Predicting mortality from biological data

Abstract

This disclosure describes systems implemented as computer programs that can provide a variety of information for a subject by processing biological (e.g., epigenetic) data, optionally in combination with image data and / or individualized data.

Description

PREDICTING MORTALITY FROM BIOLOGICAL DATACROSS-REFERENCEThis application claims the benefit of priorities under 35 U.S.C. § 119(e) to U.S. Application No. 63 / 733,949 filed on December 13, 2024 and 63 / 780,073, filed on March 28, 2025. The disclosures of the aforementioned applications are hereby incorporated by reference in their entireties.BACKGROUNDThis specification relates to processing multi-modal biological data for a subject using one or more algorithms.In particular, the one or more algorithms can be a model configured to receive an input and generate an output, e.g., a predicted output, based on the received input. For example, models are parametric models and generate the output based on the received input and on values of the parameters of the model.Some models are machine learning models that employ multiple layers of models to generate an output for a received input. For example, a deep neural network is a deep machine learning model that includes an output layer and one or more hidden layers that each apply a non-linear transformation to a received input to generate an output.SUMMARYThis specification describes a system implemented as computer programs on one or more computers in one or more locations that predicts mortality for a subject based on the presence of a disease in a subject by processing biological information.As used in this specification, the disease can be a cardiovascular disease that includes coronary heart disease, coronary artery disease, or a stroke. Although cardiovascular disease is discussed for illustrative purposes, the mortality prediction techniques are not limited to cardiovascular-related causes.As used in this specification, biological information can include image data of the subject, epigenetic data representing a biological sample of the subject, proteomic data (e.g., protein expression level, protein presence / absence, protein functional state), genetic variation data (e.g., single nucleotide polymorphisms), metabolomic data, or a combination thereof.As used in this specification, "patient," "subject," and "individual" may be used interchangeably.According to one method, the system can obtain epigenetic data representing a biological sample for the subject. The system can then process the epigenetic data and generate an output comprising a mortality risk assessment for the subject based on the processing.According to another method, the system can obtain image data representing cardiovascular disease information for the subject. The system can then process the image data and generate an output comprising a mortality risk assessment for the subject based on the processing.According to another method, the system can obtain multi-modal biological information for the subject, the multi-modal biological information comprising at least one or more modalities, where a first modality is image data and a second modality is epigenetic data representing a biological sample. The system can then process the multi-modal biological information and generate an aggregated output comprising a mortality risk assessment for the subject based on the processing.In some embodiments, the system processes a single modality, such as imaging alone, epigenetic biomarkers alone, proteomic data alone, or genetic variation alone, to generate a mortality risk assessment.Particular embodiments of the subject matter described in this specification can be implemented so as to realize one or more of the following advantages.Identification of biomarkers can be particularly helpful to predict various health outcomes, including disease conditions, treatment effectiveness, and risk factors. Biomarkers from the same gene or in linkage disequilibrium with other markers can provide insights into disease pathways and guide therapy choices, such as whether a statin or aspirin is more suitable for a patient. For example, cardiovascular disease is the most common type of heart disease and was responsible for over 900,000 deaths in the United States in 2023. For example, understanding biological factors associated with cardiovascular disease (e.g., mortality risk, treatment, drug effectiveness, etc.) is essential for improving post-discharge survival in patients hospitalized for cardiovascular disease. Existing methods have attempted to estimate, detect and manage mortality risk as well as recommend interventions for subjects.However, existing methods often lack in sensitivity and specificity. In particular, current methods have implemented performing methylation procedures of biological samples to determine predictions of a condition, such as cardiovascular disease or mortality. Prediction models that only account for epigenetic signatures in biological samples, however, fail to account for confounding genetic variation, which results in models that lack robustness with respect to generalizability. On the other hand, current methods that implement only imaging-based methods for mortality predictions may poorly predict survival due to poor scalability in image data and the lack of information on disease pathways or management strategies.In contrast, the described techniques leverage multiple modalities of data for a subject by processing image data and epigenetic data for a biological sample to determine condition predictions for a subject, specifically, a mortality risk. This approach focuses on using specific biomarkers, such as CpG sites, genes, or markers in PrecisionCHD™. This predictive capability can operate standalone or in conjunction with other biomarkers like proteins, demographics, and imaging. In particular, the system can process the image data to generate one or more image scores that can represent a first mortality risk assessment for the subject, the epigenetic data to generate one or more epigenetic scores that can represent a second mortality risk assessment for the subject, or both (e.g., as multi-modal biological information). Additionally, the described techniques provide information about disease pathways, aiding in the identification of optimal interventions, management strategies, and the quantification of intervention effectiveness. In particular, the advantages of this approach include more personalized, precise treatment strategies, better prediction of patient risk and therapy outcomes, and the potential for developing novel treatments based on biomarker insights.For example, the techniques can include processing the image data according to a Duke scoring algorithm to generate one or more image scores. As another example, the techniques can include performing a methylation assay on an isolated nucleic sample of the subject to generate the one or more epigenetic scores. In some examples, the system can feature scaling, dimensionality reduction, and / or survival modeling on the multi-modality biological information. Thus, by leveraging both types of modalities, the system can perform enhanced predictive accuracy across all time frames.In some examples, the biological information further includes genetic variation data, such as single nucleotide polymorphisms (SNPs). SNPs can represent inherited sequence differences that may influence disease mechanisms, drug metabolism, or treatment response. Genetic features may be analyzed alone or in combination with any other modality as described herein.Additionally, in some instances, the multi-scale nature of the disease can be accounted which may not be possible without multi-modal approach, further enhancing prediction performance. That is, because cardiovascular disease progresses across multiple biological scales (e.g., from early molecular changes to cellular, tissue- and organ-level dysfunction), the disclosed multimodal approach provides improved predictive performance by capturing this multi-scale nature of disease. Moreover, because methylation can both identify potentially targetable pathways and be used to monitor treatment response, these data support further investigations of approaches for improving patient outcomes based on the prediction. Importantly, the described techniques allow for prediction of an indication, a condition a factor, a choice of therapy, a driver, a mechanism, a disease pathway, or condition standalone or in conjunction with one or more other biomarkers. As such, these methods open up possibilities for developing new interventions, including drug targets, repurposing existing drugs for new indications, or recommending lifestyle changes.Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the methods and compositions of matter belong. Although methods and materials similar or equivalent to those described herein can be used in the practice or testing of the methods and compositions of matter, suitable methods and materials are described below. In addition, the materials, methods, and examples are illustrative only and not intended to be limited to predicting mortality. All publications, patent applications, patents, and other references mentioned herein are incorporated by reference in their entirety.BRIEF DESCRIPTION OF THE DRAWINGSFIG. 1 is a flow diagram of an example process for generating a mortality risk assessment for a subject by processing epigenetic data.FIG. 2 is a flow diagram of an example process for generating a mortality risk assessment for a subject by processing image data.FIG. 3 is a flow diagram of an example process for generating a mortality risk assessment for a subject by processing multi-modal biological information.FIG. 4 is a diagram of the results of generating a mortality risk assessment for a subject by processing multi-modal biological information.FIG. 5 is a flow diagram of an example process for generating an output for a subject by processing epigenetic data that includes a value indicative of one or more CpG sites.FIG. 6 is a flow diagram of an example process for generating an output for a subject by contacting cells with a test compound.FIG. 7 is a flow diagram of an example process for generating an output for a subject by processing epigenetic data from a subject at a first epigenetic data interval and at a second epigenetic data interval.FIG. 8 is a graph depicting time-dependent area under the curve (AUC) of the survival random forest model.Like reference numbers and designations in the various drawings indicate like elements.DETAILED DESCRIPTIONBiomarkers provide critical insight into multiple dimensions of disease, including health outcomes, underlying disease conditions, treatment effectiveness, modifiable and non-modifiable risk factors, and the biological pathways driving disease progression. For example, biomarkers can predict outcomes such as mortality, rehospitalization, and secondary events, and can help characterize coronary artery disease severity, plaque instability, microvascular dysfunction, inflammatory load, or endothelial injury. Because biomarkers from the same gene—or from loci in linkage disequilibrium—often reflect shared biological mechanisms, they can reveal the biological context that contributes to disease severity.Biomarkers also provide valuable information about treatment response. For instance, changes in methylation at genes such as MPO or CXCL1 may indicate whether neutrophil-targeted therapies, statins, aspirin, colchicine, or other agents are effectively modifying the patient’s biology. Accordingly, biomarker profiles can identify responders, non-responders, and subjects at elevated risk for adverse outcomes, enabling more precise therapy selection. Biomarkers further capture risk factors, including genetic susceptibility, inflammation, comorbidities (e.g., diabetes, hypertension), and lifestyle-related risks, supporting individualized risk-reduction strategies.Biomarkers also provide information for biological disease pathways relevant to disease progression and management. These pathways include inflammatory signaling, oxidative stress, metabolic dysregulation, endothelial dysfunction, thrombotic pathways, and immune-cell recruitment. Identifying which pathways are active in a given subject allows clinicians to select therapies based on the subject’s actual biological drivers, rather than relying solely on population-level assumptions.In contrast to existing approaches that typically analyze biomarkers in isolation, the described techniques leverage multiple modalities of biological data, including image data and epigenetic data, to generate more accurate, multi-scale predictions for a subject. While traditional prediction systems rely exclusively on epigenetic signatures (thereby failing to account for imaging-based disease burden) or rely solely on imaging metrics (thereby missing pathway-level insights from methylation), the disclosed system integrates both modalities to capture mortality risk. For example, image-derived metrics (e.g., Duke scores) can reflect short-term or near-term risk, while epigenetic scores derived from