Systems and methods for dynamic Raman profiling of biological diseases and disorders
Patent Information
- Application Number
- JP2023533674
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2020-12-04
- Filing Date
- 2021-12-03
- Publication Date
- 2024-12-05
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
There is a need for accurate, non-invasive methods to diagnose biological conditions such as neurological and neurodegenerative disorders and cancer using dynamic biological responses from human samples like hair, teeth, and nails, which can be applied to infants and young children.
The method involves exposing biological samples to a light source, obtaining Raman spectra, and processing them using Raman spectroscopy combined with artificial intelligence to predict diagnostic statuses by analyzing temporal and spatial dynamics, employing techniques like independent component analysis and recurrent quantitative analysis.
This approach achieves high sensitivity and specificity in predicting disorders like autism spectrum disorder, ADHD, ALS, and cancer with accuracy rates of up to 90% and area under the curve (AUC) values of 0.86-0.88, enabling early diagnosis in infants and young children.
Smart Images

Figure 00000000_0000_ABST
Abstract
Description
[Technical Field]
[0001] cross reference This application claims the benefits of U.S. Patent Application No. 63 / 121,800, filed on 4 December 2020, which is incorporated herein by reference. [Background technology]
[0002] Dynamic biological responses can represent fundamental biological processes that are structurally and functionally important to humans. For example, abnormal dynamic biological responses can be associated with many biological conditions, such as diseases and disorders. Examples of such biological conditions include neurological conditions (e.g., autism spectrum disorder, schizophrenia, or attention-deficit / hyperactivity disorder (ADHD)), neurodegenerative conditions (e.g., amyotrophic lateral sclerosis (ALS), Alzheimer's disease, Parkinson's disease, and Huntington's disease), and cancer (e.g., childhood cancer). [Overview of the project]
[0003] Given the above background, there is a need for accurate methods and systems for diagnosing biological conditions, particularly non-invasive diagnostics. Such diagnoses can be based on accurate profiling of biomarkers detectable by non-invasive methods for diagnosing biological conditions. This disclosure provides improved systems and methods for accurately diagnosing biological conditions based on the analysis of dynamic biological response data from biological samples obtained non-invasively from subjects. Such improved systems and methods for accurate diagnosis of biological conditions may be based on a combination of Raman profiling of biological samples and artificial intelligence data analysis. This disclosure addresses these needs, for example, by providing biological sample biomarkers for the diagnosis of biological conditions. Biological samples include human biological samples associated with incremental growth. Such biological samples may be hair shafts, teeth, and nails. The non-invasive biomarkers of this disclosure can be used for the diagnosis of infants, and in some cases, infants under one year of age.
[0004] In one embodiment, the present disclosure provides a method for predicting the diagnostic status of a subject with respect to a disease or disorder, the method comprising: (a) exposing a biological sample of the subject to a light source, wherein the biological sample includes a tooth sample, a hair sample, or a nail sample; (b) obtaining a plurality of Raman spectra from the exposed biological sample; (c) processing the plurality of Raman spectra to generate a spatial map of the plurality of Raman spectra; and (d) predicting the diagnostic status of the subject with respect to a disease or disorder, at least in part, based on the spatial map of the plurality of Raman spectra. In some embodiments, the light source comprises a laser.
[0005] In some embodiments, the analysis determines the temporal dynamics of an underlying biological process. In some embodiments, the analysis includes reducing the dimensionality of multiple Raman spectra (e.g., by independent component analysis) before processing. In some embodiments, the optical signal is generated by a light source (e.g., a laser). In some embodiments, the biological sample includes a dental sample. In some embodiments, the method further includes detecting or monitoring changes in a temporal stress profile that indicates the temporal response of the subject. In some embodiments, the temporal response includes a biochemical response. In some embodiments, the temporal response includes a biological response, a physiological response, an anatomical response, a therapeutic response, a stress-related response, or a combination thereof. In some embodiments, the multiple Raman spectra include wavenumbers from about 200 to about 3700. In some embodiments, acquisition includes using a Raman spectroscopic microscope. In some embodiments, the Raman spectroscopic microscope includes a 50x air-coupled objective lens, a 63x water-immersion-coupled objective lens, or any combination thereof. In some embodiments, the laser includes a wavelength of about 785 nm, a wavelength of about 532 nm, or any combination thereof. In some embodiments, acquisition is performed using an integration time of approximately 0.2 to 0.3 seconds. In some embodiments, acquisition involves moving the biological sample with a step size of approximately 2 to 5 microns after acquiring the Raman spectrum from among several Raman spectra.
[0006] In some embodiments, the disease or disorder includes autism spectrum disorder (ASD), attention deficit / hyperactivity disorder (ADHD), amyotrophic lateral sclerosis (ALS), schizophrenia, irritable bowel disease (IBD), childhood kidney disease, kidney transplant rejection, childhood cancer, or any combination thereof. In some embodiments, the disease or disorder is ASD. In some embodiments, the subjects are human. In some embodiments, the subjects are adults. In some embodiments, the subjects are approximately 12 to approximately 5 years old. In some embodiments, the humans are approximately 12, 11, 10, 9, 8, 7, 5, 4, 3, 2, or under 1 year old. In some embodiments, the humans are at least approximately 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 years old. In some embodiments, at least a portion of the time-Raman profile corresponds to the prenatal period of the subject.
[0007] In some embodiments, predicting the diagnostic status of a subject with respect to a disease or disorder involves processing a spatial map using a trained model. In some embodiments, processing includes extracting features from the spatial map (e.g., by recurrence quantification analysis) and analyzing the features using a trained model. In some embodiments, processing includes computational analysis of the temporal dynamics derived from the spatial map by applying dimensionality reduction techniques, such as independent component analysis (ICA) and / or principal component analysis (PCA), followed by applying recurrence quantification analysis (RQA) to extract computational features that describe the dimensions derived from ICA / PCA. In some embodiments, the trained model is selected from the group consisting of neural network algorithms, support vector machine algorithms, decision tree algorithms, unsupervised clustering algorithms, supervised clustering algorithms, regression algorithms, gradient boosting algorithms (e.g., gradient boosted implementations of machine learning algorithms such as gradient boosted decision trees), and any combination thereof. In some embodiments, the trained model includes gradient boosted ensemble models. In some embodiments, the trained model is configured to process one or more features selected from the group consisting of relapse rate, determinism, mean diagonal length, maximum diagonal length, divergence, Shannon entropy in diagonal length, relapse tendency, laminarity, trapping time, maximum vertical length, Shannon entropy in vertical length, mean relapse time, Shannon entropy in relapse time, number of most likely relapses, and / or any combination thereof. In some embodiments, the trained model is configured to process two or more features selected from the group consisting of relapse rate, determinism, mean diagonal length, maximum diagonal length, divergence, Shannon entropy in diagonal length, relapse tendency, laminarity, trapping time, maximum vertical length, Shannon entropy in vertical length, mean relapse time, Shannon entropy in relapse time, number of most likely relapses, and / or any combination thereof.
[0008] In some embodiments, the method further includes predicting the diagnostic status of a subject for a disease or disorder using a model having at least about 70%, 75%, 80%, 85%, or 90% sensitivity in predicting the diagnostic status for a disease or disorder across a suitable cohort population (e.g., as provided in the Examples section below).
[0009] In some embodiments, the method further includes predicting the diagnostic status of a subject for a disease or disorder using a model having a sensitivity of up to approximately 70%, 75%, 80%, 85%, or 90% in predicting the diagnostic status for a disease or disorder across a suitable cohort population.
[0010] In some embodiments, the method further includes predicting the diagnostic status of a subject for a disease or disorder using a model having specificity of at least about 70%, 75%, 80%, 85%, or 90% when predicting the diagnostic status for a disease or disorder across a suitable cohort population.
[0011] In some embodiments, the method further includes predicting the diagnostic status of a subject for a disease or disorder using a model having a specificity of up to approximately 70%, 75%, 80%, 85%, or 90% when predicting the diagnostic status for a disease or disorder across a suitable cohort population.
[0012] In some embodiments, the method further includes predicting the diagnostic status of a subject for a disease or disorder using a model having a positive predictive value of at least about 70%, 75%, 80%, 85%, or 90% when predicting the diagnostic status for a disease or disorder across a suitable cohort population.
[0013] In some embodiments, the method further includes predicting the diagnostic status of a subject for a disease or disorder using a model having a positive predictive value of up to approximately 70%, 75%, 80%, 85%, or 90% when predicting the diagnostic status for a disease or disorder across a suitable cohort population.
[0014] In some embodiments, the method further includes predicting the diagnostic status of a subject for a disease or disorder using a model that has a negative predictive value of at least about 70%, 75%, 80%, 85%, or 90% when predicting the diagnostic status for a disease or disorder across a suitable cohort population.
[0015] In some embodiments, the method further includes predicting the diagnostic status of a subject for a disease or disorder using a model having a negative predictive value of up to approximately 70%, 75%, 80%, 85%, or 90% when predicting the diagnostic status for a disease or disorder across a suitable cohort population.
[0016] In some embodiments, the method further includes predicting the diagnostic status of a subject for a disease or disorder using a model that predicts the diagnostic status for a disease or disorder having an area under the receiver operating characteristic curve (AUROC) of at least about 0.65, at least about 0.70, at least about 0.75, at least about 0.80, at least about 0.82, at least about 0.84, at least about 0.86, at least about 0.88, or at least about 0.90 for a suitable cohort population.
[0017] In another aspect, the Disclosure provides a device comprising one or more processors and a memory for storing one or more programs for execution by the one or more processors, wherein one or more programs include instructions for: (a) sampling each of a plurality of locations along a baseline on a biological specimen of a subject associated with a Raman signature of a subject, thereby obtaining a plurality of Raman spectra, wherein each Raman spectrum in the plurality of Raman spectra corresponds to a different location in the plurality of locations, and each location in the plurality of locations represents a different growth period of the biological specimen associated with the Raman signature; (b) analyzing each of the plurality of Raman spectra across the entire baseline on the biological specimen, thereby obtaining a first dataset; (c) deriving each second dataset from the corresponding plurality of Raman spectral measurements, wherein each feature in the corresponding set of features is determined by a continuous variation in the Raman spectrum; and (d) processing the features using a trained model to predict the diagnostic status of a subject with respect to a disease or disorder associated with a Raman signature. In some embodiments, each second dataset is derived by applying recurrence quantification analysis or related methods to the corresponding multiple Raman spectral measurements. In some embodiments, analyzing the Raman spectra includes cosmic ray removal, background correction, normalization, peak fitting, or any combination thereof.
[0018] In some embodiments, the biological sample includes a tooth sample, a hair sample, a nail sample, or any combination thereof. In some embodiments, the instructions further include detecting or monitoring changes in the Raman spectrum across multiple locations that indicate a temporal response of the subject. In some embodiments, the temporal response includes biological responses, physiological responses, anatomical responses, therapeutic responses, stress-related responses, or combinations thereof. In some embodiments, the multiple Raman spectra include wavenumbers from about 200 to about 3700. In some embodiments, sampling includes using a Raman spectroscopic microscope. In some embodiments, the Raman spectroscopic microscope includes a 50x air-coupled objective lens, a 63x water-immersion-coupled objective lens, or any combination thereof. In some embodiments, sampling includes exposing the biological sample to a light source to generate Raman spectra from among multiple Raman spectra at multiple locations. In some embodiments, the light source includes a laser, the laser includes a wavelength of about 785 nm, a wavelength of about 532 nm, or any combination thereof. In some embodiments, the instructions further include a translation, which involves moving a biological sample from a first position to a second position among several positions with a step size of approximately 2 to 5 microns after acquiring a Raman spectrum from among several Raman spectra. In some embodiments, the translation is performed using an integration time of approximately 0.2 to 0.3 seconds. In some embodiments, the disease or disorder includes autism spectrum disorder (ASD), attention deficit / hyperactivity disorder (ADHD), amyotrophic lateral sclerosis (ALS), schizophrenia, irritable bowel disease (IBD), pediatric kidney disease, kidney transplant rejection, pediatric cancer, or any combination thereof. In some embodiments, the disease or disorder is ASD. In some embodiments, predicting the diagnostic status of a subject with respect to a disease or disorder involves processing changes in Raman spectra across multiple positions using a trained model.In some embodiments, the trained model is selected from the group consisting of neural network algorithms, support vector machine algorithms, decision tree algorithms, unsupervised clustering algorithms, supervised clustering algorithms, regression algorithms, gradient boosting algorithms, and any combination thereof. In some embodiments, the trained model includes a gradient boosted ensemble model. In some embodiments, the trained model is configured to process one or more features selected from the group consisting of laminarity, entropy, trapping time (TT), mean diagonal length (MDL), recurrence time (RT), Vmax, determinism, Lmax, and any combination thereof. In some embodiments, the trained model is configured to process two or more features selected from the group consisting of laminarity, entropy, trapping time (TT), mean diagonal length (MDL), recurrence time (RT), Vmax, determinism, Lmax, and any combination thereof.
[0019] In another aspect, the Disclosure provides a non-temporary computer-readable storage medium and one or more computer programs embedded in the non-temporary controller-readable storage medium, which, when executed by a computer system, causes the computer system to (a) sample each of a plurality of locations along a baseline on a biological sample of subject associated with a Raman signature of subject, thereby obtaining a plurality of Raman spectra, wherein each Raman spectrum in the plurality of Raman spectra corresponds to a different location in the plurality of locations, and each location in the plurality of locations corresponds to a different location on the biological sample associated with the Raman signature The present invention provides a non-temporary computer-readable storage medium and a computer program for performing a method that includes: (b) sampling representing a growth period; (c) analyzing each of a plurality of Raman spectra across a baseline on a biological sample to obtain a first dataset; (d) deriving each second dataset from the corresponding plurality of Raman spectral measurements, wherein each feature in the corresponding set of features is determined by continuous variation in the Raman spectrum; and (e) processing the features using a trained model to predict the diagnostic status of a subject with respect to a disease or disorder associated with the Raman signature. In some embodiments, each second dataset is derived by applying recurrence quantification analysis or related methods to the corresponding plurality of Raman spectral measurements. In some embodiments, the analysis of the Raman spectra includes cosmic ray removal, background correction, normalization, peak fitting, or any combination thereof.
[0020] In some embodiments, the biological sample includes a tooth sample, a hair sample, a nail sample, or any combination thereof. In some embodiments, the method further includes detecting or monitoring changes in Raman spectra across multiple locations that indicate a temporal response of the subject. In some embodiments, the temporal response includes biological responses, physiological responses, anatomical responses, therapeutic responses, stress-related responses, or combinations thereof. In some embodiments, the multiple Raman spectra include wavenumbers from about 200 to about 3700. In some embodiments, sampling includes using a Raman spectroscopic microscope. In some embodiments, the Raman spectroscopic microscope includes a 50x air-coupled objective lens, a 63x water-immersion-coupled objective lens, or any combination thereof. In some embodiments, sampling includes exposing the biological sample to a light source to generate Raman spectra from among multiple Raman spectra at multiple locations. In some embodiments, the light source includes a laser, the laser includes a wavelength of about 785 nm, a wavelength of about 532 nm, or any combination thereof. In some embodiments, the instructions further include a translation, which involves moving a biological sample from a first position to a second position among several positions with a step size of approximately 2 to 5 microns after acquiring a Raman spectrum from among several Raman spectra. In some embodiments, the translation is performed using an integration time of approximately 0.2 to 0.3 seconds. In some embodiments, the disease or disorder includes autism spectrum disorder (ASD), attention deficit / hyperactivity disorder (ADHD), amyotrophic lateral sclerosis (ALS), schizophrenia, irritable bowel disease (IBD), pediatric kidney disease, kidney transplant rejection, pediatric cancer, or any combination thereof. In some embodiments, the disease or disorder is ASD. In some embodiments, predicting the diagnostic status of a subject with respect to a disease or disorder involves processing changes in Raman spectra across multiple positions using a trained model.In some embodiments, the trained model is selected from the group consisting of neural network algorithms, support vector machine algorithms, decision tree algorithms, unsupervised clustering algorithms, supervised clustering algorithms, regression algorithms, gradient boosting algorithms, and any combination thereof. In some embodiments, the trained model includes a gradient boosted ensemble model. In some embodiments, the trained model is configured to process one or more features selected from the group consisting of laminarity, entropy, trapping time (TT), mean diagonal length (MDL), recurrence time (RT), Vmax, determinism, Lmax, and any combination thereof. In some embodiments, the trained model is configured to process two or more features selected from the group consisting of laminarity, entropy, trapping time (TT), mean diagonal length (MDL), recurrence time (RT), Vmax, determinism, Lmax, and any combination thereof.
