain diseases.

[0014] An antibody may correspond to a protein produced by B cells of the immune system of an organism. An antibody may specifically recognize and bind to a particular antigen. An antibody may have variable regions that determine its specificity for an antigen.

[0015] An antigen may correspond to any substance that can induce an immune response. An antigen may correspond to a protein or polysaccharide, or can be any molecule that the immune system recognizes as foreign.

[0016] The formation of an antibody-antigen complex occurs when the variable region of the antibody binds to a specific epitope on the antigen. This binding is highly specific, akin to a lock and key mechanism.

[0017] The training of the model is based on data obtained by using standard techniques (also denoted “gold standard techniques”) comprising at least one immunoassay technique recording pre-defined preparations that comprise the antibody, the antigen and antibody-antigen complex, respectively wherein at least one of the antibody and the antigen is labeled for making it visible and / or detectable that the antibody-antigen complex has formed. Often, a labeled antibody-antigen complex involves the use of an antibody that is tagged / labeled with a detectable marker.

[0018] Therefore, an antibody is conjugated, i.e. labeled with a detectable label. Common labels include fluorescent dyes, enzymes, radioisotopes, or biotin. Fluorescent dye labels emit light when excited by a specific wavelength. They are commonly used in techniques like flow cytometry and fluorescence microscopy. Enzyme labels may catalyze a reaction that produces a detectable product, often a color change. Examples include horseradish peroxidase (HRP) and alkaline phosphatase (AP). Enzyme labels are often used in ELISA (Enzyme-Linked Immunosorbent Assay) and Western blotting. Radioisotope labels may emit radiation that can be detected using specialized equipment. Such labels are often used in radioimmunoassays. Biotin labels may bind with high affinity to streptavidin or avidin, which can be labeled with various detectable markers. Such labels are often used in various immunoassays and imaging techniques.

[0019] When a labeled antibody-antigen complex involves the use of an antibody that is tagged / labeled with a detectable marker, the labeled antibody binds to its specific antigen, forming the labeled antibodyantigen complex.

[0020] The label on the antibody allows for the detection of the complex through various methods, such as fluorescence in which the complex may be visualized under a fluorescence microscope; an enzymatic reaction by which the complex may be detected by a color change in an ELISA; and / or radiation by which the complex may be measured using a scintillation counter in radio-immunoassays. As a specific example, in a fluorescence-based assay, a fluorescently labeled antibody binds to a specific antigen on the surface of a cell. When viewed under a fluorescence microscope, the cells expressing the antigen will fluoresce, allowing researchers to identify and quantify the antigen.

[0021] The immunoassay training data may be obtained for a labeled antibody-antigen complex (i.e. the labeled version of the label-free antibody-antigen complex) obtained from at least one immunoassay technique, wherein the label is specific to the detection of an immunoassay technique.

[0022] The spectral training data comprising one or more training vibrational spectra may be obtained for the labeled antibody-antigen complex used to obtain the immunoassay training data. The spectral training data comprising one or more training vibrational spectra may be obtained for a differently labeled antibody-antigen complex as used to obtain the immunoassay training data, in that case a different label is used compared to the label used for the immunoassay training data. The spectral training data comprising one or more training vibrational spectra may be obtained for the antibodyantigen complex being labeled with a Raman-active label and / or an IR active label. The spectral training data comprising one or more training vibrational spectra may be obtained for a label-free antibody-antigen complex obtained from at least one vibrational spectroscopy technique.

[0023] The training of the model is based also on data obtained by using at least one vibrational spectroscopy technique such as Raman spectroscopy recording spectra of pre-defined preparations that comprise the antibody, the antigen and antibody-antigen complex, respectively wherein at least one of the antibody and the antigen is labeled for making it visible and / or detectable that the antibody-antigen complex has formed. Often, also for the vibrational spectroscopy, a labeled antibody-antigen complex involves the use of an antibody that is tagged / labeled with a detectable marker. Alternatively or in addition, an antigen may be labeled. If Raman spectroscopy is chosen as a vibrational spectroscopy technique, a Raman active label may be chosen. A Raman-active label may comprise a molecule that has a strong Raman signal and can be conjugated to either the antibody or the antigen. Common labels include Raman dyes or reporter molecules.

[0024] The Raman spectroscopy technique may in most cases comprise a common Raman spectroscopy type that measures a signal from molecules in a sample in a bulk. Alternatively, it may comprise surface- enhanced Raman scattering (SERS). SERS allows measuring a strong signal from molecules in the vicinity of a metallic surface, like gold or silver. The metallic surface may have a roughened surface (using metallic nanoparticles and / or electrochemical roughening) for further enhancement of the signal. Moreover, the metallic surface may be functionalized to capture and / or bind the antibodyantigen complex. For example, the antigen may be bound to the surface such that the antibody is also captured to the surface when forming the complex. In all cases, a specific Raman shift corresponding to the label can be used to confirm the presence of the antibody-antigen complex. Moreover, other bands in the Raman spectrum may identify molecular groups in the antibody and the antigen.

[0025] A biological sample may comprise at least one of: a bodily fluid, specifically blood and / or a component thereof such as plasma and / or serum and / or other components, cerebrospinal fluid, lymph, saliva, urine, sweat, gastric juices, bile, amniotic fluid, seminal fluid.