methylation assays can reflect long-term biological risk, including inflammatory and gene-regulatory pathways that evolve over months or years.By combining these modalities, the system generates richer, multi-temporal predictions, supports multi-scale modeling of disease, and enables improved identification of disease pathways and treatment opportunities. For example, the system can apply dimensionality reduction, scaling, or survival modeling to multimodal biological inputs, thereby enhancing predictive accuracy across timeframes. Because methylation changes can also be monitored repeatedly to evaluate treatment response, the integrated framework enables ongoing adjustment of therapeutic strategies and supports development of new interventions, including novel drug targets, drug repurposing, and individualized lifestyle recommendations.Additionally, in some examples, the multi-scale nature of the disease can be accounted which may not be possible without a multi-modal approach, further enhancing prediction performance. As used herein, “multi-scale” refers to the fact that disease develops across multiple biological layers and time horizons. Molecular-level alterations, such as DNA methylation changes, inflammatory activation, or oxidative stress, can progress to tissue-level remodeling, ultimately manifesting at the cellular, tissue and / or organ level. In some examples, multi-scale analysis can additionally incorporate genetic biomarkers, such as single nucleotide polymorphisms (SNPs), which do not change over time for a given individual but can provide contextual information that influences how other modalities contribute to risk estimation or pathway interpretation.Simply by way of example, cardiovascular disease-related events may also arise in the absence of significant stenosis, such as in ischemia with non-obstructive coronary arteries (INOCA), Myocardial Infarction with Non-Obstructive Coronary Arteries (MINOCA), Coronary Microvascular Dysfunction (CMD), Coronary Vasospasm (Variant Angina), Microvascular Heart Failure (HFpEF with microvascular involvement) demonstrating that disease biology extends beyond anatomical narrowing. Accordingly, integrating epigenetic biomarkers (capturing molecular-scale information) (see, e.g., Appendix A and B) with imaging biomarkers (reflecting cellular, tissue- and / or organ-scale pathology), and optionally proteomic (see, e.g., Appendix D), genetic (see, e.g., Appendix C) or other clinical variables, enables the system to model cardiovascular disease across these interconnected biological scales. That is, because methylation can both identify potentially targetable pathways and be used to monitor treatment response, these data support further investigations of approaches for improving patient outcomes based on the prediction.FIG. 1 is a flow diagram of an example process for generating a mortality risk assessment for a subject by processing epigenetic data. For convenience, the process 100 will be described as being performed by a system of one or more computers, located in one or more locations, and programmed appropriately in accordance with this specification.The system can obtain epigenetic data representing a biological sample for the subject (102). As used herein, epigenetic data typically refers to the methylation status of one or more genetic loci in the subject. In some examples, the epigenetic data can include numerical values at multiple time scales or longitudinal in nature.The system can process the epigenetic data (104) and generate an output comprising a mortality risk assessment for the subject (106). In particular, the system can process the epigenetic data using a particular procedure to generate one or more epigenetic scores. As described in this specification, the one or more scores can correspond to one or more elements of a vector, one or more vectors, or a combination thereof. The one or more epigenetic scores can represent a particular term (e.g., months, years) of a second mortality risk assessment for the subject. For example, the system can perform a methylation assay to generate the one or more epigenetic scores. In some examples, the one or more epigenetic scores can represent a mortality risk assessment at varying time scales for the subject.In some examples, the particular procedure can include using an algorithm to generate the one or more epigenetic scores. For example, the system can use random forest algorithm, Bayesian algorithms, machine learning models, transfer learning, or a combination thereof.In some examples, the system can output the one or more suggested interventions for the subject according to the output. Interventions are known in the art and can include, for example, pharmaceutical intervention (e.g., Statins, beta blockers, colchicine, semaglutide, etc.), epigenetic editing (e.g. CRISPR, Transcription Activator-Like Effectors etc.), lifestyle intervention (e.g., exercise, nutrition / diet plan, smoking, alcohol, stress management) intervention (e.g., stenting, coronary artery bypass graft), managing risk factors (e.g., diabetes, hypertension, cholesterol, lipoprotein) and combinations thereof.In some examples, the system can perform one or more actions according to the one or more suggested interventions for the subject. In particular, the system can obtain image data representing disease information for the subject, and the system can process the image data using a particular algorithm to generate one or more image scores. The one or more image scores can represent a first mortality risk assessment for the subject (e.g., which may correspond to short-term, intermediate-term, or long-term mortality risk in some examples, but not necessarily so). That is, the system can combine the epigenetics data with one or more image scores. In some examples, the system can combine the image data with one or more scores from the epigenetic data.In some examples, the system can combine the one or more image scores and the one or more epigenetic scores to generate an aggregated output representing a combined second mortality risk assessment and a first mortality risk assessment for the subject.In some examples, the system can determine disease pathway information based on the image data, the one or more image scores, information associated with the methylation assay, the one or more epigenetic scores, the aggregated output, or a combination thereof. In some other examples, the system can monitor the subject at a particular frequency based on the image data, the one or more image scores, information associated with the methylation assay, the one or more epigenetic scores, the aggregated output, or a combination thereof. In some examples, the system can update the frequency based on the particular modality. For example, the system can use methylation as a basis to measure more imaging data or to measure both imaging data and epigenetic data.Importantly, in some examples, the system can perform additional measurements to assess an effectiveness of the interventions. In this example, the system can leverage the epigenetic information to measure changes in epigenetic data after a certain timeframe following the particular intervention, which allows for continuous optimization of interventions.FIG. 2 is a flow diagram of an example process for generating a mortality risk assessment for a subject by processing image data. For convenience, the process 100 will be described as being performed by a system of one or more computers, located in one or more locations, and programmed appropriately in accordance with this specification.The system can obtain image data representing disease information for the subject (202). Image data can include, without limitation, Echocardiography, Magnetic Resonance Imaging, Coronary computed tomography angiography, Nuclear Imaging, Invasive Coronary Angiography, Optical Coherence Tomography, Intravascular Ultrasound, X-ray Angiography, Fluoroscopy, Near-Infrared Spectroscopy and Photoacoustic Imaging. In some examples, the image data can additionally include modalities configured to assess microvascular structure or function, tissue-level perfusion, inflammatory activity, plaque composition, or metabolic imaging signals, such as PET, SPECT, contrast-enhanced MRI, perfusion MRI, or molecular imaging techniques. Such modalities may capture microvascular dysfunction, endothelial activation, tissue edema, or inflammatory cell infiltration.By way of example, modalities related to coronary heart disease can include, without limitation, ECG (electrocardiogram measuring electrical activity of the heart), SCG (seismocardiography measuring chest vibrations), PPG (photoplethysmography measuring blood volume changes), RHC (right heart catheterization measuring pressures and CO), echocardiography (ultrasound imaging of cardiac structure and function), cardiac MRI (magnetic resonance imaging of cardiac tissue and function), coronary Artery Calcium (CAC) (CT-based calcium scoring for atherosclerosis), CTA (Coronary CT Angiography) (CT imaging for coronary stenosis and plaque), carotid ultrasound / CIMT (ultrasound measurement of carotid artery thickness), SPECT / PET (nuclear imaging for perfusion and metabolism), CPET (cardiopulmonary exercise test measuring VO2 and ventilatory efficiency), ABI (ankle-brachial index for peripheral artery disease), Pulse Wave Velocity (assessment of arterial stiffness), ABPM (ambulatory blood pressure monitoring), BCG (ballistocardiography measuring body recoil), impedance cardiography (electrical impedance to estimate CO), coronary angiography (gold standard for coronary stenosis), left heart cath (measures LV pressures and aortic gradients), FFR / iFR (physiologic assessment of coronary lesion severity), EP Study (electrophysiology mapping for arrhythmias), endomyocardial biopsy (tissue sampling for myocarditis / amyloid), FFR-CT (noninvasive estimation of fractional flow reserve), OCT (intravascular optical imaging of plaque), IVUS (ultrasound inside vessels to assess plaque), and combinations thereof.The system can process the image data (204) and generate an output comprising a mortality risk assessment for the subject (206). In particular, the system can process the image data using a particular algorithm to generate one or more image score. The one or more image scores can represent a particular term (e.g., hours, days, weeks, months) of a second mortality risk assessment for the subject. For example, the system can process the image data using a Duke scoring algorithm to generate a Duke image score. That is, the system can process the image data using any particular method configured to estimate stenosis information (e.g., a Jeopardy score).In some examples, the system can output the one or more suggested interventions for the subject according to the output. Interventions are known in the art and can include, for example, pharmaceutical intervention (e.g., Statins, beta blockers, colchicine, semaglutide, etc.), epigenetic editing (e.g. CRISPR, Transcription Activator-Like Effectors etc.), lifestyle intervention (e.g., exercise, nutrition / diet plan, smoking, alcohol, stress management) intervention (e.g., stenting, coronary artery bypass graft), managing risk factors (e.g., diabetes, hypertension, cholesterol, lipoprotein) and combinations thereof.In some examples, the system can perform one or more actions according to the one or more suggested interventions for the subject. In particular, the system can perform collection of epigenetic data representing a biological sample for the subject. The epigenetic data can include numerical values at multiple scales. The system can process the collected epigenetic data based on a methylation assay to generate one or more epigenetic scores. For example, the one or more epigenetic scores can represent a second mortality risk assessment for the subject.In some examples, the system can then combine the one or more image scores and the one or more epigenetic scores to generate an aggregated output representing a combined second mortality risk assessment and a first mortality risk assessment (e.g., which may correspond to short-term, intermediate-term, and / or long-term mortality risk in some examples, but not necessarily so).In some examples, the system can determine disease pathway information based on the image data, the one or more image scores, information associated with the methylation assay, the one or more epigenetic scores, the aggregated output, or a combination thereof. In some other examples, the system can monitor the subject at a particular frequency based on the image data, the one or more image scores, information associated with the methylation assay, the one or more epigenetic scores, the aggregated output, or a combination thereof.In some examples, the system determines information for specific disease pathways by combining multimodal biological signals. The system can identify inflammatory pathway activity by integrating imaging features that indicate vascular inflammation with methylation signatures at inflammation-related loci, such as MPO-, CXCL1-, or AHRR-associated CpG sites. The system can also determine activity in oxidative