[0021] In another aspect, the Disclosure relates to a method for training a model in a computer system having one or more processors and memory for storing one or more programs to be executed by the one or more processors, wherein (a) for each of the multiple training subjects, a first subset of the training subjects has a first diagnostic state corresponding to having a first biological state associated with a Raman signature, and a second subset of the training subjects has a second diagnostic state corresponding to not having a first biological state associated with a Raman signature, (i) sampling each of the multiple locations along a baseline on a biological sample of the subject associated with the Raman signature, thereby obtaining a plurality of Raman spectra, wherein each Raman spectrum in the plurality of Raman spectra corresponds to a different location in the plurality of locations, and each location in the plurality of locations represents a different growth period of the biological sample of the subject associated with the Raman signature. The present invention provides a method comprising: (ii) sampling; (ii) analyzing each Raman spectrum across a baseline on a biological sample to obtain a first dataset; (iii) deriving each second dataset from a plurality of corresponding Raman spectra, wherein each feature in the corresponding set of features is determined by continuous variation in the Raman spectrum; and (b) training an untrained or partially untrained model using (i) the corresponding set of features of each second dataset for each of the plurality of training subjects, and (ii) the corresponding diagnostic state for each of the plurality of training subjects, selected from a first diagnostic state and a second diagnostic state, thereby obtaining a trained model that provides an index of whether the test subject has a first biological state associated with a Raman signature, based on the value of the features in the set of features obtained from a biological sample associated with the Raman signature of the test subject.In some embodiments, each second dataset is derived by applying recurrence quantification analysis or related methods to the corresponding multiple Raman spectral measurements. In some embodiments, analyzing the Raman spectra includes cosmic ray removal, background correction, normalization, peak fitting, or any combination thereof.
[0022] In some embodiments, the trained model is a neural network algorithm, a support vector machine algorithm, a decision tree algorithm, an unsupervised clustering model algorithm, a supervised clustering model algorithm, a regression model, or a gradient boost algorithm (e.g., a gradient boost implementation of a machine learning algorithm such as a gradient boosted decision tree). In some embodiments, the trained model is a multinomial classifier. In some embodiments, the trained model is a binary classifier. In some embodiments, the trained model is a regressor. In some embodiments, the first biological state is selected from the group consisting of autism spectrum disorder (ASD), attention deficit / hyperactivity disorder (ADHD), amyotrophic lateral sclerosis (ALS), schizophrenia, irritable bowel disease (IBD), childhood kidney disease, kidney transplant rejection, and childhood cancer.
[0023] In some embodiments, evaluating a subject for a first biological state associated with a Raman signature further comprises distinguishing the presence of the first biological state associated with the Raman signature from the absence of the first biological state associated with the Raman signature. In some embodiments, evaluating a subject for a first biological state associated with a Raman signature further comprises distinguishing the first biological state associated with the Raman signature from a second biological state associated with a Raman signature different from the first biological state associated with the Raman signature. In some embodiments, the first biological state is an autism spectrum disorder and the second biological state is typical development, i.e., the absence of a neurodevelopmental disorder. In some embodiments, the first biological state is an autism spectrum disorder and the second biological state is an attention deficit / hyperactivity disorder. In some embodiments, the subject is a human. In some embodiments, the subject is an adult. In some embodiments, the human is from about 12 to about 5 years old. In some embodiments, the human is less than about 12, 11, 10, 9, 8, 7, 5, 4, 3, 2, or 1 year old. In some embodiments, the human is at least about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 years old. In some embodiments, at least a portion of the temporal profile of the Raman profile corresponds to the prenatal period of the subject.
[0024] In some embodiments, the corresponding biological sample associated with the Raman signature of each training subject is selected from the group consisting of hair shafts, teeth, and nails. In some embodiments, the corresponding biological sample associated with the Raman signature of each training subject is a hair shaft, and the reference line corresponds to the longitudinal direction of the hair shaft. In some embodiments, the corresponding biological sample associated with the Raman signature of each training subject is a tooth, and the reference line corresponds to the direction across the growth zone including the neonatal line of the tooth. In some embodiments, the corresponding plurality of positions are arranged such that the first position at the corresponding plurality of positions along the corresponding biological sample of each training subject corresponds to the position closest to the tip of the corresponding biological sample of each training subject. In some embodiments, each trace in the corresponding plurality of Raman spectrum measurements includes a plurality of data points, and each data point is an instance of each position at the plurality of positions. In some embodiments, the corresponding set of features is selected from the group consisting of recurrence rate, determinism, mean diagonal length, maximum diagonal length, divergence, Shannon entropy in diagonal length, recurrence tendency, layering, trapping time, maximum vertical length, Shannon entropy in vertical length, mean recurrence time, Shannon entropy in recurrence time, the number of most likely recurrences, and / or any combination thereof. In some embodiments, the corresponding plurality of positions includes at least 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000, 5500, 6000, 6500, 7000, 7500, 8000, 8500, 9000, 9500, 10000, or more than 10000 positions.
[0025] Another aspect of the present disclosure provides a non-transitory computer-readable medium including machine-executable code that, when executed by one or more computer processors, implements any of the methods described above or elsewhere in this specification.
[0026] Another aspect of this disclosure provides a system comprising one or more computer processors and computer memory coupled to the computer processors. The computer memory contains machine-executable code that, when executed by the one or more computer processors, implements any of the methods described above or elsewhere in this specification.
[0027] Further aspects and advantages of the present disclosure will be readily apparent to those skilled in the art from the following detailed description, where only exemplary embodiments of the present disclosure are shown and described. As will be understood, other different embodiments of the present disclosure are possible, and some of their details can be modified in various obvious ways without departing from the present disclosure. Accordingly, the drawings and description should be considered illustrative and not limiting.
[0028] Reference All publications, patents, and patent applications referenced herein are incorporated by reference to the same extent as each individual publication, patent, or patent application is specifically and individually indicated as being incorporated by reference. To the extent that the publications and patents or patent applications incorporated by reference conflict with the disclosures contained herein, this specification is intended to supersede and / or take precedence over such conflicting material. [Brief explanation of the drawing]
[0029] Novel features of the present invention are described in detail in the appended claims. A better understanding of the features and advantages of the present invention can be obtained by referring to the following detailed description, which describes exemplary embodiments in which the principles of the present invention are utilized, and to the accompanying drawings (also referred to herein as “Figures”).
[0030] [Figure 1] An example block diagram of the computing device 100 of this disclosure is shown. [Figure 2A-2C] Figures of the target hair sample (Figure 2A), tooth sample (Figure 2B), and nail sample (Figure 2C) are shown. [Figure 3]A flowchart of 300 methods for evaluating a subject's biological state is shown. [Figure 4] This describes a computer system programmed or otherwise configured to implement the methods provided herein. [Figure 5] This is an example of model accuracy for predicting the diagnostic status of autism spectrum disorder (ASD), utilizing features derived from the application of RQA to dimensions derived by ICA of Raman waveforms, as shown by experimental receiver operating characteristic (ROC) curves for evaluating the accuracy of a disclosed method for evaluating subjects of autism spectrum disorder. Device performance is measured by calculating the area under the curve (AUC) of the ROC plot, which provides a measure of performance at various classification thresholds. Here, the AUC is 0.86, indicating robust and accurate predictive performance. [Figure 6] An example of model accuracy for predicting the diagnostic status of amyotrophic lateral sclerosis (ALS) is shown, utilizing features derived from the application of RQA to dimensions derived by ICA of Raman waveforms, as indicated by experimental receiver operating characteristic (ROC) curves for evaluating the accuracy of the disclosed method for evaluating subjects for autism spectrum disorder. Device performance is measured by calculating the area under the curve (AUC) of the ROC plot, which provides a measure of performance at various classification thresholds. Here, the AUC is 0.88, indicating robust and accurate predictive performance. [Modes for carrying out the invention]
[0031] While various embodiments of the present invention are shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided only as examples. Numerous variations, modifications, and substitutions can be conceived herein without departing from the present invention. It should be understood that various alternatives to the embodiments of the present invention described herein may be used.
[0032] Dynamic biological responses can represent fundamental biological processes that are structurally and functionally important to humans. For example, abnormal dynamic biological responses can be associated with many biological conditions, such as diseases and disorders. Examples of such biological conditions include neurological conditions (e.g., autism spectrum disorder, schizophrenia, or attention-deficit / hyperactivity disorder (ADHD)), neurodegenerative conditions (e.g., amyotrophic lateral sclerosis (ALS), Alzheimer's disease, Parkinson's disease, and Huntington's disease), and cancer (e.g., childhood cancer).
[0033] Given the above background, there is a need for accurate methods and systems for diagnosing biological conditions, particularly non-invasive diagnostics. Such diagnoses can be based on accurate profiling of biomarkers detectable by non-invasive methods for diagnosing biological conditions. This disclosure provides improved systems and methods for accurately diagnosing biological conditions based on the analysis of dynamic biological response data from biological samples obtained non-invasively from subjects. Such improved systems and methods for accurate diagnosis of biological conditions may be based on a combination of Raman profiling of biological samples and artificial intelligence data analysis. This disclosure addresses these needs, for example, by providing biological sample biomarkers for the diagnosis of biological conditions. Biological samples include human biological samples associated with incremental growth. Such biological samples may be hair shafts, teeth, and nails. The non-invasive biomarkers of this disclosure can be used for the diagnosis of infants, and in some cases, infants under one year of age. In some cases, the child may be about 12 years to about 5 years of age. In some embodiments, the human may be about 12, 11, 10, 9, 8, 7, 5, 4, 3, 2, or under 1 year of age. In some embodiments, the human may be at least about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 years of age.
[0034] In one embodiment, the present disclosure provides a method for predicting the diagnostic status of a subject with respect to a disease or disorder, comprising: (a) exposing a biological sample of the subject to a light source, wherein the biological sample includes a tooth sample, a hair sample, or a nail sample; (b) obtaining a plurality of Raman spectra from the exposed biological sample; (c) processing the plurality of Raman spectra to generate a spatial map of the plurality of Raman spectra; and (d) predicting the diagnostic status of the subject with respect to a disease or disorder, at least in part, based on the spatial map of the plurality of Raman spectra. In some embodiments, the light source comprises a laser.
[0035] In some embodiments, the analysis determines the temporal dynamics of an underlying biological process. In some embodiments, the analysis includes reducing the dimensionality of multiple Raman spectra before processing (e.g., by independent component analysis). In some embodiments, the optical signal is generated by a light source (e.g., a laser). In some embodiments, the biological sample includes a dental sample. In some embodiments, the method further includes detecting or monitoring changes in the temporal stress profile that indicate the temporal response of the subject. In some embodiments, the temporal response includes a biochemical response. In some embodiments, the temporal response includes a biological response, a physiological response, an anatomical response, a therapeutic response, a stress-related response, or a combination thereof. In some embodiments, the multiple Raman spectra include wavenumbers from about 200 to about 3700. In some embodiments, acquisition includes using a Raman spectroscopic microscope. In some embodiments, the Raman spectroscopic microscope includes a 50x air-coupled objective lens, a 63x water-immersion-coupled objective lens, or any combination thereof. In some embodiments, the laser includes a wavelength of about 785 nm, a wavelength of about 532 nm, or any combination thereof. In some embodiments, acquisition is performed using an integration time of approximately 0.2 to 0.3 seconds. In some embodiments, acquisition involves moving the biological sample with a step size of approximately 2 to 5 microns after acquiring the Raman spectrum from among several Raman spectra.
[0036] In some embodiments, the systems and methods disclosed herein may be used alone or in combination with other techniques. Such techniques may include laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS), C-reactive immunohistochemical fluorescence staining, and the like. In some embodiments, combining techniques may improve the diagnostic accuracy or precision of a given technique alone. In some embodiments, the addition of LA-ICP-MS may provide multiple non-invasive metal metabolic biomarkers of a given biological sample that can complement the diagnostic capabilities of Raman spectroscopy. In some embodiments, the metal metabolic biomarkers may include zinc, tin, magnesium, copper, iodide, lithium, aluminum, phosphorus, sulfur, calcium, chromium, manganese, iron, cobalt, nickel, arsenic, strontium, cadmium, tin, iodine, barium, mercury, lead, bismuth, molybdenum, or any combination thereof. In some embodiments, the addition of C-reactive protein immunohistochemical fluorescence may complement the diagnostic capabilities of Raman spectroscopy by providing temporal variations in inflammation.
[0037] In some embodiments, the disease or disorder includes autism spectrum disorder (ASD), attention deficit / hyperactivity disorder (ADHD), amyotrophic lateral sclerosis (ALS), schizophrenia, irritable bowel disease (IBD), childhood kidney disease, kidney transplant rejection, childhood cancer, or any combination thereof. In some embodiments, the disease or disorder is ASD. In some embodiments, the subjects are human. In some embodiments, the subjects are adults. In some embodiments, the subjects are approximately 12 to approximately 5 years old. In some embodiments, the humans are approximately 12, 11, 10, 9, 8, 7, 5, 4, 3, 2, or under 1 year old. In some embodiments, the humans are at least approximately 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 years old. In some embodiments, at least a portion of the time-Raman profile corresponds to the prenatal period of the subject.
[0038] In some embodiments, predicting the diagnostic status of a subject with respect to a disease or disorder involves processing a spatial map using a trained model. In some embodiments, processing includes extracting features from the spatial map (e.g., by recurrence quantification analysis) and analyzing the features using a trained model. In some embodiments, processing includes computational analysis of the temporal dynamics derived from the spatial map by applying dimensionality reduction techniques, for example, independent component analysis (ICA) and / or principal component analysis (PCA), followed by applying recurrence quantification analysis (RQA) to extract computational features that describe the dimensions derived from ICA / PCA.