[0026] Recording the at least one sample vibrational spectrum of a biological sample using the at least one vibrational spectroscopy technique is performed to identify an antibody-antigen complex in the biological in which no component has a label. In other words, the biological sample does not require a labeling as a pre-processing technique to make the antibody-antigen complex. The sample may therefore be a native sample. This is possible because the trained model is used to help identifying whether or not an antibody-antigen complex has formed in the biological sample. The trained model is however, as previously outlined, based on pre-defined preparations which do not correspond to the sample and which have labeled compounds, either the antibody or the antigen or the antibody-antigen complex of which at least one is labeled. The pre-defined preparations are well-defined and it is known whether an antibody-antigen complex of interest has formed. Therefore, applying the trained model onto the at least one sample vibrational spectrum allows determining whether the label-free antibody-antigen complex of interest is present in the biological sample. It is to be noted that the preparations for generating the training data comprises the specific antibody-antigen complex (or antibody or antigen) of interest that is supposed to be identified in the sample. The only difference is given in that at least one of the antibody and the antigen comprises a label that is not present in the antibody and antigen of interest that may be in the sample. The spectral training data and the recorded vibrational spectra of the biological sample are based on the same type of vibrational spectroscopy. For example, the model may be trained on Raman spectra of a pre-defined preparation and the recorded sample vibrational spectra of the biological sample are then obtained by using Raman spectroscopy.

[0027] The at least one vibrational spectroscopy technique may comprise at least one of: Raman spectroscopy; Infrared spectroscopy; UV / vis spectroscopy.

[0028] The listed vibrational spectroscopy methods are non-invasive and mild standard techniques to efficiently gain information about a sample based on molecular vibrations. A sample that was measured by means of such a vibrational spectroscopy may be re-used for other diagnostic methods as it does not degrade by the influence of the measurement technique.

[0029] The at least one immunoassay technique may comprise at least one of: Electrochemoluminescence (ECL); Enzyme-Linked Immunosorbent Assay (ELISA); Radioimmunoassay (RIA); Chemiluminescent Immunoassay (CLIA); Fluorescence Immunoassay (FIA); Immunohistochemistry (IHC); Western Blotting (Immunoblotting); Immunomagnetic Separation (IMS); Lateral Flow Immunoassay (LFIA); Fluorescence Resonance Energy Transfer (FRET); Bioluminescence Resonance Energy Transfer (BRET); Proximity Immunoassays; Microsphere-based Immunoassays.

[0030] The listed immunoassay techniques are standard techniques and can reliably identify whether a complex has formed.

[0031] The immunoassay training data may comprise at least one of: a confirmation that the labeled antibody-antigen complex was formed; a measure for the strength of the binding forces in the labeled antibody-antigen complex; a number and / or a type and / or a location of binding sites in the labeled antibody-antigen complex; a conformation of the labeled antibody-antigen complex; stoichiometry of the complex; kinetic data; presence of interfering substances.

[0032] The spectral training data may comprise one or more training vibrational spectra of the isolated antibody and / or the isolated antigen which form together in a non-isolated state the labeled antibody- antigen complex, wherein one or both of the isolated antibody and the isolated antigen comprise a label.

[0033] In other words, at least one training vibrational spectrum is obtained and / or provided for a pre-defined preparation that contains the antibody but not the antigen. At least one training vibrational spectrum is obtained and / or provided for a pre-defined preparation that contains the antigen but not the antibody. And at least one training vibrational spectrum is obtained and / or provided for a pre-defined preparation that contains the antibody-antigen complex. Alternatively or in addition, at least one training vibrational spectrum is obtained and / or provided for a pre-defined preparation that contains the antigen and the antibody, wherein one or both of the antigen and the antibody are blocked to form a complex, for example having the binding sites blocked with a blocking molecule. The model therefore learns what spectra look like for situations in which the antibody and / or the antigen is present without forming a complex and how the spectra differ when a complex has formed.

[0034] The label-free antibody-antigen complex of interest may comprise at least one of: a cardiac biomarker, specifically NTproBNP, Troponin T, Myoglobin, CK-MB, BNP, FABP, Galectin-3; Alzheimer's Disease Biomarkers, specifically Amyloid-beta (AP), Tau protein, YKL-40; Cancer Biomarkers, specifically PSA, CEA, CA 125; Infectious Disease Biomarkers, specifically HIV p24 antigen, Hepatitis B surface antigen; Neurological Biomarkers, specifically SIOOB protein; Inflammatory Biomarkers, specifically C-Reactive Protein (CRP), Interleukin-6 (IL-6); Metabolic Biomarkers, specifically Hemoglobin Ale (HbAlc), Insulin.

[0035] The above listed analytes may correspond to the antibody or the antigen of an antibody-antigen complex. Further, these analytes comprise biomarkers for identifying a variety of widely spread diseases.

[0036] The trained model may be based on at least one of: a Linear Regression Model; a Decision Tree; a Random Forest; a Support Vector Machine (SVM); a Neural Network (including Convolutional Neural Networks, Recurrent Neural Networks, etc.); a Naive Bayes Classifier; a K-Nearest Neighbor (K-NN); a Gradient Boosting Machine (GBM); a Clustering Model (e.g., K-Means, Hierarchical Clustering); Partial Least Squares (PLS) Regression; Support Vector Regression (SVR); Extreme Gradient Boosting (XGBoost); Autoencoders; Self-Organizing Maps (SOMs).