stress pathways, endothelial dysfunction pathways, immune-cell recruitment pathways, metabolic dysregulation pathways, or thrombosis-related pathways by jointly analyzing imaging data and epigenetic data.FIG. 3 is a flow diagram of an example process for generating a mortality risk assessment for a subject by processing multi-modal biological information. For convenience, the process 300 will be described as being performed by a system of one or more computers, located in one or more locations, and programmed appropriately in accordance with this specification.The system can obtain multi-modal biological information for the subject (302). The multi-modal biological information includes at least one or more modalities at multiple time points. The first modality is image data, and the second modality is epigenetic data representing a biological sample. In particular, the system can obtain the second modality by isolating nucleic acid from the biological sample and performing a methylation assay of the isolated nucleic acid.The system can process the multi-modal biological information (304) and generate an aggregated output comprising a mortality risk assessment for the subject (306).In some examples, the system can process the multi-modal biological information by generating an image metric and an epigenetic metric. In particular, the system can generate the image metric by processing the image data using a first particular algorithm, and the system can generate the epigenetic metric by processing the epigenetic data using a second particular algorithm, such as a methylation assay that processes the nucleic sample. The system can then combine the image metric and the epigenetic metric to generate the aggregated output representing a combined second mortality risk assessment and a first mortality risk assessment for the subject.For example, the aggregated output can be a concordance index (e.g., a C index) that measures how well the system correctly ranked subjected based on their predicted risk of mortality.In some other examples, the system can process the image data, the epigenetic data, or both using a particular second algorithm configured to process data from the at least one or more modalities to generate the aggregated output. For example, the particular second algorithm can be a pre-trained machine learning model configured to combine the image data and the epigenetic data to generate the aggregated output. For example, the aggregated output can be a concordance index (e.g., a C index) that measures how well the system correctly ranked subjected based on their predicted risk of mortality.In some examples, the system can additionally generate one or more modality-specific metrics from additional biological data types, including but not limited to genetic biomarkers (e.g., single nucleotide polymorphisms or other germline variants), RNA expression or transcriptomic biomarkers, proteomic biomarkers (e.g., protein expression levels, protein presence / absence, or structural or functional protein states such as BDNF), serologic biomarkers (e.g., cytokines, chemokines, or cholesterol fractions), electrocardiographic biomarkers, or physiological parameters (e.g., blood pressure, heart rate variability, pulse wave velocity, or oxygen saturation).The system can combine any subset of metrics, such as imaging-derived metrics, epigenetic metrics, genetic metrics, transcriptomic metrics, proteomic metrics, serologic metrics, electrocardiographic metrics, or physiological metrics, to generate the aggregated output. Any modality described herein may be used individually or in any combination, and the system may process a single modality, multiple modalities of the same type, or heterogeneous modalities to generate predictions, pathway information, intervention recommendations, or monitoring intervals.In some examples, the aggregated output can incorporate proteomic features that provide intermediate predictive value across biological timescales. Proteomic features may include, but are not limited to, protein expression levels, presence or absence of particular proteins, or protein functional states. The proteomic features may bridge imaging-derived signals, which may capture more immediate physiological status, and epigenetic features, which may reflect longer-term biological processes. That is, the first mortality risk assessment can be a short-term mortality risk, the combined second mortality risk assessment can be a long-term mortality risk, and the proteomic features can represent an intermediate-term mortality risk. These examples illustrate potential multimodal temporal relationships but do not limit the use of any modality to a specific time horizon.Although certain modalities may empirically exhibit different predictive behaviors across biological timescales, these observations are illustrative rather than limiting. Any modality described herein may contribute predictive value at short-term, intermediate-term, or long-term intervals, and the system does not require or assume any fixed assignment between modality and timescale.In some examples, the system can output the one or more suggested interventions for the subject according to the output. Interventions are known in the art and can include, for example, pharmaceutical intervention (e.g., Statins, beta blockers, colchicine, semaglutide, etc.), epigenetic editing (e.g. CRISPR, Transcription Activator-Like Effectors etc.), lifestyle intervention (e.g., exercise, nutrition / diet plan, smoking, alcohol, stress management) intervention (e.g., stenting, coronary artery bypass graft), managing risk factors (e.g., diabetes, hypertension, cholesterol, lipoprotein) and combinations thereof.In some examples, the system selects and / or prioritizes interventions according to the multi-scale nature of disease. The system can recommend immediate, cellular, tissue and / or organ-level interventions, such as stenting or coronary artery bypass grafting, when the subject exhibits high risk based on imaging features or other rapidly changing physiological data. The system can also recommend molecular-level interventions or pathway-level interventions, such as pharmaceutical therapy or lifestyle modification, when the subject exhibits biological risk reflected in methylation signatures or other molecular biomarkers. By aligning each intervention with the biological scale and temporal window of risk, the system delivers targeted, mechanism-appropriate treatment strategies for the subject.In some examples, the system can determine disease pathway information based on the image data, the one or more image scores, information associated with the methylation assay, the one or more epigenetic scores, the aggregated output, or a combination thereof. In some other examples, the system can monitor the subject at a particular frequency based the image data, the one or more image scores, information associated with the methylation assay, the one or more epigenetic scores, the aggregated output, or a combination thereof.FIG. 4 is a diagram of the results of generating a mortality risk assessment for a subject by processing multi-modal biological information.The graph of FIG. 4 shows the performance of multiple methods in determining a mortality prediction of multiple subjects according to an accuracy metric. In particular, the graph shows a time-dependent area under the curve (AUC) analysis for mortality prediction using epigenetic data (e.g., “methylation”), image data (e.g., “duke”), and both (e.g., “methylation + duke”) for a particular number of days. That is, each of the methods are examined according to a concordance index (e.g., a C index) that measures how well the system correctly ranked subjected based on their predicted risk of mortality.As shown, combining both modalities for mortality prediction produced an increased C-index over merely using epigenetic data or image data, resulting in enhanced predictive accuracy across all time frames.FIGs. 5-8 are flow diagrams that describe example processes for processing epigenetic data associated with one or more biomarkers to generate an output. For example, the one or more biomarkers can be from a same gene and / or in linkage disequilibrium with the same kind of biomarker or a different biomarker. In this case, the output can be a particular prediction associated with a health of the subject. In particular, the prediction can be an indication, condition, factor, choice of intervention, a driver, mechanism, a disease pathway of a disease of some condition, or a combination thereof associated with the one or more biomarkers.As such, the system can generate one or more predictions, such as a prediction of whether a certain type of drug may be better suited for a particular patient. Additionally, the system can generate a prediction of a particular patient’s risk category while predicting which drug would be most effective for treating a particular condition. Importantly, by generating these predictions, along with performing measurements after implementing an interventions, the system allows for the development of new interventions. For example, by performing one or more measurements based on a particular intervention being administered to the patient, the system can process the one or more measurements to generate updated drug targets, updated lifestyle recommendations, updated drugs for a new indication in the patient, or a combination thereof.FIG. 5 is a flow diagram of an example process for generating an output for a subject by processing epigenetic data that includes a value indicative of one or more CpG sites. For convenience, the process 500 will be described as being performed by a system of one or more computers, located in one or more locations, and programmed appropriately in accordance with this specification.The system can obtain epigenetic data representing a biological sample for the subject (502). As used herein, epigenetic data typically refers to the methylation status of one or more genetic loci in the subject. In this case, the epigenetic data includes a value indicative of the presence or absence of methylation at one or more CpG sites. The one or more CpG sites are selected from cg03725309, cg12586707, cg04988978, cg17901584, cg21161138, and cg12655112. In some examples, the one or more CpG sites are within the sequence encoding MPO and / or within the sequence encoding CXCL1 and / or within the sequence encoding SARS1 and / or within the sequence encoding AHRR, and / or within the sequence encoding DHCR24, and / or within the sequence encoding DHCR24-DT, and / or within the sequence encoding EHD4.The system can process the epigenetic data (504) and generate an output (506). In particular, the system can process the epigenetic data using a particular procedure to generate one or more epigenetic scores. As described in this specification, the one or more scores can correspond to one or more elements of a vector, one or more vectors, or a combination thereof. For example, the system can perform a methylation assay to generate the one or more epigenetic scores. In some examples, the one or more scores can indicate a presence or absence of methylation at one or more CpG sites, which is predictive of mortality risk in the subject. The output can be a predicted indication, condition, factor, choice of intervention, a driver, mechanism, or disease pathway of a disease of some condition. For example, the one or more epigenetic scores can represent a particular term (e.g., months, years) of a second mortality risk assessment for the subject. In another example, the one or more epigenetic scores can represent a change in inflammation disease driven for this patient by inflammation, an effectiveness of intervention / therapy based on one or more biomarkers (e.g., protein, demographics, imaging, etc.).In some examples, the output can include one or more suggested interventions for the subject according to the output. Interventions are known in the art and can include, for example, pharmaceutical intervention (e.g., Statins, beta blockers, colchicine, semaglutide, etc.), epigenetic editing (e.g. CRISPR, Transcription Activator-Like Effectors etc.), lifestyle intervention (e.g., exercise, nutrition / diet plan, smoking, alcohol, stress management) intervention (e.g., stenting, coronary artery bypass graft), managing risk factors (e.g., diabetes, hypertension, cholesterol, lipoprotein) and combinations thereof.In some examples, the system can perform one or more actions according to the one or more suggested interventions for the subject. In particular, the system can obtain image data representing disease information for the subject, and the system can process the image data using a particular algorithm to generate one or more image scores. The one or more image scores can represent a first mortality risk assessment for the subject, a second mortality risk assessment, or both.In some examples, the system can combine the one or more image scores and the one or more epigenetic scores to generate an aggregated output representing a combined second mortality risk assessment and a first mortality risk assessment for the subject. As