[0039] In some embodiments, the trained model includes multiple parameters, and the term “parameter” refers to any coefficient or likewise any value of internal or external elements in the model (e.g., weights and / or hyperparameters) that can affect (e.g., modify, adjust, and / or tune) one or more inputs, outputs, and / or functions in the model (e.g., if the model is a regressor or classifier). For example, in some embodiments, the parameters of a model refer to any coefficients, weights, and / or hyperparameters that can be used to control, modify, adjust, and / or tune the behavior, learning, and / or performance of the model. In some examples, parameters are used to increase or decrease the influence of inputs (e.g., features) on the model. As a non-limiting example, in some embodiments, parameters are used to increase or decrease the influence of nodes (e.g., neural networks), where a node includes one or more activation functions. The assignment of parameters to specific inputs, outputs, and / or functions of a model is not limited to any one paradigm of a given model, but can be used in any suitable model for desired performance. In some embodiments, parameters have fixed values. In some embodiments, the parameter values are manually and / or automatically adjustable. In some embodiments, the parameter values are modified by a model validation and / or training process (e.g., by an error minimization and / or backpropagation method). In some embodiments, the model of this disclosure includes a plurality of parameters. In some embodiments, the number of parameters associated with a model (e.g., an untrained, partially trained, or fully trained model) is n, where n≧2, n≧5, n≧10, n≧25, n≧40, n≧50, n≧75, n≧100, n≧125, n≧150, n≧200, n≧225, n≧250, n≧350, n≧500, n≧600, n≧750, n≧1,000, n≧2,000, n≧4,000, n≧5,000, n≧7,500, n≧10,000, n≧20,000, n≧40,000, n≧75,000, n≧100,000, n≧200,000, n≧500,000, n≧1x10 6n≧5x10 6 , or n≧1x10 7 In some embodiments, n is 10,000 to 1 × 10 7 , 100,000~5×10 6 , or 500,000~1×10 6 That is the case.
[0040] In some embodiments, the trained model is selected from the group consisting of neural network algorithms, support vector machine algorithms, decision tree algorithms, unsupervised clustering algorithms, supervised clustering algorithms, regression algorithms, gradient boosting algorithms (e.g., gradient boosting implementations of machine learning algorithms such as gradient boosted decision trees), and any combination thereof. In some embodiments, the trained model includes a gradient boosted ensemble model. In some embodiments, the trained model is configured to process one or more features selected from the group consisting of recurrence rate, determinism, mean diagonal length, maximum diagonal length, divergence, Shannon entropy in diagonal length, recurrence tendency, laminarity, trapping time, maximum vertical length, Shannon entropy in vertical length, mean recurrence time, Shannon entropy in recurrence time, most likely number of recurrences, and / or any combination thereof. In some embodiments, the trained model is configured to process two or more features selected from the group consisting of relapse rate, determinism, mean diagonal length, maximum diagonal length, divergence, Shannon entropy in diagonal length, relapse tendency, stratification, trapping time, maximum vertical length, Shannon entropy in vertical length, mean relapse time, Shannon entropy in relapse time, most likely number of relapses, and / or any combination thereof. In some embodiments, the method further includes predicting the diagnostic status of a subject for disease or disorder with a sensitivity of at least about 80%. In some embodiments, the method further includes predicting the diagnostic status of a subject for disease or disorder with a sensitivity of up to about 80%. In some embodiments, the method further includes predicting the diagnostic status of a subject for disease or disorder with a specificity of at least about 80%. In some embodiments, the method further includes predicting the diagnostic status of a subject for disease or disorder with a specificity of up to about 80%. In some embodiments, the method further includes predicting the diagnostic status of a subject for disease or disorder with a positive predictive value of at least about 80%.In some embodiments, the method further includes predicting the diagnostic status of a subject for disease or disorder with a positive predictive value of up to approximately 80%. In some embodiments, the method further includes predicting the diagnostic status of a subject for disease or disorder with a negative predictive value of at least approximately 80%. In some embodiments, the method further includes predicting the diagnostic status of a subject for disease or disorder with a negative predictive value of up to approximately 80%. In some embodiments, the method further includes predicting the diagnostic status of a subject for disease or disorder with an area under the receiver operating characteristic curve (AUROC) of at least approximately 0.80.
[0041] In another aspect, the Disclosure provides a device comprising one or more processors and a memory for storing one or more programs for execution by the one or more processors, wherein one or more programs include instructions for: (a) sampling each of a plurality of locations along a baseline on a biological specimen of a subject associated with a Raman signature of a subject, thereby obtaining a plurality of Raman spectra, wherein each Raman spectrum in the plurality of Raman spectra corresponds to a different location in the plurality of locations, and each location in the plurality of locations represents a different growth period of the biological specimen associated with the Raman signature; (b) analyzing each of the plurality of Raman spectra across the entire baseline on the biological specimen, thereby obtaining a first dataset; (c) deriving each second dataset from the corresponding plurality of Raman spectral measurements, wherein each feature in the corresponding set of features is determined by a continuous variation in the Raman spectrum; and (d) processing the features using a trained model to predict the diagnostic status of a subject with respect to a disease or disorder associated with a Raman signature. In some embodiments, each second dataset is derived by applying recurrence quantification analysis or related methods to the corresponding multiple Raman spectral measurements. In some embodiments, analyzing the Raman spectra includes cosmic ray removal, background correction, normalization, peak fitting, or any combination thereof.
[0042] In another aspect, the Disclosure provides a non-temporary computer-readable storage medium and one or more computer programs embedded in the non-temporary controller-readable storage medium, which, when executed by a computer system, causes the computer system to (a) sample each of a plurality of locations along a baseline on a biological sample of subject associated with a Raman signature of subject, thereby obtaining a plurality of Raman spectra, wherein each Raman spectrum in the plurality of Raman spectra corresponds to a different location in the plurality of locations, and each location in the plurality of locations corresponds to a different location on the biological sample associated with the Raman signature The present invention provides a non-temporary computer-readable storage medium and a computer program for performing a method that includes: (b) sampling representing a growth period; (c) analyzing each of a plurality of Raman spectra across a baseline on a biological sample to obtain a first dataset; (d) deriving each second dataset from the corresponding plurality of Raman spectral measurements, wherein each feature in the corresponding set of features is determined by continuous variation in the Raman spectrum; and (e) processing the features using a trained model to predict the diagnostic status of a subject with respect to a disease or disorder associated with the Raman signature. In some embodiments, each second dataset is derived by applying recurrence quantification analysis or related methods to the corresponding plurality of Raman spectral measurements. In some embodiments, the analysis of the Raman spectra includes cosmic ray removal, background correction, normalization, peak fitting, or any combination thereof.
[0043] In another aspect, the Disclosure relates to a method for training a model in a computer system having one or more processors and memory for storing one or more programs to be executed by the one or more processors, wherein (a) for each of the multiple training subjects, a first subset of the training subjects in the multiple training subjects has a first diagnostic state corresponding to having a first biological state associated with a Raman signature, and a second subset of the training subjects in the multiple training subjects has a second diagnostic state corresponding to not having a first biological state associated with a Raman signature, (i) sampling each of the multiple locations along a baseline on a biological sample of the subject associated with the Raman signature of the subject, thereby obtaining a plurality of Raman spectra, wherein each Raman spectrum in the plurality of Raman spectra corresponds to a different location in the plurality of locations, and each location in the plurality of locations corresponds to a different growth stage of the biological sample of the subject associated with the Raman signature. A method comprising: (ii) sampling, representing an interval; (ii) analyzing each Raman spectrum across a baseline on a biological sample to obtain a first dataset; (iii) deriving each second dataset from a plurality of corresponding Raman spectra, wherein each feature in the corresponding set of features is determined by a continuous variation in the Raman spectrum; and (b) training an untrained or partially untrained model using (i) the corresponding set of features in each second dataset for each of the plurality of training subjects, and (ii) the corresponding diagnostic state for each of the plurality of training subjects, selected from a first diagnostic state and a second diagnostic state, to obtain a trained model that provides an index of whether the test subject has a first biological state associated with a Raman signature, based on the value of the features in the set of features obtained from a biological sample associated with a Raman signature of the test subject.In some embodiments, each second dataset is derived by applying recurrence quantification analysis or related methods to the corresponding multiple Raman spectral measurements. In some embodiments, analyzing the Raman spectra includes cosmic ray removal, background correction, normalization, peak fitting, or any combination thereof.
[0044] In some embodiments, the trained model is a neural network algorithm, a support vector machine algorithm, a decision tree algorithm, an unsupervised clustering model algorithm, a supervised clustering model algorithm, a regression model, or a gradient boost algorithm (e.g., a gradient boost implementation of a machine learning algorithm such as a gradient boosted decision tree). In some embodiments, the trained model is a multinomial classifier. In some embodiments, the trained model is a binary classifier. In some embodiments, the first biological state is selected from the group consisting of autism spectrum disorder (ASD), attention deficit / hyperactivity disorder (ADHD), amyotrophic lateral sclerosis (ALS), schizophrenia, irritable bowel disease (IBD), childhood kidney disease, kidney transplant rejection, and childhood cancer.
[0045] In some embodiments, evaluating a test subject for a first biological state associated with a Raman signature further includes distinguishing between the presence and absence of the first biological state associated with the Raman signature. In some embodiments, evaluating a test subject for a first biological state associated with a Raman signature further includes distinguishing between a first biological state associated with a Raman signature and a second biological state associated with a different Raman signature than the first biological state associated with the Raman signature. In some embodiments, the first biological state is autism spectrum disorder, and the second biological state is typical development, i.e., the absence of neurodevelopmental disorder. In some embodiments, the first biological state is autism spectrum disorder, and the second biological state is attention deficit / hyperactivity disorder. In some embodiments, the test subject is human. In some embodiments, the test subject is adult. In some embodiments, the human is approximately 12 to approximately 5 years old. In some embodiments, the human is approximately 12, 11, 10, 9, 8, 7, 5, 4, 3, 2, or less than 1 year old. In some embodiments, the human is at least approximately 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 years old. In some embodiments, at least a portion of the time profile of the Raman profile corresponds to the prenatal period of the subject.
[0046] In some embodiments, the corresponding biological specimen associated with each training subject's Raman signature is selected from the group consisting of hair shafts, teeth, and nails. In some embodiments, the corresponding biological specimen associated with each training subject's Raman signature is a hair shaft, and the baseline corresponds to the longitudinal direction of the hair shaft. In some embodiments, the corresponding biological specimen associated with each training subject's Raman signature is a tooth, and the baseline corresponds to the direction across the growth zone, including the neonatal line of the tooth. In some embodiments, the corresponding multiple locations are arranged such that the first location at the corresponding multiple locations along the corresponding biological specimen of each training subject corresponds to the location closest to the tip of the corresponding biological specimen of each training subject. In some embodiments, each trace in the corresponding multiple Raman spectral measurements includes multiple data points, each data point being an instance of the respective location at the multiple locations. In some embodiments, the corresponding set of features is selected from the group consisting of laminarity, entropy, trapping time (TT), mean diagonal length (MDL), recurrence time (RT), Vmax, determinism, Lmax, and any combination thereof. In some embodiments, the corresponding plurality of positions include at least positions 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000, 5500, 6000, 6500, 7000, 7500, 8000, 8500, 9000, 9500, 10000, or more than 10000.
[0047] Details of an exemplary system will be described in conjunction with Figure 1, which shows an example block diagram of the computing device 100 of this disclosure. In some implementations, the device 100 includes one or more processing unit CPUs 102 (also referred to as processors), one or more network interfaces 104, a user interface 106, non-persistent memory 111, persistent memory 112, and one or more communication buses 114 for interconnecting these components. The one or more communication buses 114 optionally include circuits (sometimes called chipsets) that interconnect and control communication between system components. Non-persistent memory 111 typically includes high-speed random-access memory such as DRAM, SRAM, DDR RAM, ROM, EEPROM, and flash memory, while persistent memory 112 typically includes CD-ROMs, digital general-purpose disks (DVDs), or other optical storage, magnetic cassettes, magnetic tapes, magnetic disk storage, or other magnetic storage devices, magnetic disk storage devices, optical disk storage devices, flash memory devices, or other non-volatile solid-state storage devices. The persistent memory 112 optionally includes one or more storage devices located remotely from the CPU 102. The non-volatile memory devices in the persistent memory 112 and non-persistent memory 112 include non-temporary computer-readable storage media.In some implementations, non-persistent memory 111, or alternatively, non-temporary computer-readable storage medium, along with possibly persistent memory 112, stores the following programs, modules, and data structures, or subsets thereof: an optional operating system 116 including procedures for handling various basic system services and performing hardware-dependent tasks; an optional network communication module (or instruction) 118 for connecting the system 100 to other devices and / or communication networks 104; an optional classifier training module 120 for trained models for evaluating subjects for biological states; and a feature data from a training subject, including feature data for one or more training subjects 124. An optional data store 122 for a dataset of biological samples, wherein the feature data includes parameters associated with each of the features 126 and a diagnostic status 128 (e.g., an indicator of whether each training subject has been diagnosed with or not with a biological status); an optional classifier validation module 130 for validating a model that distinguishes biological status; an optional data store 132 for a dataset of biological statuses and biological samples from the validation subjects; and an optional patient classification module 134 for classifying subjects with a biological status, for example, as trained using the classifier training module 120.
[0048] In various implementations, one or more of the identified elements described above are stored in one or more of the aforementioned memory devices and correspond to a set of instructions for performing the functions described above. The identified modules, data, or programs (e.g., sets of instructions) do not need to be implemented as separate software programs, procedures, datasets, or modules; therefore, various subsets of these modules and data may be combined or rearranged in various implementations. In some implementations, non-persistent memory 111 optionally stores a subset of the modules and data structures identified above. Furthermore, in some embodiments, the memory stores additional modules and data structures not described above. In some embodiments, one or more of the identified elements described above are stored in a computer system other than the computer system of the visualization system 100, which is addressable by the visualization system 100, so that the visualization system 100 can retrieve all or part of such data when needed.
[0049] In some embodiments, the system 100 is connected to or includes one or more analytical devices for performing chemical analysis. For example, an optional network communication module (or instruction) 118 is configured to connect the system 100 to one or more analytical devices, for example, via a communication network 104. In some embodiments, the one or more analytical devices include a laser ablation inductively coupled plasma mass spectrometer (LA-ICP-MS), a fluorescence image sensor, or a Raman spectrometer.
[0050] Figure 1 depicts "System 100," but the figure is intended more as a functional description of various features that may be present in a computer system than as a structural schematic of the implementation described herein. In practice, as will be recognized by those skilled in the art, items shown separately may be combined, and some items may be separated. Furthermore, Figure 1 depicts certain data and modules in non-persistent memory 111, but some or all of this data and modules may be stored in persistent memory 112.
[0051] In some embodiments, the methods of the present disclosure include obtaining a biological sample (e.g., a single hair including the hair shaft). The subject may be a human. In some embodiments, the subject is a child 12 years of age or younger (e.g., the child is 5, 4, 3, 2, 1 year, 9 months, 6 months, 3 months, or 1 month or younger). In some embodiments, the child is about 12 to about 5 years of age. In some embodiments, the human is about 12, 11, 10, 9, 8, 7, 5, 4, 3, 2, or under 1 year of age. In some embodiments, the human is at least about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 years of age. In some embodiments, the subject is an adult. In some embodiments, the subject is an adult. Figure 2A shows an example of a hair sample from a subject including the hair shaft. The hair sample may simply be cut from the subject (e.g., with the help of scissors). The method for obtaining the hair sample may be non-invasive. The obtained hair samples may have a minimum length of 1 cm (for example, hair samples may be 1 cm, 2 cm, 3 cm, 4 cm, or 5 cm in length). Hair samples may include any part of the hair (e.g., the tip or the part between the tip and the follicle). There are no special requirements for hair samples that include hair follicles. Figure 2B shows an example of a tooth sample. Figure 2C shows an example of a nail sample. In the case of nails or hair, obtaining a biological sample may refer to positioning the subject so that the nail or hair can be sampled. Nail samples may include the entire nail or a trimmed nail.