[0037] The providing of the trained model may comprise: obtaining the immunoassay training data by using the at least one immunoassay technique; and / or obtaining the spectral training data by using the at least one vibrational spectroscopy technique; and / or obtaining the trained model by training the model with the training data set. The diagnostic method may further comprise a pre-processing applied to the one or more training vibrational spectra and / or the at least one sample vibrational spectrum, specifically wherein the preprocessing comprises at least one of: a baseline correction; a normalization; a smoothing.

[0038] The diagnostic method may further comprise: analyzing the spectral training data by means of a database and / or computational tools such as DFT calculations, Normal Coordinate Analysis (NCA); and / or applying a classification algorithm to the spectral training data to characterize unknown spectra based on patterns learnt on training data, wherein the classification algorithm comprises at least one of: SVM; Random Forests; Logistic Regression; Decision Trees; Gradient Boosting Machines; K- Nearest neighbours; Convolutional Neural Networks (CNN), specifically wherein the classification algorithm is tuned for specific functions by taking into account a specific spectral region identified by means of computational tools.

[0039] The training data set may comprise: immunoassay training data of a corresponding labeled antibody and / or antigen being in an uncomplexed conformation obtained from at least one immunoassay technique; spectral training data comprising one or more training vibrational spectra of the labeled and / or the label-free antibody and / or antigen being in the un-complexed conformation obtained from at least one vibrational spectroscopy technique; and the method may further comprise applying the trained model onto the at least one sample vibrational spectrum and determine whether (or not) the label-free antibody and / or antigen being in the un-complexed conformation is present in the biological sample.

[0040] According to a second aspect, a diagnostic method of determining the presence of a property of interest of an unlabeled / label-free molecule of a biological sample comprises: providing a trained model based on a training data set processed by multivariate statistics and a machine learning algorithm, wherein the training data set comprises: non-vibrational-spectroscopic training data for a labeled molecule having the property obtained from a gold standard technique; spectral training data comprising one or more training vibrational spectra of the labeled and / or label-free molecule having the property; recording at least one sample vibrational spectrum of the biological sample; and applying the trained model onto the at least one sample vibrational spectrum and determine whether the property is present in the label-free molecule in the biological sample.

[0041] The method allows identifying a property of a molecule of a biological sample. Specifically the method allows identifying the property of a molecule without the need to process the sample and / or add labels to the molecules. Therefore, a reliable and efficient method of identifying / determining the presence of a property of interest of an label-free / unlabeled molecule of a biological sample is provided.

[0042] The spectral training data may further comprise one or more training vibrational spectra of the labeled molecule not having the property.

[0043] The gold standard technique may comprise at least one of: Electrochemoluminescence (ECL); Enzyme-Linked Immunosorbent Assay (ELISA); Radioimmunoassay (RIA); Chemiluminescent Immunoassay (CLIA); Fluorescence Immunoassay (FIA); Immunohistochemistry (IHC); Western Blotting (Immunoblotting); Immunomagnetic Separation (IMS); Lateral Flow Immunoassay (LFIA); Fluorescence Resonance Energy Transfer (FRET); Bioluminescence Resonance Energy Transfer (BRET); Proximity Immunoassays; Microsphere-based Immunoassays.

[0044] The property of interest of an unlabeled / label-free molecule of a biological sample may comprise at least one of: a presence of a parameter specifically an analyte, such as a cardiac biomarker, specifically NTproBNP, Troponin T, Myoglobin, CK-MB, BNP, FABP, Galectin-3; Alzheimer's Disease Biomarkers, specifically Amyloid-beta (AP), Tau protein, YKL-40; Cancer Biomarkers, specifically PSA, CEA, CA 125; Infectious Disease Biomarkers, specifically HIV p24 antigen, Hepatitis B surface antigen; Neurological Biomarkers, specifically SIOOB protein; Inflammatory Biomarkers, specifically C-Reactive Protein (CRP), Interleukin-6 (IL-6); Metabolic Biomarkers, specifically Hemoglobin Ale (HbAlc), Insulin, and / or a property of any one of the listed analytes, specifically a conformation of a molecule; a binding state and / or a binding partner and / or a binding strength; a charging and / or redox state of a molecule and / or component thereof such as a redox center; an surrounding environment of a molecule.

[0045] The at least one vibrational spectroscopy technique may comprise at least one of: Raman spectroscopy; Infrared spectroscopy; UV / vis spectroscopy.

[0046] The immunoassay training data may comprise at least one of: a confirmation of the presence of the property of interest of the unlabeled / label-free molecule of the biological sample; a measure for the strength of the property of interest; a number and / or a type and / or a location of the property of interest; a conformation of the molecule having the property; stoichiometry of the molecule having the property; kinetic data; presence of interfering substances.

[0047] The spectral training data may comprise one or more training vibrational spectra of a pre-defined preparation of the molecule not having the property and separated therefrom the molecule having the property.

[0048] The trained model may be based on at least one of: a Linear Regression Model; a Decision Tree; a Random Forest; a Support Vector Machine (SVM); a Neural Network (including Convolutional Neural Networks, Recurrent Neural Networks, etc.); a Naive Bayes Classifier; a K-Nearest Neighbor (K-NN); a Gradient Boosting Machine (GBM); a Clustering Model (e.g., K-Means, Hierarchical Clustering); Partial Least Squares (PLS) Regression; Support Vector Regression (SVR); Extreme Gradient Boosting (XGBoost); Autoencoders; Self-Organizing Maps (SOMs).

[0049] The providing of the trained model may comprise: obtaining the non-vibrational-spectroscopic training data for a labeled molecule having the property by using the gold standard technique; and / or obtaining the spectral training data by using the at least one vibrational spectroscopy technique; and / or obtaining the trained model by training the model with the training data set.