further described with respect to FIG. 3, the aggregated output may additionally incorporate metrics derived from other modalities, such as proteomic biomarkers, genetic variants, serologic markers, electrocardiographic data, or physiological signals. For example, the first mortality risk assessment can be a short-term mortality risk, and the combined second mortality risk assessment can be a long-term mortality risk. In this case, proteomic biomarkers, such as protein expression levels, protein presence or absence, or protein functional states, can provide biological data at, e.g., intermediate biological timescales that complement image-derived and epigenetic-derived features within the aggregated output.In some examples, the system can determine disease pathway information based on the image data, the one or more image scores, information associated with the methylation assay, the one or more epigenetic scores, the aggregated output, or a combination thereof. In some other examples, the system can monitor the subject at a particular frequency based on the image data, the one or more image scores, information associated with the methylation assay, the one or more epigenetic scores, the aggregated output, or a combination thereof.FIG. 6 is a flow diagram of an example process for generating an output for a subject by contacting cells with a test compound. For convenience, the process 600 will be described as being performed by a system of one or more computers, located in one or more locations, and programmed appropriately in accordance with this specification.The system can obtain epigenetic data representing a biological sample for the subject (502). As used herein, epigenetic data typically refers to the methylation status of one or more genetic loci in the subject. In this case, the epigenetic data includes a value indicative of the presence or absence of methylation at one or more CpG sites. The one or more CpG sites are selected from cg03725309, cg12586707, cg04988978, cg17901584, cg21161138, and cg12655112. In some examples, the one or more CpG sites are within the sequence encoding MPO and / or within the sequence encoding CXCL1 and / or within the sequence encoding SARS1 and / or within the sequence encoding AHRR, and / or within the sequence encoding DHCR24, and / or within the sequence encoding DHCR24-DT, and / or within the sequence encoding EHD4.The system can contact cells with a test compound (602). The system selects the test compound from nucleic acids (DNAs, RNAs), proteins (e.g., peptides, antibodies), small molecules, chemicals, or a combination thereof.The system can obtain epigenetic data from the cells (604). The epigenetic data includes a value indicative of the presence or absence of methylation at one or more CpG sites.The system can process the epigenetic data (606) and generate an output (608). In particular, the system can process the epigenetic data using a particular procedure to generate one or more epigenetic scores. As described in this specification, the one or more scores can correspond to one or more elements of a vector, one or more vectors, or a combination thereof. For example, the system can perform a methylation assay to generate the one or more epigenetic scores. In some examples, the one or more scores can indicate a presence or absence of methylation at one or more CpG sites, which is predictive of mortality risk in the subject, as described in further detail above with reference to FIG. 5The output can be a predicted indication, condition, factor, choice of intervention, a driver, mechanism, or disease pathway of a disease of some condition, as described in further detail above with reference to FIG. 5.FIG. 7 is a flow diagram of an example process for generating an output for a subject by processing epigenetic data from a subject at a first epigenetic data interval and at a second epigenetic data interval.The system can obtain epigenetic data (604). The epigenetic data can be obtained from the subject at a first epigenetic data interval and a second epigenetic data interval. The epigenetic data includes a value indicative of the presence or absence of methylation at one or more CpG sites. In some examples, the system can obtain epigenetic data at multiple time points to enable the detection of statistically significant changes in biomarker values over time. Such changes may reflect biological progression, regression, or response to an intervention. The system may incorporate uncertainty quantification techniques to determine whether observed temporal changes in methylation or other biomarkers exceed expected measurement variability or noise.In some examples, statistically significant temporal changes in biomarker values can be associated with specific interventions at the individual level (e.g., medication, lifestyle modification, dietary change) or at the population level (e.g., public health measures, clinical treatment guidelines). The system may further use these statistically validated changes to adjust risk estimates, recommend interventions, or determine whether a given intervention is producing a measurable biological effect.The system can process the epigenetic data (606) and generate an output (608). In particular, the system can process the epigenetic data using a particular procedure to generate one or more epigenetic scores. As described in this specification, the one or more scores can correspond to one or more elements of a vector, one or more vectors, or a combination thereof. For example, the system can perform a methylation assay to generate the one or more epigenetic scores. In some examples, the one or more scores can indicate a presence or absence of methylation at one or more CpG sites, which is predictive of mortality risk in the subject, as described in further detail above with reference to FIG. 5.The output can be a predicted indication, condition, factor, choice of intervention, a driver, mechanism, or disease pathway of a disease of some condition, as described in further detail above with reference to FIG. 5.In some examples, the system can also obtain imaging data at a first image data interval and a second image data interval, and the system can process both the epigenetic data and the imaging data to generate the output. In particular, the system can leverage the different modalities of data to generate individualized data for a subject at a first individualized data interval and a second individualized data interval.The process 500, the process 600, and the process 700 of generating an output for a subject based on one or more biomarkers can be particularly advantageous for reducing (e.g., controlling or reversing) inflammation, altering mortality risk, reversing methylation or demethylation of CpG sites within nucleic acid sequences encoding MPO and / or CXCL1.Although FIG. 7 illustrates repeated acquisition of epigenetic data at two time intervals, in some examples, the system can additionally or alternatively obtain multiple measurements of other modalities, including image data, proteomic data, metabolomic data, or clinical measurements at two or more time intervals. Multiple modalities can be collected longitudinally, independently, or in coordinated fashion to monitor treatment response, progression of disease, or effectiveness of an intervention. Thus, FIG. 7 is intended to represent a generalized multi-time, multi-modality framework rather than being limited to epigenetic measurements alone.FIG. 8 is a graph depicting time-dependent area under the curve (AUC) of the survival random forest model.The graph of FIG. 8 shows the performance of multiple methods in determining a mortality prediction of multiple subjects according to an accuracy metric. In particular, the graph shows a time-dependent area under the curve (AUC) analysis for mortality prediction using epigenetic data for a particular number of days. That is, each of the methods are examined according to a concordance index (e.g., a C index) that measures how well the system correctly ranked subjects based on their predicted risk of mortality.As shown, combining both modalities for mortality prediction produced an increased C-index over merely using epigenetic data or image data, resulting in enhanced predictive accuracy across all time frames. Furthermore, the analysis on the larger data set without image data yielded similar results 0.652 (n=201) vs. 0.646 (n=338) c-index, while using image data (e.g., DJS) was more accurate in predicting short-term mortality (<7 months) while MSdPCR performed better for long-term scales.This specification uses the term “configured” in connection with systems and computer program components. For a system of one or more computers to be configured to perform particular operations or actions means that the system has installed on it software, firmware, hardware, or a combination of them that in operation cause the system to perform the operations or actions. For one or more computer programs to be configured to perform particular operations or actions means that the one or more programs include instructions that, when executed by data processing apparatus, cause the apparatus to perform the operations or actions.Embodiments of the subject matter and the functional operations described in this specification can be implemented in digital electronic circuitry, in tangibly-embodied computer software or firmware, in computer hardware, including the structures disclosed in this specification and their structural equivalents, or in combinations of one or more of them. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible non-transitory storage medium for execution by, or to control the operation of, data processing apparatus. The computer storage medium can be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or a combination of one or more of them. Alternatively or additionally, the program instructions can be encoded on an artificially-generated propagated signal, e.g., a machine-generated electrical, optical, or electromagnetic signal, that is generated to encode information for transmission to suitable receiver apparatus for execution by a data processing apparatus.The term “data processing apparatus” refers to data processing hardware and encompasses all kinds of apparatus, devices, and machines for processing data, including by way of example a programmable processor, a computer, or multiple processors or computers. The apparatus can also be, or further include, special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application specific integrated circuit). The apparatus can optionally include, in addition to hardware, code that creates an execution environment for computer programs, e.g., code that constitutes processor firmware, a protocol stack, a database management system, an operating system, or a combination of one or more of them.A computer program, which may also be referred to or described as a program, software, a software application, an app, a module, a software module, a script, or code, can be written in any form of programming language, including compiled or interpreted languages, or declarative or procedural languages, and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. A program may, but need not, correspond to a file in a file system. A program can be stored in a portion of a file that holds other programs or data, e.g., one or more scripts stored in a markup language document, in a single file dedicated to the program in question, or in multiple coordinated files, e.g., files that store one or more modules, sub-programs, or portions of code. A computer program can be deployed to be executed on one computer or on multiple computers that are located at one site or distributed across multiple sites and interconnected by a data communication network.In this specification the term “engine” is used broadly to refer to a software-based system, subsystem, or process that is programmed to perform one or more specific functions. Generally, an engine will be implemented as one or more software modules or components, installed on one or more computers in one or more locations. In some cases, one or more computers will be dedicated to a particular engine; in other cases, multiple engines can be installed and running on the same computer or computers.The processes and logic flows described in this specification can be performed by one or more programmable computers executing one or more computer programs to perform functions by operating on input data and generating output. The processes and logic flows can also be performed by special purpose logic circuitry, e.g., an FPGA or an ASIC, or by a combination of special purpose logic circuitry and one or more programmed computers.Computers suitable for the execution of a computer program can be based on general or special purpose microprocessors or both, or any other kind of central processing unit. Generally, a central processing unit will receive instructions and data from a read-only memory or a random access memory or both. The essential elements of a computer are a central processing unit for performing or executing instructions and one or more memory devices for storing