[0052] In some embodiments, the obtained biological sample is pretreated, such as by washing and drying the biological sample with one or more solvents and / or surfactants. If the biological sample is hair, the hair sample may be washed in a solution of TRITON X-100® and ultrapure metal-free water (e.g., MILLI-Q® water) and dried overnight in an oven (e.g., at 60°C). Pretreatment may further include preparing the hair shafts for measurement by placing them on a glass slide (e.g., a microscope glass slide) having an adhesive film (e.g., double-sided tape). The hair shafts may be positioned so that the hair shafts are substantially straight. The glass slide containing the hair shafts may be placed in or near a measurement system (e.g., a laser ablation inductively coupled plasma mass spectrometer (LA-ICP-MS), a fluorescence image sensor, or a Raman spectrometer) for performing the analysis. If the biological sample is teeth or nails, the surface of the biological sample may be washed (e.g., with a surfactant, water, or one or more solvents). The sample may be divided and then placed in or near a measurement system (e.g., a laser ablation inductively coupled plasma mass spectrometer (LA-ICP-MS), a fluorescence image sensor, or a Raman spectrometer) for analysis.
[0053] Figure 3 shows a flowchart of Method 300 for evaluating a subject for its biological condition, such as a method for predicting the diagnostic status of the subject for a disease or disorder. Method 300 may include exposing a biological sample of the subject to a light source (as in operation 302). In some cases, the light source may include a laser. In some embodiments, analysis determines the temporal dynamics of the underlying biological process. In some embodiments, analysis includes reducing the dimensionality of multiple Raman spectra (e.g., by independent component analysis) before processing. In some embodiments, the optical signal is generated by the light source (e.g., a laser). The biological sample may include a tooth sample, a hair sample, or a nail sample. Next, Method 300 may include obtaining multiple Raman spectra from the exposed biological sample (as in operation 304). Next, Method 300 may include processing the multiple Raman spectra (as in operation 306) to generate a spatial map of the multiple Raman spectra. Next, method 300 may include predicting the diagnostic status of a subject with respect to a disease or disorder based at least partially on a spatial map of multiple Raman spectra (as in operation 308).
[0054] In some embodiments, multiple Raman spectra are acquired using a Raman spectroscopic microscope including a 50x air-coupled objective lens or a 63x water-immersion-coupled objective lens. In some embodiments, the laser includes a wavelength of approximately 785 nm or approximately 532 nm. In some embodiments, acquisition is performed using an integration time of approximately 0.2 to 0.3 seconds. In some embodiments, acquisition includes moving the biological sample with a step size of approximately 2 to 5 microns after acquiring one of the multiple Raman spectra.
[0055] In some embodiments, the analysis includes generating a time-Raman profile based at least partially on the acquired Raman spectrum and analyzing the time-profile of the variation in the Raman spectrum. In some embodiments, at least a portion of the time-Raman profile corresponds to the prenatal period of interest.
[0056] Measurement data can be collected sequentially from a biological sample at multiple locations along the biological sample. In some embodiments, the corresponding multiple locations include at least 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000, 5500, 6000, 6500, 7000, 7500, 8000, 8500, 9000, 9500, 10000, or more than 10000 locations. In some embodiments, each location is adjacent to one another. In this manner, each region corresponding to a different location on the biological sample can be associated with dynamic (e.g., time-varying) abundance measurements. In some embodiments, each location is separated by a predetermined distance. In some embodiments, sampling is performed along a baseline of the biological sample, starting from each location closest to the tip of the biological sample, such as a hair sample (e.g., the location corresponding to the youngest age of the subject). Generally, as long as the sampling direction is known, sampling can begin from the respective location closest to the tip or root, and a suitable trained model is used for analysis.
[0057] Sampling can generate a set of data points. Each set of data points may correspond to measured values (e.g., abundance or concentration) of a substance exhibiting a dynamic biological response, measured at multiple locations along the biological sample. Each location on the baseline of the biological sample may correspond to a specific growth time of the biological sample. In some embodiments, in the example of a hair shaft, each location corresponds to a hair growth period of approximately 20 minutes (e.g., a hair growth period calculated using a laser step size of 5 micrometers and an average hair growth rate of 1 cm per month). By correlating multiple locations along the baseline of the biological sample to the corresponding periods of growth, a first dataset containing multiple traces is obtained. Each trace contains a time-dependent abundance of measured values (e.g., abundance or concentration) of a substance exhibiting a dynamic biological response, measured from the biological sample. For example, the distance between locations may correspond to the estimated growth (e.g., biological time) of the biological sample. For example, the abundance may be measured for a hair sample along a distance of 1.2 cm, where this distance corresponds to a biological time of approximately 35 days. Biological time can be estimated by using the average rate of hair growth (e.g., 1 cm per month).
[0058] In some embodiments, the data analysis of the Raman spectrum may include cosmic ray removal, background correction, spectral normalization, peak fitting, or any combination thereof.
[0059] In some embodiments, data analysis may be performed on traces corresponding to time-dependent abundances (e.g., time-dependent concentrations) of substances exhibiting dynamic biological responses measured from biological samples. This may include customized actions to clean up the data (e.g., smoothing the data over a time span and / or removing data points that are above or below a predetermined threshold). In some embodiments, data analysis includes removing from the trace data points whose mean absolute difference between adjacent data points is at least 1, 2, or 3 times the mean absolute difference between adjacent data points.
[0060] In some embodiments, the data analysis further includes a dimensionality reduction step, thereby decomposing the high-dimensional array of Raman spectra into a low-dimensional array of derived time-varying components. Methods for dimensionality reduction include independent component analysis (ICA), principal component analysis (PCA), non-negative matrix factorization (NNMF), and related unsupervised and supervised methods.
[0061] In some embodiments, the data analysis further includes performing a recurrence quantitative analysis (RQA) on the time-dependent trace or components derived from dimensionality reduction techniques (ICA / PCA) applied to the time-dependent trace to obtain a set of features that describe the dynamic periodic characteristics of the trace. The RQA measures the variation of the time-dependent trace or components derived from the time-dependent trace. The RQA involves estimation of features that describe the periodic characteristics in a given waveform, including recurrence rate, determinism, mean diagonal length, maximum diagonal length, divergence, Shannon entropy in diagonal length, recurrence tendency, lamination, trapping time, maximum vertical length, Shannon entropy in vertical length, mean recurrence time, Shannon entropy in recurrence time, most likely number of recurrences, and / or any combination thereof. For the methods and features of RQA, please refer, for example, to Webber et al. in “Simpler Methods Do It Better: Success of Recurrence Quantification Analysis as a General Purpose Data Analysis Tool,” Physics Letters A 373, 3753-3756 (2009) and by Marwan et al. in “Recurrence Plots for the Analysis of Complex Systems,” Physics Reports 438, 237-239 (2007), the contents of which are incorporated herein by reference in their entirety. In some embodiments, time-dependent traces are analyzed by using other analytical methods such as Fourier transform, wavelet analysis, and cosinol analysis. By applying such techniques, similar metrics, including spectral analysis of frequency components and their associated powers, can be derived. These metrics and related derived measures can be used instead of features derived from RQA to analyze time-dependent traces obtained from biological samples for predictive classification purposes.
[0062] RQA includes constructing recurrence plots to visualize and analyze the dynamic temporal structure of each acquired trace. Such recurrence plots can illustrate the phase-dynamic process in continuous measurements by plotting a given sequence against the time-delayed derivation of that sequence. From a one-dimensional trace measured from the hair shaft, an additional dimension is computationally derived, and the trace is embedded in a higher-dimensional space called a phase diagram, where t refers to the value of the original trace, and the dimensions (t+τ) and (t+2τ) are derived from delaying the original time series by an interval τ. Subsequent analysis is then performed on the embedded phase diagram to construct the recurrence plot and recurrence quantitative analysis. The recurrence plot can be derived from the phase diagram by applying a threshold function to each point in the phase diagram. On the corresponding recurrence plot, which consists of a square binary matrix usually represented as a white or black space, a given point is assigned a value of 1 for each time interval, and other points in the phase diagram share the spatial limit of the assigned threshold boundary. The RQA method is applied to recurrence plots to examine the delay intervals between states in a given system, where black dots reflect the time intervals when the system revisits the same state. Periodic processes in which the system continuously repeats a given pattern of states appear as black diagonals in the recurrence plot, periods of stability appear as a square structure, spurious repeats appear as black dots, and unique events appear as white spaces.
[0063] In some embodiments, the recurrence plot is constructed for tracing a single substance or a combination of two substances (for example, constructed to visualize an interactive cyclic pattern of two substances, which may be referred to as cross-recurrence quantification or joint recurrence quantification). In some embodiments, the recurrence plot is constructed for a combination of three or more substances.
[0064] In some embodiments, the data analysis includes analyzing recurrence plots to obtain a set of features associated with the recurrence plots. Features, which may be referred to synonymously as "rhythmic features" or "dynamic features", provide a quantitative measure that describes the periodicity, predictability, and transitions present in multiple traces. The features are selected from a set that includes recurrence rate, determinism, mean diagonal length, maximum diagonal length, divergence, Shannon entropy in diagonal length, recurrence tendency, layering, trapping time, maximum vertical line length, Shannon entropy in vertical line length, mean recurrence time, Shannon entropy in recurrence time, most likely number of recurrences, and / or any combination thereof.
[0065] In some embodiments, the data analysis further includes inputting the obtained set of features into a trained model. In some embodiments, the trained model includes a predictive calculation algorithm for obtaining the probability of a subject having a biological state. In some embodiments, the predictive calculation algorithm performs the following calculations.
Equation
[0066] The weight parameters β1, …, β k can be defined based on model training. The probability p(subject) may be provided as a number in the range of 0 to 1, where 1 corresponds to a 100% probability that the subject has a biological state.
[0067] In some embodiments, data analysis includes applying a threshold to the acquired probability p(subject). If the acquired probability p(subject) is above a predetermined threshold, the subject is evaluated as having a biological state. If the acquired probability is below the threshold, the subject is evaluated as not having a biological state. In some embodiments, the threshold is between approximately 0.3 and 0.6 (for example, a predetermined threshold is approximately 0.3, 0.35, 0.4, 0.45, 0.5, 0.55, or 0.6). The value assigned to the stochastic threshold may be determined or estimated during model training via the use of a receiver operating characteristic (ROC) chart, and the optimal threshold used corresponds to the value that yields the maximum area under the curve (ROC-AUC). In some embodiments, the acquired probability is expressed in terms of the associated odds (e.g., an odds ratio (OR) which can be derived from the probability, such as OR = p / (1-p)). For example, the evaluation includes evaluating the odds that the subject has a biological state.
[0068] In some embodiments, data analysis includes distinguishing a first biological state from an alternative state, e.g., a second biological state. In some embodiments, the alternative state is not associated with any known state (e.g., a typically developing state (NT)). In some embodiments, the first biological state is associated with autism spectrum disorder (ASD), and the alternative state is associated with attention deficit / hyperactivity disorder (ADHD). In some embodiments, the alternative state is any other neurodevelopmental state, or a comorbid diagnosis of two neurodevelopmental states. Thus, data analysis may be able to distinguish between two neurodevelopmental states (e.g., distinguishing between autism spectrum disorder and ADHD, or between ASD and a comorbid (CM) case diagnosed with both ASD and ADHD).
[0069] Healthcare providers, such as the patient's physician and treatment team, may have access to patient data (e.g., dynamic biological response data or other health data), and / or predictions or assessments generated from such data. Based on the results of data analysis, healthcare providers may determine clinical decisions or outcomes.
[0070] For example, a physician may instruct a patient to undergo one or more clinical trials at a hospital or other clinical setting, at least in part, based on the expected disease or disorder in the subject. These instructions may be given when certain predetermined criteria (e.g., a minimum threshold for the likelihood of disease or disorder) are met.
[0071] Such minimum thresholds could be, for example, at least about 5% probability, at least about 10% probability, at least about 20% probability, at least about 25% probability, at least about 30% probability, at least about 35% probability, at least about 40% probability, at least about 45% probability, at least about 50% probability, at least about 55% probability, at least about 60% probability, at least about 65% probability, at least about 70% probability, at least about 75% probability, at least about 80% probability, at least about 85% probability, at least about 90% probability, at least about 95% probability, at least about 96% probability, at least about 97% probability, at least about 98% probability, or at least about 99% probability.
[0072] As another example, a physician may prescribe a therapeutically effective dose of a drug (e.g., a medication), a clinical procedure, or further clinical trials to be performed on a patient, at least in part, based on the anticipated disease or disorder in the subject. For example, a physician may prescribe an anti-inflammatory drug in response to signs of inflammation in a patient.
[0073] Model The methods and systems described herein may utilize or access the external capabilities of artificial intelligence technology to develop signatures for various diseases or disorders. These signatures can be used to accurately predict diseases or disorders (e.g., months or years earlier than standard clinical care). Using such predictive capabilities, healthcare providers (e.g., physicians) can make informed and accurate risk-based decisions, thereby improving the quality of care and monitoring provided to patients.
[0074] The methods and systems of this disclosure can analyze dynamic biological response data obtained from a subject (patient) to generate the likelihood that the subject has a disease or disability. For example, the system may apply a trained (e.g., predictive) algorithm to the obtained dynamic biological response data to generate the likelihood that a subject has a disease or disability. The trained algorithm may include an artificial intelligence-based model, such as a machine learning-based classifier, configured to process the obtained dynamic biological response data to generate the likelihood that a subject has a disease or disability. The model may be trained using clinical datasets from one or more cohorts of patients, for example, using the patient's clinical health data and / or dynamic biological response data as input and the patient's known clinical health outcome (e.g., disease or disability) as output to the model.
[0075] The model may include one or more machine learning algorithms. Examples of machine learning algorithms include support vector machines (SVMs), naive Bayes classification, random forests, neural networks (e.g., deep neural networks (DNNs), recurrent neural networks (RNNs), deep RNNs, long short-term memory (LSTM) recurrent neural networks (RNNs), or gated recurrent units (GRUs), or other supervised learning algorithms, or unsupervised machine learning, statistical, or deep learning algorithms for classification and regression. The model may also include estimation of an ensemble model consisting of multiple predictive models, for example, utilizing techniques such as gradient boosting in the construction of a gradient boosted decision tree. The model may be trained using one or more training datasets corresponding to patient data.
[0076] A training dataset may be generated, for example, from one or more cohorts of patients with common clinical characteristics (features) and clinical outcomes (labels). A training dataset may include a set of features and corresponding labels. Features may correspond to algorithmic inputs, including dynamic biological response data, patient demographic information derived from electronic medical records (EMRs), and medical observations. Features may include clinical characteristics such as a specific range or category of dynamic biological response data. Features may include patient information such as the patient's age, medical history, other medical conditions, current or past medications, and time since the last observation. For example, a set of features collected from a given patient at a given time point in time may collectively function as a signature that can indicate the patient's health status or condition at that time point.
[0077] For example, the range of dynamic biological response data and other health measurements may be represented as multiple disjoint continuous ranges of continuous measurements, and the categories of dynamic biological response data and other health measurements may be represented as multiple disjoint sets of measurements (e.g., {"high", "low"}, {"high", "normal"}, {"low", "normal"}, {"high", "borderline high", "normal", "low"}, etc.). Clinical features may also include clinical markers indicating the patient's health history, such as a diagnosis of disease or disorder, previous clinical treatments (e.g., drugs, surgery, chemotherapy, radiotherapy, immunotherapy, etc.), behavioral factors, or other health conditions (e.g., a history of hypertension, hyperglycemia, hypercholesterolemia or high blood cholesterol, allergic reactions or other side effects, etc.).