[0050] The diagnostic method may further comprise a pre-processing applied to the one or more training vibrational spectra and / or the at least one sample vibrational spectrum, specifically wherein the preprocessing comprises at least one of: a baseline correction; a normalization; a smoothing.

[0051] The diagnostic method may further comprise: analyzing the spectral training data by means of a database and / or computational tools such as DFT calculations, Normal Coordinate Analysis (NCA); and / or applying a classification algorithm to the spectral training data to characterize unknown spectra based on patterns learnt on training data, wherein the classification algorithm comprises at least one of: SVM; Random Forests; Logistic Regression; Decision Trees; Gradient Boosting Machines; K- Nearest neighbours; Convolutional Neural Networks (CNN), specifically wherein the classification algorithm is tuned for specific functions by taking into account a specific spectral region identified by means of computational tools.

[0052] According to a further aspect, a computer-program product contains machine-readable instructions which when loaded and run on a machine, a computer or a system respectively cause the machine, the computer or the system to perform the previously described methods or an embodiment thereof, i.e. the method according to the first and / or the second aspect. The present disclosure therefore describes the integration of Raman spectroscopy, multivariate statistics and machine learning for characterization antibody-antigen interactions and Raman group signaling technology for clinical diagnostics.

[0053] A molecule is a group of two or more atoms that are chemically bonded together. These atoms can be of the same element or different elements. Molecules are the smallest units of a chemical compound that retain the chemical properties of the compound.

[0054] A biomolecule is any molecule that is produced by living organisms and may play a role in the structure and function of cells. Biomolecules are often essential to various biological processes and are typically categorized into four major types:

[0055] • Carbohydrates, which are considered sugar molecules and their polymers. They provide energy and structural support. Examples include glucose, starch, and cellulose.

[0056] • Proteins, which are composed of amino acids and perform a vast array of functions, including catalyzing metabolic reactions (enzymes), providing structural support, and regulating cellular processes. Examples include enzymes, antibodies, and hemoglobin.

[0057] • Lipids, which are fats and fat-like substances that are important for storing energy, forming cell membranes, and serving as signaling molecules. Examples include triglycerides, phospholipids, and steroids.

[0058] • Nucleic Acids, which store and transmit genetic information. The two main types are DNA (deoxyribonucleic acid) and RNA (ribonucleic acid).

[0059] Additionally, there are smaller biomolecules like vitamins and hormones that play crucial roles in various biochemical processes.

[0060] An antigen is a specific form of a biomolecule, comprising protein antigens, polysaccharide antigens, lipid and nucleic acid antigens and epitopes. An antigen is a substance that can trigger an immune response in the body, typically by being recognized as foreign by the immune system. While many antigens are proteins, they can also be other types of molecules, such as polysaccharides, lipids, or nucleic acids. Often diagnostic tests use antigens for binding an antibody of interest from a sample

[0061] A biomarker, or biological marker, is a measurable indicator of a biological state or condition. Biomarkers are used extensively in medical research, clinical practice, and drug development to assess health, diagnose diseases, monitor disease progression, and evaluate responses to treatment.

[0062] Biomarkers include diagnostic biomarkers, prognostic biomarkers, predictive biomarkers, monitoring biomarkers. Diagnostic biomarkers are often used to detect or confirm the presence of a disease or condition. For example, elevated blood sugar levels as a biomarker for diabetes. Prognostic biomarkers may provide information about the likely course of a disease. For example, certain gene mutations in cancer can indicate a more aggressive disease. Predictive biomarkers may indicate the likely response to a particular treatment. For example, HER2 protein levels can predict response to certain breast cancer therapies. Monitoring biomarkers are often used to track the progress of disease or the effects of treatment. For example, PSA levels in prostate cancer monitoring.

[0063] Biomarkers may comprise molecular biomarkers, including DNA, RNA, proteins, and metabolites. For example, BRCA1 and BRCA2 gene mutations as biomarkers for breast cancer risk. Biomarkers may comprise imaging biomarkers including MRI, CT scans, and PET scans that provide visual information about the state of tissues and organs. Biomarkers may comprise physiological biomarkers including blood pressure, heart rate, and temperature.

[0064] Biomarkers are critical in the development of personalized medicine, where treatments are tailored to individual patients based on their unique biomarker profiles. The process of identifying and validating new biomarkers involves extensive research and clinical trials to ensure their accuracy, reliability, and clinical utility.

[0065] An analyte is a specific substance or chemical constituent that is being measured in a diagnostic test. It is the actual component in the sample that is of interest for medical evaluation or research. Analytes may comprise: chemical elements, molecules, biomolecules, biomarkers, antibodies, antigens, proteins, peptides, hormones, DNA, RNA, lipids etc.

[0066] A parameter, in the context of diagnostics, refers to a measurable factor or characteristic that can be quantified and used to assess a specific condition or disease state. Parameters can include the levels of analytes, but they can also encompass a broader range of measurements and observations. A parameter may include properties of analytes, samples and / or vital measures of a body. A property of a sample may include a physical property such as a fluidity, the state of aggregation, a gas pressure and many more.