instructions and data. The central processing unit and the memory can be supplemented by, or incorporated in, special purpose logic circuitry. Generally, a computer will also include, or be operatively coupled to receive data from or transfer data to, or both, one or more mass storage devices for storing data, e.g., magnetic, magneto-optical disks, or optical disks. However, a computer need not have such devices. Moreover, a computer can be embedded in another device, e.g., a mobile telephone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a Global Positioning System (GPS) receiver, or a portable storage device, e.g., a universal serial bus (USB) flash drive, to name just a few.Computer readable media suitable for storing computer program instructions and data include all forms of nonvolatile memory, media and memory devices, including by way of example semiconductor memory devices, e.g., EPROM, EEPROM, and flash memory devices; magnetic disks, e.g., internal hard disks or removable disks; magnetooptical disks; and CDROM and DVD-ROM disks.To provide for interaction with a user, embodiments of the subject matter described in this specification can be implemented on a computer having a display device, e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor, for displaying information to the user and a keyboard and a pointing device, e.g., a mouse or a trackball, by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback, e.g., visual feedback, auditory feedback, or tactile feedback; and input from the user can be received in any form, including acoustic, speech, or tactile input. In addition, a computer can interact with a user by sending documents to and receiving documents from a device that is used by the user; for example, by sending web pages to a web browser on a user’s device in response to requests received from the web browser. Also, a computer can interact with a user by sending text messages or other forms of message to a personal device, e.g., a smartphone that is running a messaging application, and receiving responsive messages from the user in return.Data processing apparatus for implementing machine learning models can also include, for example, special-purpose hardware accelerator units for processing common and compute-intensive parts of machine learning training or production, i.e., inference, workloads.Machine learning models can be implemented and deployed using a machine learning framework, e.g., a TensorFlow framework.Embodiments of the subject matter described in this specification can be implemented in a computing system that includes a back-end component, e.g., as a data server, or that includes a middleware component, e.g., an application server, or that includes a front-end component, e.g., a client computer having a graphical user interface, a web browser, or an app through which a user can interact with an implementation of the subject matter described in this specification, or any combination of one or more such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication, e.g., a communication network. Examples of communication networks include a local area network (LAN) and a wide area network (WAN), e.g., the Internet.The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. In some embodiments, a server transmits data, e.g., an HTML page, to a user device, e.g., for purposes of displaying data to and receiving user input from a user interacting with the device, which acts as a client. Data generated at the user device, e.g., a result of the user interaction, can be received at the server from the device.While this specification contains many specific implementation details, these should not be construed as limitations on the scope of any invention or on the scope of what may be claimed, but rather as descriptions of features that may be specific to particular embodiments of particular inventions. Certain features that are described in this specification in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features that are described in the context of a single embodiment can also be implemented in multiple embodiments separately or in any suitable subcombination. Moreover, although features may be described above as acting in certain combinations and even initially be claimed as such, one or more features from a claimed combination can in some cases be excised from the combination, and the claimed combination may be directed to a subcombination or variation of a subcombination.Similarly, while operations are depicted in the drawings and recited in the claims in a particular order, this should not be understood as requiring that such operations be performed in the particular order shown or in sequential order, or that all illustrated operations be performed, to achieve desirable results. In certain circumstances, multitasking and parallel processing may be advantageous. Moreover, the separation of various system modules and components in the embodiments described above should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.EXAMPLESPART AExample 1—Materials and MethodsClinicalMethods and procedures for the Iowa CHD repository were approved by the University of Iowa Institutional Review Board (IRB #201910834) and have been described previously. After informed consent was obtained from the subject or their legally authorized representative, each subject was interviewed to confirm ACS presentation, then phlebotomized. Following discharge, research assistants—blinded to methylation status—abstracted medical records to confirm CHD and ACS diagnoses and collect relevant clinical data. When available, angiographic data were used to calculate DJS values (Dash et al., 1977) from angiograms obtained during or up to six months prior to admission. Mortality status was determined as described previously via hospital notifications or obituary searches (Lund et al., 2024).MethylationMSdPCR was performed as previously described (Philibert et al., 2023). Aliquots of bisulfite-converted whole blood DNA underwent pre-amplification of each of the six target regions (cg03725309, cg12586707, cg04988978, cg17901584, cg21161138, and cg12655112). The diluted pre-amp product was then mixed with droplet digital PCR reagents specific for each locus, partitioned into droplets, and PCR amplified. A Bio-Rad QX-200 reader and associated software determined the methylation status of each droplet.Random Forest SurvivalAdministrative censoring was applied at 3 years (1,095 days) to standardize follow-up, with participants surviving beyond 1,095 days right-censored at that time. To ensure robust model estimates, a time-based cross-validation approach was employed. Specifically, a time-quantile-based stratification partitioned the survival timeline into quantile-based bins, each containing at least one event, and 10-repeated 4-fold cross-validation was performed with consistent event-time distributions across folds. Random Survival Forest models (Ishwaran et al., 2008) were implemented via the scikit-survival package (Pölsterl, 2020). Cumulative dynamic area under the ROC curve was calculated at time points determined by event quantiles (Blanche et al., 2013).Cox Proportional HazardsTo assess the individual impact of MSdPCR indices, we built Cox proportional hazards models (Cox, 1972) for all 338 subjects (48 deaths). Model performance was assessed using Akaike’s Information Criterion (AICc), with features added sequentially based on their individual AICc (Akaike, 2011). Features failing to decrease AICc were excluded.Example 2—ResultsTwo overlapping populations of ACS subjects were used in these analyses. The smaller subset of subjects (n = 201, 23 deaths) included those for whom both angiographic and epigenetic data were available (Table 1). The second dataset included a total of 338 subjects (48 deaths) who had epigenetic data but did not necessarily have angiographic data. In the overall dataset, the ACS subjects were disproportionately male (67%) and tended to be in their mid-sixties. Nearly half presented with NSTEMIs, and males were more likely to present with STEMIs than females (p < 0.03).All possible combinations of MSdPCR, age, and DJS were used to calculate the c-index and cumulative dynamic AUC using RSF models (Table 2). From a univariate perspective, MSdPCR demonstrated the strongest predictive performance with a c-index of 0.652, followed by DJS at 0.642. Combining MSdPCR, age, and DJS yielded the highest c-index of 0.693, followed by MSdPCR + DJS (0.682). Applying the same analysis to the larger dataset (where DJS was not available, 48 deaths vs. 23 deaths) yielded similar results, indicating good generalizability as shown in Table 3 (0.652 vs. 0.646). Models using only DJS showed good short-term performance (~<130 days), but their predictive power dropped significantly thereafter, while MSdPCR retained predictive utility over longer follow-up (FIG. 8). This demonstrates that mortality modeling is both multi-modal and multi-scale, highlighting that imaging tools such as DJS may be best suited for earlier interventions, whereas epigenetic approaches may be better targets for longer-term risk reduction and management. Moreover, epigenetic biomarkers are sensitive to pharmacological interventions—changes can be observed within 90 days (Philibert et al., 2023)—making them ideal targets for non-invasive therapeutic adjustment.To better understand the contribution of specific MSdPCR markers to mortality, we conducted Cox proportional hazards modeling, adding features based on AICc values (Table 4). The best-performing model (Model 5) contained age, MPO, CXCL1, and an MPO × CXCL1 interaction term, demonstrating the non-linear and often contextual nature of methylation indices.Mechanistically, CXCL1 ligand binding and CXCR2 receptor dimerization in neutrophils and macrophages activates phospholipase C, generating inositol triphosphate (IP3). This process elevates intracellular Ca2+, stimulating assembly of the inflammasome subunits and promoting microtubule-mediated transport of mitochondria to the cell periphery. The ASC motif found on the mitochondrial membrane can then fully interact with other protein subunits of the NRPL3 inflammasome and cleave procaspase into active caspase. This, in turn, activates pro-inflammatory cytokines (IL-1B, IL-18) and Gasdermin D–mediated pyroptosis. Colchicine can inhibit neutrophil migration and block inflammasome activation by inhibiting the microtubule-mediated transport of mitochondria, while canakinumab binds directly to IL-1β, preventing downstream inflammatory signaling.Example 3—Conclusions• MSdPCR Predicts Long-Term Risk: The six MSdPCR assays from PrecisionCHD™ strongly predict all-cause mortality in patients admitted with ACS.• Complementary Value of DJS and MSdPCR: Combining MSdPCR indices with DJS information increases predictive performance, especially over longer time frames.• Multi-Scale Risk: These results underscore the need for approaching multi-modal tests as multi-scale problems. DJS offers strong short-term predictive ability, while epigenetic biomarkers add insight over extended follow-up.• Inflammatory Pathway Targets: Consistent with the known importance of lipoprotein oxidation and inflammation in CHD pathogenesis, Cox regression modeling showed that methylation at MPO and CXCL1 is a major predictor of all-cause mortality.