System and method for determining biological age of an individual using metabolomic and proteomic data

Abstract

System, methods, and non-transitory computer readable medium (CRM) are provided that are directed to a biological age model utilizing metabolomic and proteomic data. Additional clinical variables and DNA methylation data may also be delivered to the methods, systems, and CRM. The biological age model determines a morbidity parameter based on a comparison between a patient's determined biological age and their chronological age to determine a time until death and possible treatment strategies.

Description

129319.01056 SYSTEM^AND^METHOD^FOR^DETERMINING^BIOLOGICAL^AGE^OF^AN^INDIVIDUAL^USING^ METABOLOMIC^AND^PROTEOMIC^DATA Cross Reference to Related Applications

[0001] The present application is based on, claims priority to, and incorporates herein by reference in its entirety for all purposes, US Provisional Application Serial No. 63 / 590,602, filed October 16, 2023. Statement of Government Support

[0002] This invention was made with government support under R01HL155742 awarded by the National Institutes of Health. The government has certain rights in the invention. Background

[0003] A goal of geroscience research is to define ‘biomarkers of aging’ that capture inter- individual differences in the timing of disease onset, functional decline, and death over the life course that are not captured via chronological age. Multiple molecular biomarker “clocks” have been developed to describe hallmarks of aging that span from telomere length, facial morphology, neuro-imaging data, immune cell counts, and large-scale omics including DNA methylation (DNAm), metabolomics, glycomics, and proteomics.

[0004] The strong molecular link between DNA methylation and the aging process has resulted in the widespread development and success of DNAm clocks with various biologic aging phenotypes aimed to reflect clinical biomarkers (e.g., PhenoAge), mortality (e.g., GrimAge), and the pace of aging (e.g., DunedinPACE). Despite their broad use across both research and commercial settings, DNAm clocks have notable limitations. The relative precision of these clocks can be a major issue due to the technical noise in measuring individual CpGs. The heterogeneity of immune cell subsets is also a confounder of DNAm aging estimates; while cellular deconvolution methods have been applied, to date immune deconvolution has been limited to 7 cell types. Elucidating causal CpGs in DNAm clocks remains and an important knowledge gap that currently limits their clinical potential.

[0005] Other omic platforms have the distinct advantage of providing direct biological links that may inform clinical strategies to reduce the aging process, prevent the development of chronic disease, and extend health aging. Blood plasma contains proteins from nearly all organs and cell 1   129319.01056 types, making it feasible to associate findings with specific tissues and organs. The proteome is also altered by hallmarks of aging including loss of proteostasis, dysregulated nutrient sensing, and altered intercellular communication and cellular senescence. The metabolome not only provides critical information about metabolic processes, but it also provides a snapshot of environmental exposures that may be critically linked to the aging process. The metabolome also carries information from multiple tissues across the body, increasing the potential aging information of metabolomics relative to methylation and transcriptomic clocks of blood cells.

[0006] Despite the important advantages of other omics, the development of transcriptomic, proteomic, and metabolomic clocks with biologic aging phenotypes has been limited. Initial work has demonstrated that while individual omics clocks share commonalities; each omic data type provides distinct clues into the aging process, suggesting that the best and most clinically informative approach would combine information from multiple omic measurements to create an optimized aging clock.

[0007] Thus, there is a continuing need for new and better ways to project or to optimize and further inform performance of the biological processes, including processes of DNAm clocks. Summary

[0008] The present disclosure provides systems and methods that overcome the aforementioned challenges by providing systems and methods for utilizing a versatile and clinically relevant biological aging phenotype. The methods, system, and non-transitory computer readable medium (CRM) disclosed herein are based on the development and validation of multiple distinct and clinically relevant measures of biological age. These may include EMRAge, a clinically-based death predictor; DNAm EMRAge, a DNA methylation clock trained to predict EMRAge; and OMICm Age, a composite methylation clock which uses Epigenetic Biomarker Proxy (EBP) from proteomic, metabolomic, and clinical data trained to predict EMRAge, as just a few examples. The systems methods, and non-transitory CRM provide an omic-informed DNAm clock.

[0009] The present disclosure provides methods, systems and non-transitory CRM that overcome the aforementioned drawbacks by utilizing a versatile and clinically relevant biological aging phenotype that can be broadly recapitulated across EMRs, that is further optimized with a DNAm clock to this phenotype and biologically-relevant metabolomic and proteomic data to inform the underlying biology, while increasing precision and robustly adjusting for immune cell subsets. By 2   129319.01056 outperforming current methylation-based clocks in associations with chronic disease outcomes and mortality, the value of EMRAge, DNAm EMRAge, and OMICm Age is shown herein, while further substantiating that value of integrating multiple omic data in one biological aging phenotype.

[0010] According to one aspect of the present disclosure, a method of determining a biological age of a patient is described. The method comprises assembling metabolomic data and proteomic data related to a patient and delivering the metabolomic and proteomic data to a system for biological age analysis. The method further comprises receiving a biological age of the patient from the system and generating a report including the biological age of the patient.