[0067] The concentration or presence of an analyte in a sample (such as blood, urine, or tissue) can serve as a diagnostic parameter. For instance, blood glucose levels (analyte) are used as a parameter to diagnose and monitor diabetes. Some parameters may be derived from multiple analytes or through calculations based on analyte levels. For example, the estimated glomerular filtration rate (eGFR) is a parameter derived from serum creatinine (analyte) levels to assess kidney function. Both analytes and parameters may be important in the field of diagnostics, providing valuable information for the diagnosis, monitoring, and management of diseases. As used in the following, the terms “have”, “comprise” or “include” or any arbitrary grammatical variations thereof are used in a non-exclusive way. Thus, these terms may both refer to a situation in which, besides the feature introduced by these terms, no further features are present in the entity described in this context and to a situation in which one or more further features are present. As an example, the expressions “A has B”, “A comprises B” and “A includes B” may both refer to a situation in which, besides B, no other element is present in A (i.e. a situation in which A solely and exclusively consists of B) and to a situation in which, besides B, one or more further elements are present in entity A, such as element C, elements C and D or even further elements.

[0068] Further, it shall be noted that the terms “at least one”, “one or more” or similar expressions indicating that a feature or element may be present once or more than once typically will be used only once when introducing the respective feature or element. In the following, in most cases, when referring to the respective feature or element, the expressions “at least one” or “one or more” will not be repeated, non-withstanding the fact that the respective feature or element may be present once or more than once.

[0069] Further, as used in the following, the terms "preferably", "more preferably", "particularly", "more particularly", "specifically", "more specifically" or similar terms are used in conjunction with optional features, without restricting alternative possibilities. Thus, features introduced by these terms are optional features and are not intended to restrict the scope of the claims in any way. The invention may, as the skilled person will recognize, be performed by using alternative features. Similarly, features introduced by "in an embodiment of the invention" or similar expressions are intended to be optional features, without any restriction regarding alternative embodiments of the invention, without any restrictions regarding the scope of the invention and without any restriction regarding the possibility of combining the features introduced in such way with other optional or non-optional features of the invention.

[0070] It is to be understood that the present invention is not limited to the particular embodiments and examples described herein as these may vary. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to limit the scope of the present invention which will be limited only by the appended claims. Unless defined otherwise, all technical and scientific terms used herein have the same meanings as commonly understood by one of ordinary skill in the art.

[0071] Detailed Description of the Invention In the following, some embodiments will be described in detail, wherein the invention should not be understood to be limited to the embodiments described. The following embodiments and figures are provided to aid the understanding of the present invention, the true scope of which is set forth in the appended claims. Single features being described in a particular embodiment may be arbitrarily combined, given that they are not excluding each other. In addition, different features, which are provided together in the example embodiments, are not to be considered restrictive to the invention.

[0072] Skilled artisans appreciate that elements in the figures are illustrated for simplicity and clarity and have not necessarily been drawn to scale. For example, the dimensions of some of the elements in the figures may be exaggerated relative to other elements whereas other elements may have been left out or represented in a reduced number in order to enhance clarity and improve understanding of the aspects of the present disclosure.

[0073] The same reference numerals are used among different embodiments and examples for the same or similar elements or elements that have similar or the same effects.

[0074] Description of the Figures

[0075] Fig- 1 is a flowchart reflecting the diagnostic method of determining the presence of a label-free antibody-antigen complex of interest in a biological sample, according to an embodiment;

[0076] Fig- 2 is a flowchart reflecting the diagnostic method of determining the presence of a property of interest of a label-free molecule in a biological sample, according to another embodiment;

[0077] Fig. 3a is a scheme illustrating the measurement of an antibody, an antigen and an antibody-antigen complex with vibrational spectroscopy according to an embodiment;

[0078] Fig. 3b shows vibrational spectra of the antibody, the antigen and the antibody-antigen complex according to an embodiment;

[0079] Fig. 3c is a scheme illustrating the measurement of an antibody, an antigen and an antibody-antigen complex with one or more gold standard technique, specifically comprising an immunoassay technique according to an embodiment;

[0080] Fig. 3d is a scheme illustrating the training of the model using the training data sets obtained from the vibrational spectroscopy and the gold standard technique, specifically comprising an immunoassay technique according to an embodiment;

[0081] Fig. 4a is a scheme of the gold standard technique, specifically comprising an immunoassay technique according to an embodiment; Fig. 4b is a scheme of the vibrational spectroscopy, specifically comprising Raman spectroscopy according to an embodiment; and

[0082] Fig. 4c is a scheme illustrating the training of the model using the training data sets obtained from the Raman spectroscopy and the immunoassay technique together with multivariate statistics and machine learning and / or artificial intelligence.

[0083] Fig- 1 is a flowchart reflecting the diagnostic method 100 of determining the presence of a label-free antibody-antigen complex 3 of interest in a biological sample, according to an embodiment. The diagnostic method 100 of determining the presence of a label-free antibody-antigen complex 3 of interest in a biological sample comprises the steps: providing 101 a trained model based on a training data set for a corresponding labeled antibody-antigen complex 3, wherein the training data set comprises: immunoassay training data of the labeled antibody-antigen complex 3 obtained from at least one immunoassay technique 6; spectral training data comprising one or more training vibrational spectra 4 of the labeled antibody-antigen complex 3 obtained from at least one vibrational spectroscopy technique 5. Further, the method 100 comprises recording 102 at least one sample vibrational spectrum 4 of the biological sample using the at least one vibrational spectroscopy technique 5; and applying 103 the trained model onto the at least one sample vibrational spectrum 4 and determine whether the label-free antibody-antigen complex 3 is present in the biological sample. Providing 101 a trained model based on a training data set may comprise providing the trained model to a customer wherein the training of the model and / or the generation of the training data set is fully or at least partially performed elsewhere, for example by the provider of one or more instruments which are used and / or a software that helps performing the method. The providing 101 of the trained model may alternatively or in addition comprise: obtaining 101a the immunoassay training data by using the at least one immunoassay technique 6; and / or obtaining 101b the spectral training data by using the at least one vibrational spectroscopy technique 5; and / or obtaining 101c the trained model by training the model with the training data set.