• Interactive and Contextual Effects: The non-linear, interactive, and contextual nature of the MPO and CXCL1 signals suggests that monitoring and modulating these pathways could improve medication selection and survival in those admitted for ACS.Example 4—ReferencesAkaike (2011). Akaike’s information criterion. In M. Lovric (Ed.), International Encyclopedia of Statistical Science (pp. 25–25). Springer.Blanche et al. (2013). Estimating and comparing time-dependent areas under receiver operating characteristic curves for censored event times with competing risks. Statistics in Medicine, 32(30), 5381–5397.Cox (1972). Regression models and life-tables. Journal of the Royal Statistical Society: Series B (Methodological), 34(2), 187–220.Dash et al. (1977). The Duke Jeopardy Score: An angiographic assessment of coronary artery disease severity. Circulation, 55(3), 450–455.Graham et al. (2001). Validation of three coronary artery jeopardy scores in a population-based cardiac catheterization cohort. American Heart Journal, 142(2), 254–261.Ishwaran et al. (2008). Random survival forests. The Annals of Applied Statistics, 2(3), 841–860.Kumar & Cannon (2009). Acute coronary syndromes: Diagnosis and management, part 1. Mayo Clinic Proceedings, 84(10), 917–938.Liu et al. (2023). Prognostic value of coronary artery jeopardy scores in patients with acute coronary syndrome over multiple years of follow-up. Journal of Cardiology, 36(2), 115–123.Lund et al. (2024). [Title of the work]. [Journal Name], [Volume], [Page Range].Philibert et al. (2023). AI-guided epigenetic methylation-based testing for coronary heart disease in a high-risk population. Journal of the American Heart Association, 12(5), e030934.Pölsterl (2020). scikit-survival: A library for time-to-event analysis built on top of scikit-learn. Journal of Machine Learning Research, 21(212), 1–6.PART BExample 5—BackgroundOptimal medical therapy (OMT) is the cornerstone of treatment for those with coronary heart disease (CHD) (Ref 1). The centerpiece of OMT is the use of statins, the first of which was approved by the Food and Drug Administration in 1987 (Ref 2). In addition, depending on the clinical presentation, the use of beta blockers, antiplatelet agents, aldosterone antagonists and antihypertensives, most of which have been in clinical use for two or three decades, can be added. Sadly, despite the richness of these and other pharmaceutical options, CHD still has a high mortality even with optimal treatment.In efforts to further reduce CHD related mortality, a new generation of medications targeting pathways critical to the pathogenesis of CHD are being developed. In particular, medications targeting pathways related to inflammation and neutrophil activation, such as colchicine and the myeloperoxidase inhibitors such as AZM198, have been developed (Ref 3, 4). However, the clinical development and implementation of these agents has been hindered by the high side effect profile of these agents. For example, colchicine is effective in CHD related mortality (Ref 5), however, this reduction in cardiac mortality is nearly, if not completely, offset by an increase in mortality from non-cardiovascular deaths (Ref 6, 7). Similarly, while MPO inhibitors can be effective in some cases, their use is also associated with predisposition to infections and paradoxically, may lead to increased atherosclerosis (Ref 8). As a result of these low benefit-to-risk ratios, the use of these agents and other agents, such as canakinumab (Ref 9), in the treatment of CHD has been slowed if not completely arrested.The use of Precision Epigenetic approaches may illuminate the reasons for these therapeutic failures and help to identify new pathways for increasing the benefit-to-risk ratios. Specifically, we have developed PrecisionCHD™, an artificial intelligence guided integrated genetic-epigenetic test for CHD. At the heart of PrecisionCHD™ are six methylation-sensitive digital polymerase chain reaction (MSdPCR) assessment of DNA methylation at six cytosine-phospho-guanine (CpG) sites predictive of CHD status. Two of these MSdPCR assays, one that targets a CpG site in myeloperoxidase (MPO) and another that targets a CpG site in a key inflammatory gene named C-X-C motif chemokine ligand 1 (CXCL1), directly map to pathways targeted by the new generation of CHD pharmaceuticals. However, the relationship of these two pathways and the other four pathways to CHD is not additive. In fact, it is highly complex and interactive. As a result, the net effect of demethylation in a given pathway with respect to CHD is contextual on the methylation status in the other six pathways.Based on these and other observations, such as the well-known ability of methylation indices to predict survival, we hypothesized that methylation status in these pathways could be used to better understand and predict the likelihood of survival of patients. Furthermore, because significant changes from baseline can indicate the potential for interventions to change activity in those pathways, we hypothesized that the likelihood that targeting a given pathway with an intervention will result in a favorable therapeutic outcome will be contextual on the methylation status in the other MSdPCR pathways.To test this hypothesis, we examined methylation status of 338 subjects admitted to the University of Iowa Hospitals and Clinics (UIHC) for acute coronary syndrome (ACS). ACS is a set of conditions that includes unstable angina, ST-elevation myocardial infarction (STEMI) and non-ST-elevation myocardial infarction.Example 6—MethodsIn brief, subjects over the age of 18 years who were admitted to the UIHC for the evaluation and treatment of ACS were solicited for participation in the study. If interested, informed consent was obtained from the subject or their legally authorized representative, and each subject was interviewed, when possible, to confirm the CHD presentation history, and phlebotomized. After discharge, their medical records were abstracted by research assistants blinded to methylation status to confirm the clinical diagnosis of both CHD and ACS, and to gather associated clinical data. Type of ACS presentation, either STEMI, NSTEMI or other was determined by review of hospital discharge summary and progress notes. If not explicitly specified as either a STEMI or NSTEMI in the medical record, the encounter was categorized as “other.”Mortality status for each subject was determined as previously described (Ref 7, 9). As a first step, the UIHC electronic health record (EHR, Epic, Verona, WI) was abstracted to determine whether the system had been notified of the subject’s death and if so, the date of the death. As a second step, a research assistant blind to epigenetic status conducted a Google search for obituaries using the subject name and hometown as the key words. The resulting obituaries were downloaded, and the identification of the deceased individual confirmed by the use of other personal identifying information, such as age and birth date.MSdPCR was conducted as previously described. In brief, first DNA was extracted from whole blood phlebotomy samples using our standard procedures (Ref 10). One µg aliquots of each DNA sample was bisulfite converted using Qiagen Fast 96 Well EpiTect Bisulfite kits (Hilden, Germany) then eluted in 70 µl volumes following manufacturer’s directions. Next, 14 cycles of high stringency PCR amplification of the each of the six target regions (cg03725309, cg12586707, cg04988978, cg17901584, cg21161138 and cg12655112) were conducted on a 3 µl aliquot of each bisulfite-converted sample using a set of amplicon-specific primers. Finally, an aliquot of the enriched target solution was diluted 1:1500, mixed with primers and probes specific for the targeted loci as well as droplet digital PCR reagents, partitioned into droplets with a Bio-Rad droplet generator, and then PCR amplified. The methylation status of each droplet was determined using a Bio-Rad QX-200 Reader and the percent methylation status of each sample determined using the Bio-Rad QuantaSoft™ software.To understand the relationship of methylation to mortality after admission, survival analysis was conducted using Cox proportional hazards regression modeling (Ref 10). In brief, the values of the six MSdPCR indices were added to the base model consisting of age based on their Akaike’s Information Criterion score (AICc) (Ref 11). The performance of each model was examined, with features that did not decrease AICc values being excluded from the final model.Example 7—ResultsThe clinical and demographic features of the study cohort is provided in Table 5. Consistent with the demography of CHD, the average age of the subject was in the mid-sixties. The majority of the study participants were male, with all but 5 of the subjects reporting White ancestry.Nearly half of the subjects presented for hospitalization in the context of a NSTEMI (168 of 338). Another quarter of the subjects presented in the context of a STEMI (85 of 338), while the rest presented with other ACS-related presentations such as chest pain. In those with documented myocardial infarctions, men were significantly more likely than women to present with STEMIs (63 of 164, 38% vs 22 of 89, 25%, Chi Square p<0.03).Over the course of the follow up, 14% (48 of 338) of the subjects died, with 48 of these deaths occurring in the 3-year follow up window. There was no difference in the death rate of males as opposed to females (36 of 227 vs 12 of 111, Chi Square p<0.25).Table 6 illustrates the results of the Cox proportional hazards regression analyses. A model (Model 4) consisting of age, cg12586707 and cg04988978 exhibited the best performance (AIC 469.0) of the additive models, with the addition of other MSdPCR markers from Table 6 only increasing the AICc score. Interestingly, when the MSdPCR marker information from MPO and CXLC1 were added individually to age (Models 2 and 3), the directionality of the parameter estimate was negative for both markers, consistent with demethylation conferring increased risk. When CXCL1 information was added to the model containing age and MPO status, the directionality of the CXCL1 parameter estimate changed and the magnitude of the parameter estimate for both CXCL1 and MPO increased.Finally, because of our prior demonstration of the contextuality of methylation effects and the flip in the direction of the CXCL1 parameter estimates, we added an interaction effect between MPO and CXCL1 to the model (Model 5). Although the effect was not statistically significant (p<0.051, the addition of the interaction term increased the significance of each of the other markers in the model and significantly decreased the AICc score for the overall model to 467.3.Example 8—SummaryIn summary, we have shown that a model that includes age, cg12586707 data and cg04988978 data, and an interaction term between cg12586707 and cg04988978 gave the best performance. Significantly, when cg12586707 was added to the model containing both age and cg04988978, but not age alone, the directionality of the effect was reversed, with the interaction term also have a negative slope. In other words, the relationship of demethylation to survival is contextual on the methylation status at cg04988978.The MPO mediated generation of free radicals such as hypochlorous acid by neutrophils is a key mechanism in eliminating invasive bacteria. This activity of neutrophil is often greatly augmented by the CXCL1-mediated recruitment of neutrophils via the release of cytokines to those sites. Because excessive recruitment of neutrophils can actually promote inflammatory illness (Ref 12), organisms must carefully regulate the balance between the need for innate immunity and excess inflammation.Though blockade of MPO activity can be beneficial under certain circumstances, complete elimination of MPO activity is associated with increased mortality, elevated levels of blood cytokines and chemokines, and higher bacterial burden in models of endotoxemia (Ref 13). Because the release of those cytokines is, in part, regulated by the CXCL1 pathway in neutrophils, this suggests the need to consider the entire context of neutrophil immune function when predicting mortality outcomes.These results have direct bearing on the use of medications that target the MPO and CXCL1 pathways. In essence, these results suggest that by considering methylation status in both pathways simultaneously when selecting therapy, and by monitoring that status during therapy, clinicians will be to determine whether a medication is likely to be harmful or not. For example, for an individual considering an MPO inhibitor, increases in methylation at CXCL1 status during exposure to the MPO inhibitor may predict greater likelihood of treatment-related death. Similarly, a decrease in MPO methylation in those undergoing blockade of the CXCL1 pathway may indicate greater likelihood of fatality.Conceivably, these methods also could be applied to the selection of patients for initiation of treatment. Because methylation rarely increases above a biological set point, and effective blockade of a system is generally associated with a reversion of disease-associated methylation, it may be that a patient who is already demethylated at MPO may be a better candidate for CXCL1 agents than those with higher methylation at MPO.In summary, these results demonstrate the feasibility of using epigenetics in a method to predict the benefit of an MPO antagonist; a method to predict the benefit of an CXCL1 pathway antagonist (e.g., colchicine or canakinumab); and a method to predict the benefit of combination interventions such as MPO and CXCL1 pathway antagonists.Example 9—Bibliography1. Reed et al., 2022, Baylor University Medical Center Proceedings. Taylor & Francis.2. Harrington, 2017, JAMA Cardiology, 2(1):66-66.3. Chaikijurajai and Tang, 2020, Expert Opin. Ther. Targets, 24(7):695-705.4. Shamsuzzaman et al., 2024, Circulation, 150(9):687-705.5. Nidorf et al., 2020, New England J. Med., 383(19):1838-47.6. Fiolet et al., 2021, European Heart J., 42(28):2765-75.7. Galli et al., 2021, European Heart J. - Cardiovascular Pharmacotherapy, 7(4):e72-e73.8. Cheng et al., 2019, Arteriosclerosis, Thrombosis, and Vascular Biology, 39(7):1448-1457.9. Ridker et al., 2011, American Heart J., 162(4):597-605.10. Cox, 1972, J. Royal Statistical Society: Series B (Methodological), 34(2):187-202.11. Akaike, 2011, International Encyclopedia of Statistical Science. Springer. p. 25-25.12. Herrero-Cervera et al., 2022, Cellular & Molecular Immunology, 19(2):177-191.13. Reber et al., 2017, J. Exp. Med., 214(5):1249-58.Particular embodiments of the subject matter have been described. Other embodiments are within the scope of the following claims. For example, the actions recited in the claims can be performed in a different order and still achieve desirable results. As one example, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some cases, multitasking and parallel processing may be advantageous.