[0078] The indicators may include, for example, clinical outcomes such as the presence or absence, diagnosis, or prognosis of a disease or disorder in the subject (e.g., a patient). Clinical outcomes may include temporal features associated with the presence or absence, diagnosis, or prognosis of a disease or disorder in the patient. For example, temporal features may indicate that the disease or disorder developed within a specific period following a previous clinical outcome (e.g., discharge from hospital, administration of therapeutic drugs, or receiving a clinical procedure such as surgery). Such periods may be, for example, about 1 hour, about 2 hours, about 3 hours, about 4 hours, about 6 hours, about 8 hours, about 10 hours, about 12 hours, about 14 hours, about 16 hours, about 18 hours, about 20 hours, about 22 hours, about 24 hours, about 2 days, about 3 days, about 4 days, about 5 days, about 6 days, about 7 days, about 10 days, about 2 weeks, about 3 weeks, about 4 weeks, about 1 month, about 2 months, about 3 months, about 4 months, about 6 months, about 8 months, about 10 months, about 1 year, or more than about 1 year.
[0079] Input features can be structured by aggregating data into bins, or alternatively by using one-hot encoding. The input may also include feature values or vectors derived from the aforementioned input, such as cross-correlations calculated between separate dynamic biological response data or other measurements over a fixed period, and discrete derivatives or finite differences between consecutive measurements. Such periods may be, for example, about 1 hour, about 2 hours, about 3 hours, about 4 hours, about 6 hours, about 8 hours, about 10 hours, about 12 hours, about 14 hours, about 16 hours, about 18 hours, about 20 hours, about 22 hours, about 24 hours, about 2 days, about 3 days, about 4 days, about 5 days, about 6 days, about 7 days, about 10 days, about 2 weeks, about 3 weeks, about 4 weeks, about 1 month, about 2 months, about 3 months, about 4 months, about 6 months, about 8 months, about 10 months, about 1 year, or more than about 1 year.
[0080] Training records can be constructed from a sequence of observations. Such sequences may have a fixed length to facilitate data processing. For example, a sequence may be zero-padded or selected as an independent subset of a single patient's record.
[0081] A model may process input features to produce output values that include one or more classifications, one or more predictions, or a combination thereof. For example, such classifications or predictions may include binary classifications of healthy / normal health status (e.g., absence of disease or disability) or adverse health status (e.g., presence of disease or disability), intergroup classifications of category labels (e.g., "no disease or disability," "obvious disease or disability," and "suspected disease or disability"), the likelihood of developing a particular disease or disability (e.g., relative likelihood or probability), a score indicating the presence of disease or disability, a score indicating the level of systemic inflammation experienced by the patient, a "risk factor" indicating the probability of death for the patient, a prediction of the time at which the patient is expected to develop disease or disability, and confidence intervals for any numerical predictions. Various machine learning techniques may be cascaded so that the output of the machine learning technique can also be used as input features to subsequent layers or subsections of the model.
[0082] A model can be trained using a dataset to generate real-time classifications or predictions (for example, by determining the model's weights and correlations). Such a dataset may be large enough to generate statistically significant classifications or predictions. For example, a dataset may include a database of non-identifiable data, including dynamic biological response data and other measurements, as well as dynamic biological response data and other measurements from a hospital or other clinical setting.
[0083] A dataset can be divided into subsets (e.g., discrete or overlapping) such as a training dataset, a development dataset, and a test dataset. For example, a dataset can be divided into a training dataset containing 80% of the dataset, a development dataset containing 10% of the dataset, and a test dataset containing 10% of the dataset. The training dataset may contain about 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of the dataset. The development dataset may contain about 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of the dataset. The test dataset may contain about 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of the dataset. The training set (e.g., the training dataset) may be selected by random sampling of sets of data corresponding to one or more patient cohorts to ensure sampling independence. Alternatively, the training set (e.g., the training dataset) may be selected by proportional sampling of sets of data corresponding to one or more patient cohorts to ensure sampling independence.
[0084] To improve the accuracy of model predictions and reduce model overfitting, datasets can be augmented to increase the number of samples in the training set. For example, data augmentation may include rearranging the order of observations in the training records. To address datasets with missing observations, methods for imputing missing data, such as forward fitting, back fitting, linear interpolation, and multitask Gaussian processes, may be used. To remove confounding factors, datasets can be filtered. For example, subsets of patients may be excluded from the database.
[0085] A model may include one or more neural networks, such as a neural network, a convolutional neural network (CNN), a deep neural network (DNN), a recurrent neural network (RNN), or a deep RNN. A recurrent neural network may include units that may be long short-term memory (LSTM) units or gated recurrent units (GRUs). For example, a model may include an algorithmic architecture that includes a neural network with a set of input features such as vital signs and other measurements, a patient's medical history, and / or patient demographics. Neural network techniques, such as dropout or regularization, may be used while training the model to prevent overfitting. A neural network may include multiple subnetworks, each configured to produce a classification or prediction of a different type of output information (for example, they may be combined to form the overall output of the neural network). Alternatively, machine learning models may utilize statistical algorithms or related algorithms, including random forests, classification and regression trees, support vector machines, discriminant analysis, regression techniques, and ensembles and gradient-boosted variations thereof.
[0086] When the model generates a classification or prediction of a disease or disorder, a notification (e.g., an alert or alarm) may be generated and sent to healthcare providers, such as doctors, nurses, or other members of the patient's treatment team within a hospital. The notification may be sent via automated phone calls, short message service (SMS) or multimedia messaging service (MMS) messages, email, or alerts within a dashboard. The notification may include output information such as a prediction of the disease or disorder, the likelihood of the predicted disease or disorder, the expected time to onset of the disease or disorder, a confidence interval for the likelihood or time, or a recommended course of treatment for the disease or disorder.
[0087] Different performance metrics can be generated to validate the model's performance. For example, the area under the receiver operating curve (AUROC) can be used to determine the model's diagnostic capability. For instance, the model may use adjustable classification thresholds, thereby allowing for adjustable specificity and sensitivity, and the receiver operating curve (ROC) can be used to identify different operating points corresponding to different values of specificity and sensitivity.
[0088] In some cases, such as when the dataset is not large enough, cross-validation may be performed to evaluate the robustness of the model across different training and test datasets.
[0089] To calculate performance metrics such as sensitivity, specificity, precision, positive predictor value (PPV), negative predictor value (NPV), AUPRC, and AUROC, the following definitions may be used: "False positive" may refer to an outcome in which a positive outcome or result is incorrectly or prematurely generated (e.g., before or without the actual onset of the disease or disability). "True positive" may refer to an outcome in which a positive outcome or result is correctly generated when the patient has the disease or disability (e.g., the patient exhibits symptoms of the disease or disability, or the patient's records indicate the disease or disability). "False negative" may refer to an outcome in which a negative result or result is generated, but the patient has the disease or disability (e.g., the patient exhibits symptoms of the disease or disability, or the patient's records indicate the disease or disability). "True negative" may refer to an outcome in which a negative outcome or result is generated (e.g., before or without the actual onset of the disease or disability).
[0090] The model may be trained until certain predetermined conditions for accuracy or performance are met, such as having a minimum desired value corresponding to the diagnostic accuracy metric. For example, the diagnostic accuracy metric may correspond to a prediction of the likelihood of disease or disability occurring in a subject. Alternatively, the diagnostic accuracy metric may correspond to a prediction of the likelihood of exacerbation or recurrence of a previously treated disease or disability in the subject. Examples of diagnostic accuracy metrics may include sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, area under the accuracy-recall curve (AUPRC), and area under the receiver operating characteristic (ROC) curve (AUROC), corresponding to the diagnostic accuracy of detecting or predicting disease or disability.
[0091] For example, such a given condition may have a sensitivity of predicting disease or disorder that includes values such as at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, at least about 90%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, or at least about 99%.
[0092] As another example, such a given condition may have a specificity that predicts a disease or disorder that includes values of, for example, at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, at least about 90%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, or at least about 99%.
[0093] As another example, such a given condition may have a positive predictive value (PPV) that predicts disease or impairment, which may include values of, for example, at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, at least about 90%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, or at least about 99%.
[0094] As another example, such a given condition may have a negative predictive value (NPV) that predicts disease or impairment that includes values such as at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, at least about 90%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, or at least about 99%.
[0095] As another example, such a given condition may include a receiver operating characteristic (ROC) area under the curve (AUC) (AUROC) predicting disease or impairment that is at least about 0.50, at least about 0.55, at least about 0.60, at least about 0.65, at least about 0.70, at least about 0.75, at least about 0.80, at least about 0.85, at least about 0.90, at least about 0.95, at least about 0.96, at least about 0.97, at least about 0.98, or at least about 0.99.
[0096] As another example, such a given condition may include an area under the precision-recall curve (AUPRC) predicting the disease or disorder that is at least about 0.10, at least about 0.15, at least about 0.20, at least about 0.25, at least about 0.30, at least about 0.35, at least about 0.40, at least about 0.45, at least about 0.50, at least about 0.55, at least about 0.60, at least about 0.65, at least about 0.70, at least about 0.75, at least about 0.80, at least about 0.85, at least about 0.90, at least about 0.95, at least about 0.96, at least about 0.97, at least about 0.98, or at least about 0.99.
[0097] In some embodiments, the trained model may be trained or configured to predict a disease or disorder with a sensitivity of at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, at least about 90%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, or at least about 99%.
[0098] In some embodiments, the trained model may be trained or configured to predict a disease or disorder with specificity of at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, at least about 90%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, or at least about 99%.
[0099] In some embodiments, the trained model may be trained or configured to predict a disease or disability with positive predictive values (PPVs) of at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, at least about 90%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, or at least about 99%.
[0100] In some embodiments, the trained model may be trained or configured to predict a disease or disability with negative predictive values (NPVs) of at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, at least about 90%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, or at least about 99%.
[0101] In some embodiments, the trained model may be trained or configured to predict a disease or impairment with an area under the receiver operating characteristic (ROC) curve (AUC) (AUROC) of at least about 0.50, at least about 0.55, at least about 0.60, at least about 0.65, at least about 0.70, at least about 0.75, at least about 0.80, at least about 0.85, at least about 0.90, at least about 0.95, at least about 0.96, at least about 0.97, at least about 0.98, or at least about 0.99.
[0102] In some embodiments, the trained model may be trained or configured to predict a disease or disorder having an area under the precision-recall curve (AUPRC) of at least about 0.10, at least about 0.15, at least about 0.20, at least about 0.25, at least about 0.30, at least about 0.35, at least about 0.40, at least about 0.45, at least about 0.50, at least about 0.55, at least about 0.60, at least about 0.65, at least about 0.70, at least about 0.75, at least about 0.80, at least about 0.85, at least about 0.90, at least about 0.95, at least about 0.96, at least about 0.97, at least about 0.97, at least about 0.98, or at least about 0.99.
[0103] Training datasets may be collected from training subjects (e.g., humans). Each training has a diagnostic status indicating whether or not a biological condition has been diagnosed. In some embodiments, training subjects are children aged 12 years or younger (e.g., 5, 4, 3, 2, 1, 9, 6, 3, or 1 month or younger). In some embodiments, children are approximately 12 to approximately 5 years old. In some embodiments, humans are approximately 12, 11, 10, 9, 8, 7, 5, 4, 3, 2, or under 1 year old. In some embodiments, humans are at least approximately 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 years old. The following training procedure can be performed for each training subject in multiple training subjects.
[0104] Multiple dynamic biological reaction samples can be obtained by sampling multiple positions on a baseline on the biological sample under training. Each dynamic biological reaction sample in the corresponding set of dynamic biological reaction samples represents a different position in the corresponding set of dynamic biological reaction samples, and each position in the corresponding set of dynamic biological reaction samples represents a different growth period of the corresponding biological sample. Next, each position of the biological reaction sample is analyzed (e.g., using a laser ablation inductively coupled plasma mass spectrometer (LA-ICP-MS), fluorescence image sensor, or Raman spectrometer) to obtain multiple traces. Each trace in the corresponding set of traces corresponds to a measurement of the abundance of the corresponding substance, which is determined collectively over time from the corresponding set of dynamic biological reaction samples.
[0105] Next, each second dataset may be obtained from a corresponding set of traces containing a corresponding set of features, and each feature in the corresponding set of features is determined by the change in the abundance of one or more substances in the corresponding traces, which is evaluated by applying recurrence quantification analysis or related methods to the dimensions derived from the Raman waveform via the Raman waveform or ICA / PCA or related dimensionality reduction techniques.
[0106] Next, an untrained or partially untrained model may be generated having (i) a corresponding set of features from each of the second datasets for each of the multiple training subjects, and (ii) a corresponding diagnostic state for each of the multiple training subjects, selected from a first diagnostic state and a second diagnostic state, thereby obtaining a trained model. The trained model provides an index of whether the test subject has a first biological state, based on the feature values of the set of features obtained from the biological sample of the test subject. In some embodiments, the trained model is a neural network algorithm, a support vector machine algorithm, a decision tree algorithm, an unsupervised clustering algorithm, a supervised clustering algorithm, a regression algorithm, or any combination or variation thereof, in particular including gradient-boosted implementations of the algorithms described, e.g., gradient-boosted decision trees. In some embodiments, the trained machine learning model utilizes a gradient-boosted ensemble algorithm. In some embodiments, the trained model is a multinomial or binary classifier. In some embodiments, a trained model can be used to make a binary prediction as to whether a sample was derived from an object having a first biological state, or a multinomial prediction as to distinguish an object without a diagnosis from an object having a first biological state or a second biological state different from the first biological state.
[0107] In some embodiments, the model is a neural network or a convolutional neural network. See Vincent et al., 2010, “Stacked denoising autoencoders: Learning useful representation in a deep network with a local denoising criterion,” J Mach Learn Res 11, pp. 3371-3408, Larochelle et al., 2009, “Exploring strategies for training deep neural networks,” J Mach Learn Res 10, pp. 1-40, and Hassoun, 1995, Fundamentals of Artificial Neural Networks, Massachusetts Institute of Technology. These are incorporated herein by reference, respectively.
[0108] Independent component analysis (ICA) as described herein in unsupervised dimensionality reduction of Raman waveforms is described in Lee, T.-W. (1998): Independent component analysis: Theory and applications, Boston, Mass: Kluwer Academic Publishers, ISBN 0-7923-8261-7, and Hyvarinen, A.; Karhunen, J.; Oja, E. (2001): Independent Component Analysis, New York: Wiley, ISBN 978-0-471-40540-5. These documents are incorporated herein by reference in their entirety.
[0109] The principal component analysis (PCA) described herein for unsupervised dimensionality reduction of Raman waveforms is described in Jolliffe, IT (2002). Principal Component Analysis. Springer Series in Statistics. New York: Springer-Verlag. doi:10.1007 / b98835. ISBN 978-0-387-95442-4. This document is incorporated herein by reference in its entirety.
[0110] SVM is a 5-dimensional model that is used in Cristianini and Shawe-Taylor, 2000, “An Introduction to Support Vector Machines,” Cambridge University Press, Cambridge, Boser et al., 1992, “A training algorithm for optimal margin classifiers,” in Proceedings of the 5 thThis is described in *Annual ACM Workshop on Computational Learning Theory*, ACM Press, Pittsburgh, Pa., pp. 142-152; *Vapnik*, 1998, *Statistical Learning Theory*, Wiley, New York; *Mount*, 2001, *Bioinformatics: Sequence and Genome Analysis*, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY; *Duda*, *Pattern Classification*, Second Edition, 2001, John Wiley & Sons, nc., pp. 259, 262-265; and *Hastie*, 2001, *The Elements of Statistical Learning*, Springer, New York; and *Furey et al.*, 2000, *Bioinformatics* 16, 906-914, each of which is incorporated herein by reference in its entirety. When used for classification, SVMs isolate a given set of binary-labeled data using a hyperplane as far away as possible from the labeled data. When linear separation is not possible, SVM can work in combination with "kernel" techniques that automatically implement nonlinear mapping to the feature space. The hyperplane found by the SVM in the feature space corresponds to the nonlinear decision boundary in the input space.