[0011] In another aspect of the present disclosure, a system for determining a biological age of a patient is disclosed. The system includes a processor configured to receive metabolomic data and proteomic data from a patient, deliver the metabolomic data and proteomic data to a biological age model, receive a biological age of the patient from the biological age model, and generate a report indicating the biological age of the patient.

[0012] In another aspect of the present disclosure, a non-transitory computer readable medium is described. The CRM has instructions stored thereon, that when executed by a computer processor, cause the computer processor to carry out steps comprising delivering metabolomic data and proteomic data related to a patient to a system for biological age analysis. The instructions further comprise receiving a biological age of the patient from the system, and generating a report including the biological age of the patient.

[0013] These aspects are nonlimiting. Other aspects and features of the systems and methods described herein will be provided below. Brief Description of the Drawings

[0014] The foregoing features of embodiments will be more readily understood by reference to the following detailed description, taken with reference to the accompanying drawings, in which:

[0015] FIG. 1 is a block diagram of an example system for determining a biological age of a patient, according to aspects of the present disclosure.

[0016] FIG. 2 is a block diagram of example components that can implement the system for determining a biological age of a patients of FIG.1. 3   129319.01056

[0017] FIG. 3 is a flow chart setting forth a non-limiting example of steps of a process in accordance with the present disclosure.

[0018] FIG. 4 is a schematic of the overall study design, according to aspects of the present disclosure. MGB Biobank: Massachusetts General Brigham Biobank. EBP: Epigenetic Biomarker Proxy.

[0019] FIG.5 is a flowchart for inclusion of participants from the Massachusetts General Brigham (MGB) Biobank for the development of EMRAge.

[0020] FIG.6 is a Pearson’s correlation between EMRAge and chronological age in the test and train sets.

[0021] FIG.7A is a forest plot of hazard ratio and odds ratio between EMRAge and aging-related health outcomes.

[0022] FIG.7B is a pairwise correlation between different 4 estimates of EMRAge at timepoints Jan 1st of 2008, 2010, 2012, and 2014.

[0023] FIG.7C is a Kaplan Meier Plot of EMRAge vs. PhenoAge.

[0024] FIG. 7D is a hazard ratio of one standard deviation change to onset of aging-related diseases. These values were estimated in the testing dataset from the MGB Cohort (N = 5,148)

[0025] FIG.8A is correlation plot of EMRAge for DNAm EMRAge (left column) and OMICm Age (right column) in the training set (n = 2762).

[0026] FIG.8B is a correlation plot of EMRAge for DNAm EMRAge (left column) and OMICm Age (right column) in the testing set (n = 689).

[0027] FIG. 8C is a correlation plot of intra-class correlation coefficients (ICC) for DNAm EMRAge (left column) and OMICm Age (right column).

[0028] FIG.9A is a distribution of super pathways for all the metabolites.

[0029] FIG.9B is a distribution of subpathways for all the metabolites.

[0030] FIG. 10 is a comparison between the CpG sites included in every epigenetic clock. The horizontal bars represent the total number of CpG sites included in each epigenetic clock. The vertical bars represent the number of unique or shared CpG sites between clocks. PhenoAge clock refers to the DNA methylation version.

[0031] FIG. 11 is a schematic of Epigenetic Biomarker Proxy (EBP) scores included in the OMICm Age clock segmented by a system approach. This figure shows the EBPs that were selected after penalization of the OMICm Age clock. Each color represents one of the 7 systems. 4   129319.01056

[0032] FIG. 12A is a heatmap of Epigenetic Biomarker Proxy (EBPs) scores associated with diseases and lifestyle factors of the MGB-Aging Biobank Cohort. The association between each EBP and 6 diseases (Depression, Stroke, COPD, Cancer, CVD, and Type-2 Diabetes) and multiple lifestyle factors, including alcohol consumption, education level, and exercise per week, among others were evaluated. The strength of the color is proportional to the estimate of the linear regression adjusted by age, sex, ethnicity, BMI, and tobacco smoking. In red, positive associations. In blue, negative associations.

[0033] FIG. 12B is a heatmap of Epigenetic Biomarker Proxy (EBPs) scores associated with diseases and lifestyle factors of the TruDiagnostic Biobank. The association between each EBP and 6 diseases (Depression, Stroke, COPD, Cancer, CVD, and Type-2 Diabetes) and multiple lifestyle factors, including alcohol consumption, education level, and exercise per week, among others were evaluated. The strength of the color is proportional to the estimate of the linear regression adjusted by age, sex, ethnicity, BMI, and tobacco smoking. In red, positive associations. In blue, negative associations.

[0034] FIG. 13A is a comparison of OMICm Age and DNAm EMRAge to previous clocks. Horizontal bar plot of odds / hazard ratios of each methylation clock to aging-related diseases in the testing set of MGB-ABC cohort.

[0035] FIG.13B is a comparison of OMICm Age and DNAm EMRAge to previous clocks. ROC curves for 5-year and 10-year survival prediction classifiers utilizing a single methylation clock or chronological age.

[0036] FIG. 14 shows the correlation plots between epigenetic clocks and immune cells in the TruDiagnostic Biobank. Each column represents a covariate among age, gender male, gender female, and immune cells. The size is related to the magnitude of the correlation and the color to the direction (positive or negative).