Claims

WHAT IS CLAIMED IS:

1. A method for predicting mortality in a subject, the method comprising:obtaining multi-modal biological information for the subject, the multi-modal biological information comprising at least one or more modalities, wherein a first modality is image data and a second modality is epigenetic data representing a biological sample, and wherein the epigenetic data is comprised of numerical values at multiple scales;processing the multi-modal biological information; andgenerating an aggregated output comprising a mortality risk assessment for the subject based on the processing.

2. The method of claim 1, further comprising:outputting one or more suggested interventions for the subject according to the aggregated output.

3. The method of claim 1, wherein the multi-modal biological information further comprises genetic variation data comprising one or more single nucleotide polymorphisms (SNPs).

4. The method of claim 1, wherein processing the multi-modal biological information comprises:generating one or more image scores by processing the image data using a first particular algorithm;generating one or more epigenetic scores by processing the epigenetic data using a second particular algorithm; andcombining the image data, the one or more image scores, information associated with a methylation assay procedure, the one or more epigenetic scores, or a combination thereof, to generate the aggregated output wherein each modality provides predictive information at one or more distinct or overlapping time frames, and wherein the aggregated output represents a mortality risk assessment for the subject across multiple temporal scales.

5. The method of claim 3, wherein at least one modality comprises a static biomarker that does not change over time, including a SNP.

6. The method of claim 3, wherein the one or more image scores represent a first mortality risk assessment for the subject, and wherein the one or more epigenetic scores represent a second mortality risk assessment for the subject.

7. The method of claim 1, further comprising a third modality, wherein the third modality comprises proteomic data including protein expression level, protein presence / absence, or protein functional state.

8. The method of claim 3, wherein obtaining the multi-modal biological information for the subject comprises:obtaining the second modality by isolating a nucleic sample from the biological sample.

9. The method of claim 8, further comprising:performing a methylation assay of the nucleic sample to generate the one or more epigenetic scores.

10. The method of claim 6, wherein processing the multi-modal biological information comprises:processing the image data, the epigenetic data, or both using a third particular algorithm configured to process data from the at least one or more modalities to generate the aggregated output.

11. The method of claim 10, wherein the third particular algorithm is a pre-trained machine learning model configured to combine the image data and the epigenetic data to generate the aggregated output.

12. The method of claim 1, wherein the multi-modal biological information is obtained at a plurality of time points.

13. The method of claim 9, further comprising:determining disease pathway information based on the image data, the one or more image scores, information associated with the methylation assay, the one or more epigenetic scores, the aggregated output, or a combination thereof.

14. The method of claim 13, further comprising:monitoring the subject at a particular frequency based on the image data, the one or more image scores, information associated with the methylation assay, the one or more epigenetic scores, the aggregated output, or a combination thereof.

15. A method for predicting mortality based on the presence of disease in a subject, the method comprising:obtaining epigenetic data representing a biological sample for the subject, wherein the epigenetic data is comprised of numerical values at multiple scales;processing the epigenetic data; andgenerating an output comprising a mortality risk assessment for the subject based on the processing.

16. The method of claim 15, further comprising:outputting one or more suggested interventions for the subject according to the output.

17. The method of claim 16, wherein processing the epigenetic data comprises:processing the epigenetic data using a first particular algorithm to generate one or more epigenetic scores.

18. The method of claim 17, wherein the first particular algorithm is a methylation assay.

19. The method of claim 18, further comprising:performing one or more actions according to the one or more suggested interventions for the subject.

20. The method of claim 19, wherein performing one or more actions according to the one or more suggested interventions for the subject comprises:obtaining image data representing disease information for the subject; andprocessing the image data using a second particular algorithm to generate one or more image scores representing a first mortality risk assessment for the subject.

21. The method of claim 20, wherein the one or more epigenetic scores represent a second mortality risk assessment for the subject.

22. The method of claim 21, further comprising:combining the one or more image scores and the one or more epigenetic scores to generate an aggregated output, wherein each modality provides predictive information at one or more distinct or overlapping time frames, and wherein the aggregated output represents a mortality risk assessment for the subject at the multiple scales.