[0111] Decision tree classifiers are outlined by Duda, 2001, Pattern Classification, John Wiley & Sons, Inc., New York, pp. 395–396, which are incorporated herein by reference. Decision tree-based methods divide the feature space into sets of rectangles and fit a model (such as a constant) to each. In some embodiments, the decision tree is random forest regression. One particular algorithm that may be used is Classification and Regression Tree (CART). Other particular decision tree algorithms include, but are not limited to, ID3, C4.5, MART, and Random Forest. CART, ID3, and C4.5 are described by Duda, 2001, Pattern Classification, John Wiley & Sons, Inc., New York, pp. 396–408 and pp. 411–412, which are incorporated herein by reference. CART, MART, and C4.5 are described in their entirety by reference in Hastie et al., 2001, The Elements of Statistical Learning, Springer-Verlag, New York, Chapter 9. Random forests are described in their entirety by reference in Breiman, 1999, “Random Forests—Random Features,” Technical Report 567, Statistics Department, UC Berkeley, September 1999.
[0112] Clustering (e.g., unsupervised and supervised clustering model algorithms) is described on pages 211–256 of Duda and Hart, Pattern Classification and Scene Analysis, 1973, John Wiley & Sons, Inc., New York (hereinafter "Duda 1973"), which is incorporated herein by reference in its entirety. As described in Section 6.7 of Duda 1973, the clustering problem is described as one of finding natural groups within a dataset. Two problems are addressed in identifying natural groups. First, a method is determined for measuring the similarity (or dissimilarity) between two samples. This metric (similarity measure) is used to ensure that samples in one cluster are more similar to each other than samples in other clusters. Second, a mechanism is determined for partitioning the data into clusters using the similarity measure. Similarity measures are discussed in Section 6.7 of Duda 1973, and one way to begin a clustering investigation is to define a distance function and compute a matrix of distances between all sample pairs in the training set. If distance is a good measure of similarity, the distance between reference entities in the same cluster will be significantly shorter than the distance between reference entities in different clusters. However, as described on page 215 of Duda 1973, clustering does not require the use of a distance metric. For example, a nonmetric similarity function s(x,x') can be used to compare two vectors x and x'. Traditionally, s(x,x') is a symmetric function whose value is large when x and x' are "similar" in some way. An example of a nonmetric similarity function s(x,x') is provided on page 218 of Duda 1973. Once a method is chosen for measuring the "similarity" or "dissimilarity" between points in a dataset, clustering requires a baseline function to measure the clustering quality of any partition of the data. Partitions of the dataset that extremize the baseline function are used to cluster the data. See page 217 of Duda 1973.The reference function is discussed in section 6.8 of Duda 1973. More recently, see Duda et al., Pattern Classification, 2. nd The edition, published by John Wiley & Sons, Inc., New York, describes clustering in detail on pages 537–563. For further details on clustering methods, see Kaufman and Rousseeuw, 1990, Finding Groups in Data: An Introduction to Cluster Analysis, Wiley, New York, NY; Everitt, 1993, Cluster analysis (3rd ed.), Wiley, New York, NY; and Backer, 1995, Computer-Assisted Reasoning in Cluster Analysis, Prentice Hall, Upper Saddle River, New Jersey. These are incorporated herein by reference, respectively. Specific exemplary clustering techniques that may be used in this disclosure include, but are not limited to, hierarchical clustering (agglomerative clustering using nearest neighbor, farthest neighbor, mean linkage, centroid, or sum-of-squares algorithms), k-means clustering, fuzzy k-means clustering algorithms, and Jarvis-Patrick clustering. In some embodiments, the clustering includes unsupervised clustering, in which case there is no preconceived notion about which clusters should be formed when the training set is clustered.
[0113] Regression models, such as multi-category logit models, are described in their entirety in Agresti, An Introduction to Categorical Data Analysis, 1996, John Wiley & Sons, Inc., New York, Chapter 8, which is incorporated herein by reference. In some embodiments, this model utilizes the regressor model disclosed in Hastie et al., 2001, The Elements of Statistical Learning, Springer-Verlag, New York, which is incorporated herein by reference. In some embodiments, gradient boosting models are used, for example, for the classification algorithms described herein, and these gradient boosting models are described in Boehmke, Bradley; Greenwell, Brandon (2019). “Gradient Boosting”. Hands-On Machine Learning with R. Chapman & Hall. pp.221-245. ISBN 978-1-138-49568-5, which is incorporated herein by reference. In some embodiments, ensemble modeling techniques are used, for example, toward the classification algorithms described herein, and these ensemble modeling techniques are described in the implementation forms of the classification models herein and are described in Zhou Zhihua (2012). Ensemble Methods: Foundations and Algorithms. Chapman and Hall / CRC. ISBN 978-1-439-83003-1. This document is incorporated herein in its entirety by reference.
[0114] In some embodiments, machine learning analysis is performed by a device that runs one or more programs (e.g., one or more programs stored in non-persistent memory 111 or persistent memory 112 in Figure 1) that include instructions for performing data analysis. In some embodiments, data analysis is performed by a system that includes at least one processor (e.g., a processing core 102) and memory (e.g., one or more programs stored in non-persistent memory 111 or persistent memory 112) that include instructions for performing data analysis.
[0115] Computer system This disclosure provides a computer system programmed to implement the methods of this disclosure. Figure 4 shows a computer system 401 programmed, or otherwise configured, to, for example, acquire a Raman signature of a tooth sample, spatially analyze the Raman spectrum across the tooth sample, generate a time-Raman profile, process the data using a trained model, and predict the diagnostic status of a subject with respect to disease or disorder. The computer system 401 can be configured to perform various aspects of the sensor data analysis of this disclosure, such as, for example, staining a tooth sample, acquiring a fluorescence image of a stained tooth sample, spatially analyzing the fluorescence intensity across the stained tooth sample, generating a time-time profile of inflammation, measuring the dynamics of the time-time profile, processing the data using a trained model, and predicting the diagnostic status of a subject with respect to disease or disorder. The computer system 401 may be a computer system located remotely from a user's electronic device or from an electronic device. The electronic device may be a mobile electronic device.
[0116] The computer system 401 includes a central processing unit (CPU, also known herein as “processor” and “computer processor”) 405, which may be a single-core or multi-core processor, or multiple processors for parallel processing. The computer system 401 also includes memory or memory locations 410 (e.g., random-access memory, read-only memory, flash memory), an electronic storage unit 415 (e.g., a hard disk), a communication interface 420 for communicating with one or more other systems (e.g., a network adapter), and peripheral devices 425 such as a cache, other memory, data storage and / or an electronic display adapter. The memory 410, storage unit 415, interface 420, and peripheral devices 425 communicate with the CPU 405 through a communication bus (solid line), such as a motherboard. The storage unit 415 may be a data storage unit (or data repository) for storing data. The computer system 401 may be operably coupled to a computer network ("network") 430 with the help of the communication interface 420. Network 430 may be the Internet, the Internet and / or an extranet, or an intranet and / or extranet communicating with the Internet. In some cases, network 430 is a telecommunications and / or data network. Network 430 may include one or more computer servers, which can enable distributed computing such as cloud computing. In some cases, network 430 may implement a peer-to-peer network with the help of computer system 401, which may allow devices connected to computer system 401 to act as clients or servers.
[0117] The CPU 405 can execute a series of machine-readable instructions, which may be embodied in a program or software. The instructions may be stored in a memory location, such as memory 410. The instructions can instruct the CPU 405, which can then be programmed or otherwise configured to implement the methods of this disclosure. Examples of operations performed by the CPU 405 may include fetching, decoding, executing, and writing back.
[0118] The CPU 405 may be part of a circuit, such as an integrated circuit. One or more other components of system 401 may be included in the circuit. In some cases, the circuit is an application-specific integrated circuit (ASIC).
[0119] The storage unit 415 can store files such as drivers, libraries, and saved programs. The storage unit 415 can also store user data, such as user preferences and user programs. In some cases, the computer system 401 may include one or more additional data storage units located outside the computer system 401, such as those located on a remote server communicating with the computer system 401 via an intranet or the internet.
[0120] Computer system 401 can communicate with one or more remote computer systems via network 430. For example, computer system 401 can communicate with a user's (e.g., healthcare provider's) remote computer system. Examples of remote computer systems include personal computers (e.g., portable PCs), slate or tablet PCs (e.g., Apple® iPad, Samsung® Galaxy Tab), telephones, smartphones (e.g., Apple® iPhone, Android-enabled devices, Blackberry®), or personal digital assistants. Users can access computer system 401 via network 430.
[0121] The methods described herein may be implemented by machine-executable code (e.g., a computer processor) stored on an electronic storage location of a computer system 401, such as on memory 410 or an electronic storage unit 415. The machine-executable code or machine-readable code may be provided in the form of software. During use, the code may be executed by the processor 405. In some cases, the code may be retrieved from the storage unit 415 and stored in memory 410 for immediate access by the processor 405. In some situations, the electronic storage unit 415 may be excluded, and the machine-executable instructions are stored in memory 410.
[0122] The code may be pre-compiled and configured for use with a machine that has a processor adapted to run the code, or it may be compiled during runtime. The code may be provided in a programming language that can be chosen to allow execution in a pre-compiled manner or as-compiled manner.
[0123] Embodiments of systems and methods provided herein, such as computer system 401, can be embodied in programming. Various embodiments of the technology can typically be considered “products” or “manufactured articles,” which are in the form of machine (or processor) executable code and / or related data carried or embodied on some kind of machine-readable medium. Machine-executable code can be stored in memory (e.g., read-only memory, random-access memory, flash memory) or electronic storage units such as hard disks. The “storage” medium can include any or all of tangible memory such as computers and processors, or related modules of tangible memory such as various semiconductor memories, tape drives, disk drives, etc., which can at any time provide non-temporary storage for software programming. All or part of the software can be communicated through the Internet or various other telecommunication networks. Such communication can, for example, enable the loading of software from one computer or processor to another computer platform, such as from a management server or host computer to an application server. Therefore, other types of media that can carry software elements include optical waves, radio waves, and electromagnetic waves used through various air links via wired and optical fixed-line networks, through physical interfaces between local devices. Physical elements that carry such waves, such as wired or wireless links and optical links, may also be considered media that carry software. As used herein, unless limited to non-temporary tangible “storage” media, terms such as computer or machine “readable media” refer to any medium involved in providing instructions to a processor for execution.
[0124] Therefore, machine-readable media such as computer executable code can take many forms, including but not limited to tangible storage media, carrier media, or physical transmission media. Non-volatile storage media include optical or magnetic disks, such as any storage device in any computer, which may be used to implement, for example, a database as shown in the drawings. Volatile storage media include dynamic memory, such as the main memory of such a computer platform. Tangible transmission media include copper wires, including coaxial cables and wires that make up buses in computer systems, and optical fibers. Carrier media can take the form of electrical or electromagnetic signals, or acoustic or light waves, such as those generated during radio frequency (RF) and infrared (IR) data communications. Therefore, common forms of computer-readable media include, for example, floppy disks, flexible disks, hard disks, magnetic tapes, any other magnetic media, CD-ROMs, DVDs, or DVD-ROMs, any other optical media, punched card paper tapes, any other physical storage media having a pattern of holes, RAM, ROMs, PROMs, and EPROMs, FLASH-EPROMs, any other memory chips or cartridges, carriers that carry data or instructions, cables or links that carry such carriers, or any other media from which a computer can read programming code and / or data. Many of these forms of computer-readable media can be involved in transporting one or more sequences of one or more instructions to a processor for execution.
[0125] The computer system 401 includes, or can communicate with, an electronic display 435, which includes a user interface (UI) 440 for providing, for example, fluorescence image data, Raman image data, Raman spectral data, time-Raman profiles, and models. Examples of UIs include, but are not limited to, graphical user interfaces (GUIs) and web-based user interfaces.
[0126] The methods and systems of this disclosure can be implemented by one or more algorithms. The algorithms can be implemented by software at runtime by the central processing unit 405. For example, the algorithms can acquire Raman images of a dental sample, spatially analyze the Raman spectrum across the dental sample, generate a time-Raman profile, process the data using a trained model, and predict the subject's diagnostic status for disease or disorder. [Examples]
[0127] Example 1: Dynamic Raman spectroscopy profile in a dental sample for determining the risk of autism spectrum disorder (ASD) Using the methods and systems described herein, dynamic Raman spectroscopy profiles were generated in dental samples and subsequently analyzed to determine disease risk in subjects. Generally, the temporal dynamics of biological responses (e.g., physiological responses) were found to be imprinted on the sample (e.g., dental samples). By analyzing these temporal dynamics, the disease risk of subjects can be determined. Dynamic Raman spectroscopy profiles were generated in two sets of children: a first set with autism spectrum disorder and a second set without autism spectrum disorder (ASD), during periods including fetal (prenatal) development and infancy. Analysis of the dynamic Raman spectroscopy profiles revealed novel features. These features accurately distinguished cases of autism from controls. For example, juvenile spectroscopy signatures were found to reveal disease risk for ASD later in life. In contrast, the clinical diagnosis of autism is typically made around 3-4 years of age.
[0128] Primary tooth samples were obtained from each pediatric subject. The tooth samples were segmented, and Raman spectroscopy signals were measured on the tooth samples to develop time-Raman spectroscopy profiles showing prenatal and postnatal physiological responses. The time-Raman profiles were analyzed using the machine learning algorithms of this disclosure to train highly accurate models for determining disease risk (e.g., autism).
[0129] Figure 5 shows an example of classifier accuracy for diagnosing autism spectrum disorder (ASD) using features derived from the application of RQA to dimensions derived by ICA of Raman waveforms, as shown by an experimental receiver operating characteristic (ROC) curve to evaluate the accuracy of a disclosed method for evaluating subjects with autism spectrum disorder. The ROC curve can be used to evaluate the performance of a binary classifier. The ROC curve is plotted as sensitivity (also called true positive rate) against specificity (also called true negative rate). A perfect classifier may have 100% sensitivity and 100% specificity, as well as an area under the curve (AUC) of 1.0. As shown in Figure 5, a classifier configured to determine the presence of ASD in a subject based on a dynamic Raman RQA dynamic profile had a receiver operating characteristic (ROC) area under the curve (AUC) of 0.861, with a 95% confidence interval (CI) of 0.769–0.954. Receiver Operating Characteristics (ROC) describe how the sensitivity and specificity values of a classifier, as varying thresholds, are assigned to probabilistic predictions.
[0130] Therefore, analysis of Raman spectroscopic signatures using the methods and systems of this disclosure has successfully determined the disease risk of autism with a deviation of 0.86 AUC using only spectroscopic signatures measured on non-invasively acquired biological samples (e.g., dental samples) from pediatric subjects. These results demonstrate that the dynamics of early human physiology are later associated with disease, and that this can be accurately detected and profiled using the methods and systems of this disclosure.