[0084] Fig- 2 is a flowchart reflecting the diagnostic method 200 of determining the presence a property of interest of a label-free molecule of a biological sample, according to another embodiment. The diagnostic method 200 of determining the presence of a property of interest of a label-free molecule of a biological sample comprises: providing 201 a trained model based on a training data set, wherein the training data set comprises: non-vibrational-spectroscopic training data for a labeled molecule having the property obtained from a gold standard technique; spectral training data comprising one or more training vibrational spectra of the labeled molecule having the property. Further, the method 200 comprises recording 202 at least one sample vibrational spectrum of the biological sample; and applying 203 the trained model onto the at least one sample vibrational spectrum and determine whether the property is present in the label-free molecule in the biological sample.

[0085] Fig. 3a is a scheme illustrating the measurement of an antibody 1, an antigen 2 and an antibodyantigen complex 3 with vibrational spectroscopy 5 according to an embodiment. Fig. 3b shows vibrational spectra 4 of the antibody 1, the antigen 2 and the antibody-antigen complex 3 according to an embodiment. Fig. 3c is a scheme illustrating the measurement of an antibody 1, an antigen 2 and an antibody-antigen complex 3 with one or more gold standard techniques 6, specifically comprising an immunoassay technique according to an embodiment. Fig. 3d is a scheme illustrating the training of the model using the training data sets obtained from the vibrational spectroscopy 5 and the gold standard technique 6, specifically comprising an immunoassay technique according to an embodiment.

[0086] In Fig. 3a it is indicated that, using a vibrational spectrometer 5, spectra 4 are recorded of an antibody 1, an anitigen 2 and an antibody-antigen complex 3. This may be the case when training data is generated. In that case, a spectrum of a preparation having an antibody 1 is recorded and a spectrum of a preparation having an antigen 2 is recorded and a spectrum of a preparation having an antibodyantigen complex 3 is recorded, wherein the antibody-antigen complex 3 comprises a label that is either attached on the antibody 1 and / or on the antigen 2.

[0087] Alternatively, a spectrum of a preparation having an antibody 1 and an antigen 2 is recorded, wherein at least one of them is blocked to avoid the complex formation and a spectrum of a preparation having an antibody-antigen complex 3 is recorded. In addition, in this case, the antibody-antigen complex 3 comprises a label that is either attached on the antibody 1 and / or on the antigen 2.

[0088] The indication of Fig. 3a may however also refer to the situation in which a sample is analyzed that comprises an antibody 1, an antigen 2 and / or an antibody-antigen complex 3, being free of a label. At least the same method is used for recording spectra to generate the training data and for analyzing the sample.

[0089] The gold standard technique 6 specifically the immunoassay technique as shown in Fig. 3c typically requires a label on the antibody and / or the antigen to make the formed complex of interest detectable. Therefore, the immunoassay technique 6 is only used to generate training data. In Fig. 3d, the training of a model according to an embodiment is described / indicated. Machine learning 7 is used as a training method for the model and it may comprise supervised learning 8, unsupervised learning 9 and / or reinforcement learning 10. Supervised learning 8 refers to a model training with labeled data, i.e. data obtained for labeled molecules. Based on supervised learning 8, a classification 11 and / or a regression 12 may be performed. Unsupervised learning 9 refers to model training with unlabeled data, i.e. data obtained for unlabeled molecules. For example, measurement data of a real sample may therefore be used in the training itself. Based on unsupervised learning 9, a clustering may be performed. Reinforcement learning 10 refers to a situation in which a model takes actions 18 in an environment 14 and then received state updates 17 and feedbacks 16 are released to the model agent 15 to further train the model.

[0090] In a specific embodiment, an integrated system employs Raman spectroscopy 5 in combination with multivariate statistics 11, 12, 13 and machine learning 7, 8, 9, 10 with existing gold standard techniques allows to provide less complex methods as shown in Fig. 3a-3d for disease detection. Instead of the determination of whether antibody-antigen complexes have formed, another type of property of a molecule, specifically a biomarker may be analyzed to detect a disease or at least an indication therefore.

[0091] Discriminant analysis for disease identification usually involves a cascade of tests carried out on biofluids via highly sophisticated automated analyzers. Most of the methods used in these analyzers 6 involve a series of steps that are necessary to capture the analytes, improve the signaling process, separate out irrelevant materials, etc.

[0092] According to an embodiment of the present disclosure, the volume of data generated by these instruments 6 offers immense scope and / or potential for machine learning (ML) assisted Raman spectroscopic investigations. Using well-characterized samples can provide the necessary labeled data to use multivariate statistics 11, 12, 13 and ML approaches 7, 8, 9, 10 to identify spectral features that can serve as disease markers. As indicated in Fig. 3b, it is difficult to distinguish the spectra 4 between the antibody 1, the antigen 2 and the antibody-antigen complex 3. Therefore, the multivariate statistics 11, 12, 13 and ML approaches 7, 8, 9, 10 allow extracting the information from the spectra 4 that would otherwise not be analyzed properly.