23. The method of claim 22, wherein at least one modality comprises a static biomarker that does not change over time, including a SNP.

24. The method of claim 23, further comprising:determining disease pathway information based on the image data, the one or more image scores, information associated with the methylation assay, the one or more epigenetic scores, the aggregated output, or a combination thereof.

25. The method of claim 24, further comprising:monitoring the subject at a particular frequency based on the image data, the one or more image scores, information associated with the methylation assay, the one or more epigenetic scores, the aggregated output, or a combination thereof.

26. The method of claim 20, further comprising a third modality, wherein the third modality comprises proteomic data including protein expression level, protein presence / absence, or protein functional state.

27. A method for predicting mortality based on the presence of disease or risk factors in a subject, the method comprising:obtaining image data representing disease information for the subject;processing the image data; andgenerating an output comprising a mortality risk assessment for the subject based on the processing.

28. The method of claim 27, further comprising:outputting one or more suggested interventions for the subject according to the output.

29. The method of claim 28, wherein processing the image data comprises:processing the image data using a first particular algorithm to generate one or more image scores.

30. The method of claim 31, wherein the one or more image scores represent a first mortality risk assessment for the subject, a second mortality risk assessment for the subject, or both.

31. The method of claim 29, wherein the first particular algorithm is an algorithm configured to process a plurality of image features of the image data.

32. The method of claim 29, wherein the first particular algorithm is a Duke Jeopardy scoring algorithm, a plaque analysis algorithm, a Fractional Flow Reserve Computed Tomography (FFRCT) algorithm, or a combination thereof.

33. The method of claim 28, further comprising:performing one or more actions according to the one or more suggested interventions for the subject.

34. The method of claim 31, wherein performing one or more actions according to the one or more suggested interventions for the subject comprises:performing collection of epigenetic data representing a biological sample, wherein the epigenetic data is comprised of numerical values at multiple scales; andprocessing the collected epigenetic data according to a methylation assay to generate one or more epigenetic scores representing a second mortality risk assessment for the subject.

35. The method of claim 34, further comprising:combining the one or more image scores, the one or more epigenetic scores, information associated with the methylation assay, or a combination thereof to generate an aggregated output representing a combined second mortality risk assessment and a first mortality risk assessment for the subject.

36. The method of claim 35, further comprising:determining disease pathway information based on image data, the one or more image scores, information associated with the methylation assay, the one or more epigenetic scores, the aggregated output, or a combination thereof.

37. The method of claim 35, further comprising:monitoring the subject at a particular frequency based on the image data, the one or more image scores, information associated with the methylation assay, the one or more epigenetic scores, the aggregated output, or a combination thereof.

38. A method for predicting mortality risk in a subject presenting with a disease, the method comprising:obtaining epigenetic data from a biological sample from the subject, wherein the epigenetic data comprises a value indicative of the presence or absence or amount of methylation at one or more CpG sites; wherein the presence or absence or amount of methylation at one or more CpG sites is predictive of mortality risk in the subject.

39. The method of claim 38, wherein the one or more CpG sites are selected from cg03725309, cg12586707, cg04988978, cg17901584, cg21161138, and cg12655112 or one or more CpG sites in linkage disequilibrium with any one of cg03725309, cg12586707, cg04988978, cg17901584, cg21161138, and cg12655112.

40. The method of claim 39, wherein the one or more CpG sites are associated, or in linkage disequilibrium, with a nucleic acid sequence that encodes for MPO, CXCL1, SARS1, AHRR, DHCR24, DHCR24-DT, or EHD4.

41. The method of any one of claims 36-38, further comprising processing the epigenetic data.

42. The method of any one of claims 36-39, further comprising obtaining image data and processing the epigenetic data and the image data.

43. The method of any one of claims 36-40, further comprising obtaining individualized data, wherein individualized data comprises one or more of(i) genetic data;(ii) proteomic data;(iii) metabolites;(iv) age;(v) sex;(vi) race;(vii) BMI;(viii) vital signs (e.g., heart rate (beats / min), blood oxygen, body temperature, blood pressure (e.g., systolic BP (mm Hg), diastolic BP (mm Hg)), respiratory rate, ECG (heart rate, resting heart rate, HRV), sleep time and quality);(ix) medical history (e.g., smoking, atrial fibrillation / flutter, hypertension, coronary heart disease, myocardical infarction, heart failure, peripheral artery disease, COPD, diabetes (type 1 or type 2), CVA / TIA, chronic kidney disease, hemodialysis, angioplasty (peripheral or coronary), stent (peripheral or coronary), CABG, percutaneous coronary intervention);(x) medications (ACE-I / ARB, beta blocker, aldosterone antagonist, loop diuretics, nitrates, CCB, statin, aspririn, warfarin, clopidogrel);(xi) echocardiographic results (e.g., LVEF (%), RSVP (mm Hg));(xii) stress test results (e.g., ischemia on scan, ischemia on ECG);(xiii) angiography results (e.g., ≥ 70% coronary stenosis in ≥ 2 vessels, ≥ 70% coronary stenosis in ≥ 3 vessels); and / or(xiv) laboratory measures (e.g., sodium, blood urea nitrogen (mg / dL), creatinine (mg / dL), eGFR (median, CKDEPI), total cholesterol (mg / dL), LDL cholesterol (mg / dL), glycohemoglobin (%), glucose (mg / dL), and HGB (mg / dL)), andprocessing the epigenetic data, the image data, and the individualized data.

44. The method of any one of claims 36-41, wherein the epigenetic data comprises determining a single nucleotide polymorphism (SNP) corresponding to rs2869675, rs4376434, rs12129789, rs7585056, rs710987, rs4639796, rs1333048, rs12714414, rs942317, or rs1441433 or a SNP in linkage disequilibrium with rs2869675, rs4376434, rs12129789, rs7585056, rs710987, rs4639796, rs1333048, rs12714414, rs942317, or rs1441433.

45. The method of any one of claims 36-41, further comprising generating an output comprising a mortality risk assessment for the subject based on the processing.

46. A method of screening for compounds involved in• reducing (e.g., controlling or reversing) inflammation or other factors associated with CVD (e.g., neutrophil activation, smoking, alcohol use);• altering mortality risk due to CVD;• reversing methylation OR demethylation of CpG sites within nucleic acid sequences encoding MPO, CXCL1, SARS1, AHRR, DHCR24, DHCR24-DT, or EHD4 polypeptides or within nucleic acid sequences in linkage disequilibrium with the nucleic acid sequences encoding MPO, CXCL1, SARS1, AHRR, DHCR24, DHCR24-DT, or EHD4 polypeptides;• biochemical pathways involving MPO, CXCL1, SARS1, AHRR, DHCR24, DHCR24-DT, or EHD4 polypeptides;• biochemical pathways involved in neutrophil activation, smoking, and alcohol use; and / or• develop new interventions (e.g., new drug, new drug target, current drug for new indication, lifestyle recommendations);the method comprising:contacting cells with a test compound;obtaining epigenetic data from the cells, wherein the epigenetic data comprises a value indicative of the presence or absence or amount of methylation at one or more CpG sites;wherein a change in the presence, absence or amount of methylation at one or more CpG sites is predictive of a compound involved in one or more of the above bullet points.

47. The method of claim 44, wherein the compounds are selected from nucleic acids (DNAs, RNAs), proteins (e.g., peptides, antibodies), small molecules, chemicals, and pharmaceuticals.

48. A method for monitoring the efficacy of one or more interventions in a subject presenting with a disease, the method comprising:obtaining epigenetic data using a biological sample from the subject at a first epigenetic data interval and a second epigenetic data interval, wherein the epigenetic data comprises a value indicative of the presence or absence or amount of methylation at one or more CpG sites;wherein a change in the presence, absence or amount of methylation at one or more CpG sites between the first epigenetic data interval and the second epigenetic data interval is indictive of the efficacy of the one or more interventions in the subject.

49. The method of claim 46, further comprising obtaining image data at a first image data interval and a second image data interval.

50. The method of claim 46 or 47, further comprising obtaining individualized data at a first individualized data interval and a second individualized data interval.

51. The method of any one of claims 46-49, further comprising processing the first epigenetic data and the second epigenetic data and the first image data and the second image data in a multi-modal manner.

52. A method of predicting an indication, a condition, a mechanism or pathway of a disease or condition, or a choice of intervention in a subject, comprising:obtaining epigenetic data from a biological sample from the subject, wherein the epigenetic data comprises a value indicative of the presence or absence or amount of methylation at one or more CpG sites;wherein the presence or absence or amount of methylation at one or more CpG sites is predictive of an indication, a condition, a mechanism or pathway of a disease or condition, or a choice of therapy in the subject.

53. The method of claim 52, further comprising processing the epigenetic data to determine a score associated with mortality.

54. The method of claim 52 or 53, further comprising processing the epigenetic data to determine whether or not the subject has experienced a change in inflammation.

55. The method of any one of claims 52-54, further comprising processing the epigenetic data to determine whether inflammation is driving additional disease pathways in the subject.

56. The method of any one of claims 52-55, further comprising processing the epigenetic data to determine the intervention.

57. The method of any one of claims 52-56, further comprising processing the epigenetic data to determine the effectiveness of intervention.

58. The method of any one of claims 52-57, further comprising processing the epigenetic data in combination with imaging data and / or individualized data.

Images