[0131] Example 2: Dynamic Raman spectroscopy profile in a dental sample for determining the risk of amyotrophic lateral sclerosis (ALS) Using the methods and systems described herein, dynamic Raman spectroscopic profiles were generated in dental samples and subsequently analyzed to determine disease risk in subjects. Generally, the temporal dynamics of biological responses (e.g., physiological responses) are found to be imprinted on the sample (e.g., dental samples), and these temporal dynamics can be analyzed to determine disease risk in subjects. Dynamic Raman spectroscopic profiles were generated in two sets of adults—a first set with amyotrophic lateral sclerosis (ALS) and a second set without ALS—over a period including infancy and adolescence. Analysis of the dynamic Raman spectroscopic profiles revealed novel features that accurately distinguished ALS cases from controls. For example, juvenile spectroscopic signatures were found to reveal the disease risk of ALS later in life.
[0132] Permanent tooth samples were obtained from each adult subject. The tooth samples were segmented, and Raman spectral signals were measured on the tooth samples to develop time-Raman spectral profiles showing physiological responses in infancy and adolescence. The time profiles were analyzed using the machine learning algorithms of this disclosure to train highly accurate models for determining disease risk (e.g., ALS).
[0133] Figure 6 shows an example of classifier accuracy for diagnosing ALS, utilizing features derived from the application of RQA to dimensions derived by ICA of Raman waveforms, as shown by an experimental receiver operating characteristic (ROC) curve for evaluating the accuracy of a disclosed method for evaluating subjects with autism spectrum disorder. The ROC curve can be used to evaluate the performance of a binary classifier. The ROC curve is plotted as sensitivity (also called true positive rate) against specificity (also called true negative rate). A perfect classifier may have 100% sensitivity and 100% specificity, as well as an area under the curve (AUC) of 1.0. As shown in Figure 6, a classifier configured to determine the presence of ASD in subjects based on a dynamic Raman RQA dynamic profile had a receiver operating characteristic (ROC) area under the curve (AUC) of 0.880, with a 95% confidence interval (CI) of 0.658 to 1.000. The receiver operating characteristic (ROC) shows how the sensitivity and specificity values of the classifier, as varying thresholds, are assigned to probabilistic predictions.
[0134] Therefore, analysis of Raman spectroscopic signatures using the methods and systems of this disclosure has successfully determined the disease risk of ALS with a spectral signature greater than 0.88 AUC, using only spectroscopic signatures measured on biological samples from adults (e.g., dental samples). These results demonstrate that the dynamics of human physiology in adolescence are later associated with disease, and that this can be accurately detected and profiled using the methods and systems of this disclosure.
[0135] Methods described elsewhere in this specification represent steps or sets of operations according to embodiments, but those skilled in the art will recognize many variations based on the teachings set forth herein. These steps can be completed in different orders. Steps can be added or omitted. Some steps may include substeps. Many steps can be repeated any number of times if beneficial.
[0136] One or more steps of each method or set of operations may be performed using one or more circuits as described herein, such as logic circuits, for example, a processor for a field-programmable gate array or programmable array logic. For example, a circuit may be programmed to provide one or more steps of each method or set of operations, and the program may include program instructions stored in computer-readable memory, or programmed steps of logic circuits, such as programmable array logic or a field-programmable gate array.
[0137] While preferred embodiments of the present invention are shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided only as examples. The present invention is not intended to be limited by any specific examples provided herein. Although the present invention has been described with reference to the foregoing specification, the descriptions and illustrations of embodiments herein are not intended to be constrained. Numerous variations, modifications, and substitutions will be conjured here upon those skilled in the art without departing from the present invention. Furthermore, it should be understood that all aspects of the present invention are not limited to any specific descriptions, configurations, or relative proportions described herein, depending on various conditions and variables. It should be understood that various alternatives to the embodiments of the present invention described herein may be used in the practice of the present invention. Accordingly, the present invention is intended to cover such alternatives, modifications, variations, or equivalents. The following claims define the scope of the present invention, and methods and structures within the scope of these claims and their equivalents are intended to be encompassed thereby.
[0138] Embodiment Embodiment 1. A method for predicting the diagnostic status of a subject with respect to a disease or disorder, comprising: (a) exposing a biological sample of the subject to a light source; (b) obtaining a plurality of Raman spectra from the biological sample; (c) processing the plurality of Raman spectra to generate a spatial map of the plurality of Raman spectra; and (d) predicting the diagnostic status of the subject with respect to a disease or disorder based at least partially on the spatial map of the plurality of Raman spectra. Embodiment 2. The method according to Embodiment 1, wherein the biological sample includes a tooth sample, a hair sample, a nail sample, or any combination thereof. Embodiment 3. The method according to Embodiment 1 or 2, further comprising detecting or monitoring changes in the temporal stress profile of a spatial map showing the temporal response of a subject. Embodiment 4. The method according to Embodiment 3, wherein the temporal response includes a biological response, a physiological response, an anatomical response, a therapeutic response, a stress-related response, or a combination thereof. Embodiment 5. The method according to any one of Embodiments 1 to 4, wherein the multiple Raman spectra include wavenumbers from about 200 to about 3700. Embodiment 6. The method according to any one of Embodiments 1 to 5, wherein obtaining the result involves using a Raman spectroscopic microscope. Embodiment 7. The method according to Embodiment 6, wherein the Raman spectroscopic microscope includes a 50x air-coupled objective lens, a 63x water-immersion-coupled objective lens, or any combination thereof. Embodiment 8. The method according to any one of Embodiments 1 to 7, wherein the light source includes a laser, and the laser has a wavelength of about 785 nm, a wavelength of about 532 nm, or any combination thereof. Embodiment 9. The method according to any one of Embodiments 1 to 8, wherein acquisition is performed using an integration time of approximately 0.2 seconds to approximately 0.3 seconds. Embodiment 10. The method according to any one of Embodiments 1 to 9, wherein the acquisition of a Raman spectrum from among several Raman spectra is performed, and then the biological sample is moved with a step size of about 2 microns to about 5 microns. Embodiment 11. The method according to any one of Embodiments 1 to 10, wherein the disease or disorder includes autism spectrum disorder (ASD), attention deficit / hyperactivity disorder (ADHD), amyotrophic lateral sclerosis (ALS), schizophrenia, irritable bowel disease (IBD), childhood kidney disease, kidney transplant rejection, childhood cancer, or any combination thereof. Embodiment 12. The method according to any one of Embodiments 1 to 10, wherein the disease or disorder includes ASD. Embodiment 13. The method according to any one of Embodiments 1 to 12, wherein predicting the diagnostic status of an object relating to a disease or disorder includes processing a spatial map using a trained model. Embodiment 14. The method according to Embodiment 13, wherein the trained model is selected from the group consisting of neural network algorithms, support vector machine algorithms, decision tree algorithms, unsupervised clustering algorithms, supervised clustering algorithms, regression algorithms, gradient boosting algorithms, and any combination thereof. Embodiment 15. The method according to Embodiment 13, wherein the trained model includes a gradient-boosted ensemble model. Embodiment 16. The method according to Embodiment 13, wherein the trained model is configured to process one or more features selected from the group consisting of laminarity, entropy, trapping time (TT), mean diagonal length (MDL), recurrence time (RT), Vmax, determinism, Lmax, and any combination thereof. Embodiment 17. The method according to Embodiment 16, wherein the trained model is configured to process two or more features selected from the group consisting of laminarity, entropy, trapping time (TT), mean diagonal length (MDL), recurrence time (RT), Vmax, determinism, Lmax, and any combination thereof. Embodiment 18. The method according to any one of Embodiments 1 to 17, wherein a trained model predicts a diagnostic status for a disease or disorder with a sensitivity of at least about 80%. Embodiment 19. The method according to Embodiment 1, wherein a trained model predicts a diagnostic status for a disease or disorder with at least about 80% specificity. Embodiment 20. The method according to Embodiment 1, wherein a trained model predicts a diagnostic status for a disease or disorder with a positive prediction of at least about 80%. Embodiment 21. The method according to Embodiment 1, wherein a trained model predicts a diagnostic status for a disease or disorder with at least about 80% negative prediction accuracy. Embodiment 22. The method according to Embodiment 1, wherein the trained model predicts a diagnostic status for a disease or disorder with an area under the receiver operating characteristic curve (AUROC) of at least about 0.80. Embodiment 23. A device comprising one or more processors and a memory for storing one or more programs for execution by the one or more processors, wherein one or more programs include instructions for performing: (a) sampling each of a plurality of locations along a baseline on a biological sample of a subject associated with a Raman signature of a subject, thereby obtaining a plurality of Raman spectra, wherein each Raman spectrum in the plurality of Raman spectra corresponds to a different location in the plurality of locations, and each location in the plurality of locations represents a different growth period of the biological sample associated with the Raman signature; (b) analyzing each of the plurality of Raman spectra across the entire baseline on the biological sample, thereby obtaining a first dataset; (c) deriving each second dataset from the corresponding plurality of Raman spectral measurements, wherein each feature in the corresponding set of features is determined by a continuous variation in the Raman spectrum; and (d) processing the features using a trained model to predict the diagnostic status of a subject with respect to a disease or disorder associated with a Raman signature. Embodiment 24. The device according to Embodiment 23, wherein the biological sample includes a tooth sample, a hair sample, a nail sample, or any combination thereof. Embodiment 25. The device according to Embodiment 23 or 24, further comprising detecting or monitoring changes in the Raman spectrum across multiple locations exhibiting a temporal response of a subject. Embodiment 26. The device according to Embodiment 25, wherein the temporal response includes a biological response, a physiological response, an anatomical response, a therapeutic response, a stress-related response, or a combination thereof. Embodiment 27. The device according to any one of Embodiments 23 to 26, wherein the multiple Raman spectra include wavenumbers from about 200 to about 3700. Embodiment 28. The device according to any one of Embodiments 23 to 27, wherein sampling includes using a Raman spectroscopic microscope. Embodiment 29. The device according to Embodiment 28, wherein the Raman spectroscopic microscope includes a 50x air-coupled objective lens, a 63x water-immersion coupled objective lens, or any combination thereof. Embodiment 30. The device according to Embodiment 23, wherein sampling includes exposing a biological sample to a light source to generate a Raman spectrum from among multiple Raman spectra at multiple locations. Embodiment 31. The device according to Embodiment 30, wherein the light source includes a laser, the laser having a wavelength of about 785 nm, a wavelength of about 532 nm, or any combination thereof. Embodiment 32. The device according to any one of Embodiments 23 to 31, wherein the instruction further includes a translation, the translation of which, after obtaining a Raman spectrum from a plurality of Raman spectra, moves a biological sample from a first position to a second position of a plurality of positions with a step size of about 2 microns to about 5 microns. Embodiment 33. The device according to Embodiment 32, wherein the translation is performed using an integration time of approximately 0.2 seconds to approximately 0.3 seconds. Embodiment 34. The device according to any one of Embodiments 23 to 33, wherein the disease or disorder includes autism spectrum disorder (ASD), attention deficit / hyperactivity disorder (ADHD), amyotrophic lateral sclerosis (ALS), schizophrenia, irritable bowel disease (IBD), pediatric renal disease, kidney transplant rejection, childhood cancer, or any combination thereof. Embodiment 35. The device according to any one of Embodiments 23 to 33, wherein the disease or disorder includes autism spectrum disorder (ASD). Embodiment 36. The device according to any one of Embodiments 23 to 35, wherein predicting the diagnostic status of an object relating to a disease or disorder involves processing changes in Raman spectra across multiple locations using a trained model. Embodiment 37. The device according to Embodiment 36, wherein the trained model is selected from the group consisting of neural network algorithms, support vector machine algorithms, decision tree algorithms, unsupervised clustering algorithms, supervised clustering algorithms, regression algorithms, gradient boosting algorithms, and any combination thereof. Embodiment 38. The device according to Embodiment 36, wherein the trained model includes a gradient-boosted ensemble model. Embodiment 39. The device according to Embodiment 36, wherein the trained model is configured to process one or more features selected from the group consisting of laminarity, entropy, trapping time (TT), mean diagonal length (MDL), recurrence time (RT), Vmax, determinism, Lmax, and any combination thereof. Embodiment 40. The device according to Embodiment 36, wherein the trained model is configured to process two or more features selected from the group consisting of laminarity, entropy, trapping time (TT), mean diagonal length (MDL), recurrence time (RT), Vmax, determinism, Lmax, and any combination thereof. Embodiment 41. The device according to Embodiment 23, wherein a trained model predicts a diagnostic status for a disease or disorder with at least about 80% sensitivity. Embodiment 42. The device according to Embodiment 23, wherein a trained model predicts a diagnostic status for a disease or disorder with at least about 80% specificity. Embodiment 43. The device according to Embodiment 23, wherein a trained model predicts a diagnostic status for a disease or disorder with a positive prediction of at least about 80%. Embodiment 44. The device according to Embodiment 23, wherein a trained model predicts a diagnostic status for a disease or disorder with at least about 80% negative prediction accuracy. Embodiment 45. The device according to Embodiment 23, wherein a trained model predicts a diagnostic status for a disease or disorder with an area under the receiver operating characteristic curve (AUROC) of at least about 0.80. Embodiment 46. A non-temporary computer-readable storage medium and one or more computer programs embedded in the non-temporary computer-readable storage medium, wherein when one or more computer programs are executed by a computer system, the computer system (a) samples each of each of a plurality of locations along a baseline on a biological sample of a subject associated with a Raman signature of the subject, thereby obtaining a plurality of Raman spectra, wherein each Raman spectrum in the plurality of Raman spectra corresponds to a different location in the plurality of locations, and each location in the plurality of locations corresponds to a biological sample associated with a Raman signature. A non-temporary computer-readable storage medium that performs a method comprising: (b) sampling representing different growth periods of a sample; (c) analyzing each of a plurality of Raman spectra across a baseline on a biological sample to obtain a first dataset; (d) deriving each second dataset from the corresponding plurality of Raman spectral measurements, wherein each feature in the corresponding set of features is determined by continuous variation in the Raman spectrum; and (d) processing the features using a trained model to predict the diagnostic status of a subject with respect to a disease or disorder associated with the Raman signature. Embodiment 47. A non-temporary computer-readable storage medium according to Embodiment 46, wherein the biological sample includes a tooth sample, a hair sample, a nail sample, or any combination thereof. Embodiment 48. A non-temporary computer-readable storage medium according to Embodiment 46 or 47, further comprising detecting or monitoring changes in Raman spectra across multiple locations exhibiting a temporal response of a subject. Embodiment 49. A non-temporary computer-readable storage medium according to Embodiment 48, wherein the temporal response includes a biological response, a physiological response, an anatomical response, a therapeutic response, a stress-related response, or a combination thereof. Embodiment 50. A non-temporary computer-readable storage medium according to any one of Embodiments 46 to 49, wherein the plurality of Raman spectra include wavenumbers from about 200 to about 3700. Embodiment 51. A non-temporary computer-readable storage medium according to any one of Embodiments 46 to 50, wherein sampling includes using a Raman spectroscopic microscope. Embodiment 52. A non-temporary computer-readable storage medium according to Embodiment 51, wherein the Raman spectroscopic microscope includes a 50x air-coupled objective lens, a 63x water-immersion coupled objective lens, or any combination thereof. Embodiment 53. A non-temporary computer-readable storage medium according to any one of Embodiments 46 to 52, wherein sampling includes exposing a biological sample to a light source to generate a Raman spectrum from a plurality of Raman spectra at a plurality of locations. Embodiment 54. A non-temporary computer-readable storage medium according to Embodiment 53, wherein the light source includes a laser, the laser having a wavelength of about 785 nm, a wavelength of about 532 nm, or any combination thereof. Embodiment 55. A non-temporary computer-readable storage medium according to any