[0093] This approach may help in designing simpler systems based on existing gold standard techniques 6. Vibrational spectroscopy 5 such as Raman spectroscopy in combination with multivariate statistics 11, 12, 13 and machine learning 7, 8, 9, 10 can provide a nondestructive, fast and robust analysis of these interactions. The spectral information 4 obtained via Raman spectroscopy 5 is rich in information and when subjected to data analysis and ML tools can not only differentiate between individual antigen / antibody and antigen-antibody complex but also on the strength of these binding interactions.

[0094] Moreover, in cases in which there are multiple sites for binding (on the antigen), the Raman-ML platform can be used to grade the strength of these interactions. Raman spectroscopic measurements can serve on a judicious selection of antibodies, antigens and antibody-antigen combinations as an excellent data set for multivariate statistics and ML methods. The resultant Raman-ML platform may interface with existing gold standard methods for characterizing antibody-antigen interactions and may then develop into an independent method. This method may allow researchers working on the development of antibodies to rely on such a platform that may allow informed decision making and potentially reduce the number of steps to arrive at the right configuration.

[0095] Fig. 4a is a scheme of the gold standard technique 6, specifically comprising an immunoassay technique according to an embodiment. The gold standard technique 6 is provided by one or more automated analyzer in which bio-fluids are analyzed and a multitude of parameters may be identified.

[0096] Fig. 4b is a scheme of the vibrational spectroscopy 5, specifically comprising Raman spectroscopy according to an embodiment. The vibrational spectroscopy is therefore provided by a Raman spectrometer on which Raman spectroscopic measurements may be performed. The bio-fluids may therefore be investigated by means of the Raman spectrometer. Visible spectral features if any are identified and features are selected in required.

[0097] Fig. 4c is a scheme illustrating the training of the model using the training data sets obtained from the Raman spectroscopy 5 and the immunoassay technique 6 together with multivariate statistics and machine learning 7 and / or artificial intelligence. Spectral data may be processed using a host of statistical approaches. An ML platform is created to identify disease markers.

[0098] Reference list

[0099] Antibody

[0100] Antigen

[0101] Antibody-Antigen Complex

[0102] Vibrational spectra (e.g. Raman spectra)

[0103] Spectrometer

[0104] Gold standard technique analyzer / Immunoassay analyzer

[0105] Machine Learning

[0106] Supervised Learning

[0107] Unsupervised Learning

[0108] Reinforcement Learning

[0109] Classification

[0110] Regression

[0111] Clustering

[0112] Environment

[0113] Model agent

[0114] Feedbacks

[0115] State updates

[0116] Actions

[0117] Diagnostic method of determining the presence of a label-free antibodyantigen complex of interest in a biological sample -103 Method steps of method 100

[0118] Diagnostic method of determining the presence of a property of interest of an label-free molecule of a biological sample -203 Method steps of method 200

Claims

Patent Claims1. Diagnostic method (100) of determining the presence of a label -free antibody-antigen complex (3) of interest in a biological sample, the method (100) comprising:Providing (101) a trained model based on a training data set processed by multivariate statistics and a machine learning algorithm, wherein the training data set comprises: immunoassay training data of a corresponding labeled antibody-antigen complex (3) obtained from at least one immunoassay technique (6); spectral training data comprising one or more training vibrational spectra (4) of the labeled and / or the label-free antibody-antigen complex (3) obtained from at least one vibrational spectroscopy technique (5);Recording (102) at least one sample vibrational spectrum (4) of the biological sample using the at least one vibrational spectroscopy technique (5); andApplying (103) the trained model onto the at least one sample vibrational spectrum (4) and determine whether the label-free antibody-antigen complex (3) is present in the biological sample.

2. Diagnostic method (100) of claim 1, wherein the at least one vibrational spectroscopy technique (5) comprises at least one of: Raman spectroscopy; Infrared spectroscopy; UV / vis spectroscopy.

3. Diagnostic method (100) of claim 1 or 2, wherein the at least one immunoassay technique (6) comprises at least one of: Electrochemoluminescence (ECL); Enzyme-Linked Immunosorbent Assay (ELISA); Radioimmunoassay (RIA); Chemiluminescent Immunoassay (CLIA); Fluorescence Immunoassay (FIA); Immunohistochemistry (IHC); Western Blotting (Immunoblotting); Immunomagnetic Separation (IMS); Lateral Flow Immunoassay (LFIA); Fluorescence Resonance Energy Transfer (FRET); Bioluminescence Resonance Energy Transfer (BRET); Proximity Immunoassays; Microsphere-based Immunoassays.

4. Diagnostic method (100) of any one of the preceding claims, wherein the immunoassay training data comprises at least one of: a confirmation that the labeled antibody-antigen complex (3) was formed; a measure for the strength of the binding forces in the labeled antibody-antigen complex (3); a number and / or a type and / or a location of binding sites in the labeled antibody-antigen complex (3); a conformation of the labeled antibody-antigencomplex (3); stoichiometry of the complex (3); kinetic data; presence of interfering substances.

5. Diagnostic method (100) of any one of the preceding claims, wherein the spectral training data (4) comprise one or more training vibrational spectra of the isolated antibody (1) and / or the isolated antigen (2) which form together in a non-isolated state the labeled antibodyantigen complex (3), wherein one or both of the isolated antibody (1) and the isolated antigen (2) comprise a label.