one of Embodiments 46 to 54, wherein the instruction further includes a translation, the translation of which moves a biological sample from a first position to a second position of a plurality of positions with a step size of about 2 microns to about 5 microns, following the acquisition of Raman spectra of a plurality of Raman spectra. Embodiment 56. A non-temporary computer-readable storage medium according to Embodiment 55, wherein the translation is performed using an integration time of approximately 0.2 seconds to approximately 0.3 seconds. Embodiment 57. A non-temporary computer-readable storage medium according to any one of Embodiments 46 to 56, wherein the disease or disorder includes autism spectrum disorder (ASD), attention deficit / hyperactivity disorder (ADHD), amyotrophic lateral sclerosis (ALS), schizophrenia, irritable bowel disease (IBD), pediatric renal disease, kidney transplant rejection, childhood cancer, or any combination thereof. Embodiment 58. A non-temporary computer-readable storage medium according to any one of Embodiments 46 to 56, wherein the disease or disorder includes autism spectrum disorder (ASD). Embodiment 59. A non-temporary computer-readable storage medium according to any one of Embodiments 46 to 58, wherein predicting the diagnostic status of an object relating to a disease or disorder involves processing changes in a Raman spectrum across multiple locations using a trained model. Embodiment 60. A non-temporary computer-readable storage medium according to Embodiment 59, wherein the trained model is selected from the group consisting of neural network algorithms, support vector machine algorithms, decision tree algorithms, unsupervised clustering algorithms, supervised clustering algorithms, regression algorithms, gradient boosting algorithms, and any combination thereof. Embodiment 61. A non-temporary computer-readable storage medium according to Embodiment 59, wherein the trained model includes a gradient-boosted ensemble model. Embodiment 62. A non-temporary computer-readable storage medium according to Embodiment 59, wherein the trained model is configured to process one or more features selected from the group consisting of laminarity, entropy, trapping time (TT), mean diagonal length (MDL), recurrence time (RT), Vmax, determinism, Lmax, and any combination thereof. Embodiment 63. A non-temporary computer-readable storage medium according to Embodiment 59, wherein the trained model is configured to process two or more features selected from the group consisting of laminarity, entropy, trapping time (TT), mean diagonal length (MDL), recurrence time (RT), Vmax, determinism, Lmax, and any combination thereof. Embodiment 64. A non-temporary computer-readable storage medium according to Embodiment 46, wherein a trained model predicts a diagnostic status relating to a disease or disorder with a sensitivity of at least about 80%. Embodiment 65. A non-temporary computer-readable storage medium according to Embodiment 46, wherein a trained model predicts a diagnostic status relating to a disease or disorder with at least about 80% specificity. Embodiment 66. A non-temporary computer-readable storage medium according to Embodiment 46, further comprising the instruction predicting a patient's diagnostic status for a disease or disorder with a positive prediction of at least about 80%. Embodiment 67. A non-temporary computer-readable storage medium according to Embodiment 46, wherein a trained model predicts a diagnostic status for a disease or disorder with a positive prediction of at least about 80%. Embodiment 68. A non-temporary computer-readable storage medium according to Embodiment 46, wherein a trained model predicts a diagnostic status for a disease or disorder with at least about 80% negative prediction accuracy. Embodiment 69. A method for training a model, comprising a computer system having one or more processors and a memory for storing one or more programs to be executed by the one or more processors, wherein (a) for each of the multiple training subjects, a first subset of the training subjects has a first diagnostic state corresponding to having a first biological state associated with a Raman signature, and a second subset of the training subjects has a second diagnostic state corresponding to not having a first biological state associated with a Raman signature, (i) sampling each of the multiple locations along a baseline on the biological sample of the subject associated with the Raman signature, thereby obtaining multiple Raman spectra, wherein each Raman spectrum in the multiple Raman spectra corresponds to a different location in the multiple locations, and each location in the multiple locations corresponds to a different growth period of the biological sample of the subject associated with the Raman signature. A method comprising: (ii) sampling; (iii) analyzing each Raman spectrum across a baseline on a biological sample to obtain a first dataset; (iii) deriving each second dataset from a plurality of corresponding Raman spectra, wherein each feature in the corresponding set of features is determined by a continuous variation in the Raman spectrum; and (b) training an untrained or partially untrained model using (i) the corresponding set of features in each second dataset for each of the plurality of training subjects, and (ii) the corresponding diagnostic state for each of the plurality of training subjects, selected from a first diagnostic state and a second diagnostic state, thereby obtaining a trained model that provides an index of whether the test subject has a first biological state associated with a Raman signature, based on the value of the features in the set of features obtained from a biological sample associated with a Raman signature of the test subject. Embodiment 70. The method according to Embodiment 69, wherein the trained model is selected from the group consisting of neural network algorithms, support vector machine algorithms, decision tree algorithms, unsupervised clustering algorithms, supervised clustering algorithms, regression algorithms, gradient boosting algorithms, and any combination thereof. Embodiment 71. The method according to Embodiment 69, wherein the trained model is a multi-classifier. Embodiment 72. The method according to Embodiment 69, wherein the trained model is a binary classifier. Embodiment 73. The method according to Embodiment 69, wherein the first biological condition is selected from the group consisting of autism spectrum disorder (ASD), attention deficit / hyperactivity disorder (ADHD), amyotrophic lateral sclerosis (ALS), schizophrenia, irritable bowel disease (IBD), pediatric renal disease, kidney transplant rejection, and pediatric cancer. Embodiment 74. The method according to any one of Embodiments 69 to 73, further comprising evaluating a test subject for a first biological state associated with a Raman signature, thereby distinguishing between a first biological state associated with a Raman signature and a second biological state associated with a different Raman signature than the first biological state associated with a Raman signature. Embodiment 75. The method according to Embodiment 74, wherein the first biological state is autism spectrum disorder and the second biological state is attention deficit / hyperactivity disorder. Embodiment 76. The method according to any one of Embodiments 69 to 75, wherein the test subject is a human. Embodiment 77. The method according to Embodiment 76, wherein the human is under 12 years of age. Embodiment 78. The method according to Embodiment 76, wherein the human is less than one year old. Embodiment 79. The method according to any one of Embodiments 69 to 78, wherein the corresponding biological specimen associated with each training subject's Raman signature is selected from the group consisting of hair shafts, teeth, and nails. Embodiment 80. The method according to Embodiment 79, wherein the corresponding biological sample associated with each training subject's Raman signature is a hair shaft, and the reference line corresponds to the longitudinal direction of the hair shaft. Embodiment 81. The method according to Embodiment 79, wherein the corresponding biological specimen associated with each training subject's Raman signature is a tooth, and the reference line corresponds to a direction across the growth zone, including the neonatal line of the tooth. Embodiment 82. The method according to any one of Embodiments 69 to 81, wherein the corresponding multiple positions are arranged such that the first position at a corresponding multiple positions along the corresponding biological sample of each training subject corresponds to the position closest to the tip of the corresponding biological sample of each training subject. Embodiment 83. The method according to any one of Embodiments 69 to 82, wherein each trace in a plurality of corresponding Raman spectral measurements includes a plurality of data points, each data point being an instance of a respective position in a plurality of locations. Embodiment 84. The method according to any one of Embodiments 69 to 83, wherein the corresponding set of features is selected from the group consisting of laminarity, entropy, trapping time (TT), mean diagonal length (MDL), recurrence time (RT), Vmax, determinism, and Lmax. Embodiment 85. The method according to any one of Embodiments 69 to 83, wherein the corresponding plurality of positions include at least positions 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, or 5000, 5500, 6000, 6500, 7000, 7500, 8000, 8500, 9000, 9500, 10000, or more than 10000.
Claims
1. 1. A method for predicting a subject's diagnostic status with respect to a disease or disorder, comprising: (a) exposing a biological sample of a subject to a light source; (b) acquiring a plurality of Raman spectra from the biological sample; and (c) processing the plurality of Raman spectra to generate a spatial map of the plurality of Raman spectra; and (d) predicting a diagnostic status of the subject with respect to the disease or disorder based at least in part on the spatial map of the plurality of Raman spectra.
2. 10. The method of claim 1, wherein the biological sample comprises a tooth sample, a hair sample, a nail sample, or any combination thereof.
3. The method of claim 1 , further comprising detecting or monitoring changes in a temporal stress profile of the spatial map indicative of the subject's temporal response.
4. The method of claim 3 , wherein the temporal response comprises a biological response, a physiological response, an anatomical response, a therapeutic response, a stress-related response, or a combination thereof.
5. 2. The method of claim 1, wherein the plurality of Raman spectra comprises wavenumbers from about 200 to about 3700.
6. The method of claim 1 , wherein obtaining comprises using a Raman spectroscopic microscope.
7. 7. The method of claim 6, wherein the Raman spectromicroscope comprises a 50x air-coupled objective, a 63x water-immersion coupled objective, or any combination thereof.
8. The method of claim 1 , wherein the light source comprises a laser, the laser comprising a wavelength of about 785 nm, a wavelength of about 532 nm, or any combination thereof.
9. The method of claim 1 , wherein the acquiring is performed using an integration time of about 0.2 seconds to about 0.3 seconds.
10. 10. The method of claim 1, wherein the acquiring comprises moving the biological sample with a step size of about 2 microns to about 5 microns after acquiring a Raman spectrum of the plurality of Raman spectra.
11. 2. The method of claim 1, wherein the disease or disorder comprises autism spectrum disorder (ASD), attention-deficit / hyperactivity disorder (ADHD), amyotrophic lateral sclerosis (ALS), schizophrenia, irritable bowel disease (IBD), pediatric kidney disease, kidney transplant rejection, pediatric cancer, or any combination thereof.
12. 10. The method of claim 1, wherein the disease or disorder comprises ASD.
13. 10. The method of claim 1, wherein predicting the subject's diagnostic status with respect to the disease or disorder comprises processing the spatial map using a trained model.
14. 14. The method of claim 13, wherein the trained model is selected from the group consisting of a neural network algorithm, a support vector machine algorithm, a decision tree algorithm, an unsupervised clustering algorithm, a supervised clustering algorithm, a regression algorithm, a gradient boosting algorithm, and any combination thereof.
15. The method of claim 13 , wherein the trained model comprises a gradient boosted ensemble model.
16. 14. The method of claim 13, wherein the trained model is configured to process one or more features selected from the group consisting of stratiformity, entropy, trapping time (TT), mean diagonal length (MDL), recurrence time (RT), Vmax, determinism, Lmax, and any combination thereof.
17. 1. A device comprising one or more processors and a memory storing one or more programs for execution by said one or more processors, said one or more programs comprising: (a) sampling each respective position in a plurality of positions along a reference line on a biological sample of a subject associated with a Raman signature of the subject, thereby obtaining a plurality of Raman spectra, each Raman spectrum in the plurality of positions corresponding to a different position in the plurality of positions, each position in the plurality of positions representing a different growth period of the biological sample associated with the Raman signature; (b) analyzing each of the plurality of Raman spectra across a reference line on the biological sample, thereby obtaining a first data set; (c) deriving respective second data sets from the corresponding plurality of the Raman spectral measurements, each respective feature in the corresponding set of features being determined by a successive variation in the Raman spectrum; (d) processing the features using the trained model to predict a subject's diagnostic status for a disease or disorder associated with the Raman signature.
18. 20. The device of claim 17, wherein the biological sample comprises a tooth sample, a hair sample, a nail sample, or any combination thereof.
19. 20. The device of claim 17, wherein the instructions further comprise detecting or monitoring a change in the Raman spectrum across the multiple locations indicative of a temporal response of the subject.
20. 20. The device of claim 19, wherein the temporal response comprises a biological response, a physiological response, an anatomical response, a therapeutic response, a stress-related response, or a combination thereof.
21. 18. The device of claim 17, wherein the plurality of Raman spectra comprises wavenumbers from about 200 to about 3700.
22. The device of claim 17 , wherein sampling comprises using a Raman spectroscopic microscope.
23. 23. The device of claim 22, wherein the Raman spectromicroscope comprises a 50x air-coupling objective, a 63x water-immersion coupling objective, or any combination thereof.
24. 20. The device of claim 17, wherein the sampling comprises exposing the biological sample to a light source to generate the Raman spectrum of the plurality of Raman spectra at the plurality of locations.
25. 25. The device of claim 24, wherein the light source comprises a laser, the laser comprising a wavelength of about 785 nm, a wavelength of about 532 nm, or any combination thereof.
26. 20. The device of claim 17, wherein the instructions further comprise translating, the translating comprising moving the biological sample from a first location to a second location of the plurality of locations with a step size of about 2 microns to about 5 microns after acquiring a Raman spectrum of the plurality of Raman spectra.
27. The device of claim 26, wherein the translation is performed using an integration time of about 0.2 seconds to about 0.3 seconds.
28. 18. The device of claim 17, wherein the disease or disorder comprises autism spectrum disorder (ASD), attention-deficit / hyperactivity disorder (ADHD), amyotrophic lateral sclerosis (ALS), schizophrenia, irritable bowel disease (IBD), pediatric kidney disease, kidney transplant rejection, childhood cancer, or any combination thereof.
29. A non-transitory computer-readable storage medium and one or more computer programs embodied in the non-transitory computer-readable storage medium, the one or more computer programs, when executed by a computer system, causing the computer system to: (a) sampling each respective position in a plurality of positions along a reference line on a biological sample of a subject associated with a Raman signature of the subject, thereby obtaining a plurality of Raman spectra, each Raman spectrum in the plurality of positions corresponding to a different position in the plurality of positions, each position in the plurality of positions representing a different growth period of the biological sample associated with the Raman signature; (b) analyzing each of the plurality of Raman spectra across a reference line on the biological sample, thereby obtaining a first data set; (c) deriving respective second data sets from the corresponding plurality of the Raman spectral measurements, each respective feature in the corresponding set of features being determined by a successive variation in the Raman spectrum; (d) processing the features using the trained model to predict a subject's diagnostic status with respect to a disease or disorder associated with the Raman signature.
30. 1. A method for training a model, comprising:
1. A computer system having one or more processors and a memory storing one or more programs for execution by said one or more processors, (a) for each respective training object in a plurality of training objects, a first subset of training objects in the plurality of training objects having a first diagnostic status corresponding to having a first biological state associated with a Raman signature, and a second subset of training objects in the plurality of training objects having a second diagnostic status corresponding to not having the first biological state associated with the Raman signature; (i) sampling each respective position in a plurality of positions along a reference line on the biological sample of the object associated with the Raman signature of the object, thereby obtaining a plurality of Raman spectra, each Raman spectrum in the plurality of positions corresponding to a different position in the plurality of positions, each position in the plurality of positions representing a different growth period of the biological sample of the object associated with the Raman signature; (ii) analyzing each Raman spectrum across a reference line on the biological sample, thereby obtaining a first data set; (iii) deriving respective second data sets from the corresponding plurality of Raman spectra, each respective feature in the corresponding set of features being determined by a successive variation in the Raman spectra; (b) training an untrained or partially untrained model using (i) the corresponding set of features of each respective second dataset for each training subject in the plurality of training subjects and (ii) the corresponding diagnostic state of each training subject in the plurality of training subjects selected from among the first diagnostic state and the second diagnostic state, thereby obtaining a trained model that provides an indication as to whether the test subject has the first biological state associated with the Raman signature based on feature values in a set of features acquired from a biological sample associated with the Raman signature of a test subject.