6. Diagnostic method (100) of any one of the preceding claims, wherein the label-free antibodyantigen complex (3) of interest comprises at least one of: a cardiac biomarker, specifically NTproBNP, Troponin T, Myoglobin, CK-MB, BNP, FABP, Galectin-3; Alzheimer's Disease Biomarkers, specifically Amyloid-beta (AP), Tau protein, YKL-40; Cancer Biomarkers, specifically PSA, CEA, CA 125; Infectious Disease Biomarkers, specifically HIV p24 antigen, Hepatitis B surface antigen; Neurological Biomarkers, specifically SIOOB protein; Inflammatory Biomarkers, specifically C-Reactive Protein (CRP), Interleukin-6 (IL-6); Metabolic Biomarkers, specifically Hemoglobin Ale (HbAlc), Insulin.

7. Diagnostic method (100) of any one of the preceding claims, wherein the trained model is based on at least one of: a Linear Regression Model; a Decision Tree; a Random Forest; a Support Vector Machine (SVM); a Neural Network (including Convolutional Neural Networks, Recurrent Neural Networks, etc.); a Naive Bayes Classifier; a K-Nearest Neighbor (K-NN); a Gradient Boosting Machine (GBM); a Clustering Model (e.g., K-Means, Hierarchical Clustering); Partial Least Squares (PLS) Regression; Support Vector Regression (SVR); Extreme Gradient Boosting (XGBoost); Autoencoders; Self-Organizing Maps (SOMs).

8. Diagnostic method (100) of any one of the preceding claims, wherein the providing (101) of the trained model comprisesObtaining (101a) the immunoassay training data by using the at least one immunoassay technique (6); and / orObtaining (101b) the spectral training data by using the at least one vibrational spectroscopy technique (5); and / orObtaining (101c) the trained model by training the model with the training data set.

9. Diagnostic method (100) of any one of the preceding claims, further comprising a preprocessing applied to the one or more training vibrational spectra (4) and / or the at least one sample vibrational spectrum (4), specifically wherein the pre-processing comprises at least one of: a baseline correction; a normalization; a smoothing.

10. Diagnostic method (100) of any one of the preceding claims, further comprising:Analyzing the spectral training data (4) by means of a database and / or computational tools such as DFT calculations, Normal Coordinate Analysis (NCA); and / orApplying a classification algorithm to the spectral training data to characterize unknown spectra based on patterns learnt on training data, wherein the classification algorithm comprises at least one of: SVM; Random Forests; Logistic Regression; Decision Trees; Gradient Boosting Machines; K- Nearest neighbours; Convolutional Neural Networks (CNN), specifically wherein the classification algorithm is tuned for specific functions by taking into account a specific spectral region identified by means of computational tools.

11. Diagnostic method (100) of any one of the preceding claims, wherein the training data set comprises: immunoassay training data of a corresponding labeled antibody and / or antigen being in an uncomplexed conformation obtained from at least one immunoassay technique (6); spectral training data comprising one or more training vibrational spectra (4) of the labeled and / or the label-free antibody and / or antigen being in the un-complexed conformation obtained from at least one vibrational spectroscopy technique (5), and wherein the method (100) further comprises applying (103) the trained model onto the at least one sample vibrational spectrum (4) and determine whether the label-free antibody and / or antigen being in the un-complexed conformation is present in the biological sample.

12. Diagnostic method (200) of determining the presence of a property of interest of an label-free molecule of a biological sample, the method comprising:Providing (201) a trained model based on a training data set processed by multivariate statistics and a machine learning algorithm, wherein the training data set comprises: non-vibrational-spectroscopic training data for a labeled molecule having the property obtained from a gold standard technique;spectral training data comprising one or more training vibrational spectra of the labeled and / or the unlabeled molecule having the property;Recording (202) at least one sample vibrational spectrum of the biological sample; andApplying (203) the trained model onto the at least one sample vibrational spectrum and determine whether the property is present in the label-free molecule in the biological sample.

13. Diagnostic method (200) of claim 12, wherein the spectral training data (4) further comprise one or more training vibrational spectra of the labeled molecule not having the property.

14. Diagnostic method (200) of claim 12 or 13, wherein the gold standard technique (6) comprises at least one of: Electrochemoluminescence (ECL); Enzyme-Linked Immunosorbent Assay (ELISA); Radioimmunoassay (RIA); Chemiluminescent Immunoassay (CLIA); Fluorescence Immunoassay (FIA); Immunohistochemistry (IHC); Western Blotting (Immunoblotting); Immunomagnetic Separation (IMS); Lateral Flow Immunoassay (LFIA); Fluorescence Resonance Energy Transfer (FRET); Bioluminescence Resonance Energy Transfer (BRET); Proximity Immunoassays; Microsphere-based Immunoassays.

15. Diagnostic method (200) of any one of claims 12 to 14, wherein the property of interest of an unlabeled molecule of a biological sample comprises a property of at least one of: a cardiac biomarker, specifically NTproBNP, Troponin T, Myoglobin, CK-MB, BNP, FABP, Galectin- 3; Alzheimer's Disease Biomarkers, specifically Amyloid-beta (AP), Tau protein, YKL-40; Cancer Biomarkers, specifically PSA, CEA, CA 125; Infectious Disease Biomarkers, specifically HIV p24 antigen, Hepatitis B surface antigen; Neurological Biomarkers, specifically SIOOB protein; Inflammatory Biomarkers, specifically C-Reactive Protein (CRP), Interleukin-6 (IL-6); Metabolic Biomarkers, specifically Hemoglobin Ale (HbAlc), Insulin.

Images