[0037] FIG. 15A is a bubble plot for the representation of the lifestyle factors associated with OMICm Age clock for the TruDiagnostic Biobank. Visual representation of the effect sizes found with the four epigenetic clocks. Circle diameter represents the calculated estimated value of the factor. Positive associations are represented in green and negative associations in red. All the associations are adjusted by chronological age, biological sex, ethnicity, body mass index (BMI), and tobacco use. 5   129319.01056

[0038] FIG. 15B is a bubble plot for the representation of the lifestyle factors associated with OMICm Age clock for the MGB-ABC Biobank. Visual representation of the effect sizes found with the four epigenetic clocks. Circle diameter represents the calculated estimated value of the factor. Positive associations are represented in green and negative associations in red. All the associations are adjusted by chronological age, biological sex, ethnicity, body mass index (BMI), and tobacco use. Detailed Description

[0039] Referring now to FIG.1, an example of a system 100 for determining a biological age of a patient from clinical data in accordance with some embodiments of the systems and methods described in the present disclosure is shown. As shown in FIG. 1, a computing device 150 can receive one or more types of data (e.g., metabolomic data, proteomic data, genomic data, demographic data, risk indices, substance use information, vital sign measurements, metabolic measurements, hepatic function measurements, lipid measurements, complete blood count and blood count differential) from data source 102. In some embodiments, computing device 150 can execute at least a portion of a morbidity prediction system 104 to generate a report comparing a patient’s biological age determined from the clinical data source 102 with the patient’s chronological age.

[0040] Additionally or alternatively, in some embodiments, the computing device 150 can communicate information about data received from the data source 102 to a server 152 over a communication network 154, which can execute at least a portion of the morbidity prediction system 104. In such embodiments, the server 152 can return information to the computing device 150 (and / or any other suitable computing device) indicative of an output of the morbidity prediction system 104.

[0041] In some embodiments, computing device 150 and / or server 152 can be any suitable computing device or combination of devices, such as a desktop computer, a laptop computer, a smartphone, a tablet computer, a wearable computer, a server computer, a virtual machine being executed by a physical computing device, and so on.

[0042] In some embodiments, data source 102 can be any suitable source of clinical data, such as another computing device (e.g., a server storing clinical data). In some embodiments, data source 102 can be local to computing device 150. For example, data source 102 can be 6   129319.01056 incorporated with computing device 150. As another example, data source 102 can be connected to computing device 150 by a cable, a direct wireless link, and so on, Additionally or alternatively, in some embodiments, data source 102 can be located locally and / or remotely from computing device 150, and can communicate data to computing device 150 (and / or server 152) via a communication network (e.g., communication network 154).

[0043] In sonic embodiments, communication network 154 can be any suitable communication network or combination of communication networks. For example, communication network 154 can include a Wi-Fi network (which can include one or more wireless routers, one or more switches, etc.), a peer-to-peer network (e.g., a Bluetooth network), a cellular network (e.g., a 3G network, a 4G network, etc., complying with any suitable standard, such as CDMA, GSM, LTE, LTE Advanced, WiMAX, etc.), a wired network, and so on. In some embodiments, communication network 154 can be a local area network, a wide area network, a public network (e.g., the Internet), a private or semi-private network (e.g., a corporate or university intranet), any other suitable type of network, or any suitable combination of networks. Communications links shown in FIG. 1 can each be any suitable communications link or combination of communications links, such as wired links, fiber optic links, Wi-Fi links, Bluetooth links, cellular links, and so on.

[0044] Referring now to FIG.2, an example of hardware 200 that can be used to implement data source 102, computing device 150, and server 152 in accordance with some embodiments of the systems and methods described in the present disclosure is shown. As shown in FIG.2, in some embodiments, computing device 150 can include a processor 202, a display 204, one or more inputs 206, one or more communication systems 208, and / or memory 210. In some embodiments, processor 202 can be any suitable hardware processor or combination of processors, such as a central processing unit (“CPU”), a graphics processing unit (“GPU”), and so on. In some embodiments, display 204 can include any suitable display devices, such as a computer monitor, a touchscreen, a television, and so on. In some embodiments, inputs 206 can include any suitable input devices and / or sensors that can be used to receive user input, such as a keyboard, a mouse, a touchscreen, a microphone, and so on.

[0045] In some embodiments, communications systems 208 can include any suitable hardware, firmware, and / or software for communicating information over communication network 154 and / or any other suitable communication networks. For example, communications 7   129319.01056 systems 208 can include one or more transceivers, one or more communication chips and / or chip sets, and so on. In a more particular example, communications systems 208 can include hardware, firmware and / or software that can be used to establish a Wi-Fi connection, a Bluetooth connection, a cellular connection, an Ethernet connection, and so on.

[0046] In some embodiments, memory 210 can include any suitable storage device or devices that can be used to store instructions, values, data, or the like, that can be used, for example, by processor 202 to present content using display 204, to communicate with server 152 via communications system(s) 208, and so on. Memory 210 can include any suitable volatile memory, non-volatile memory, storage, or any suitable combination thereof. For example, memory 210 can include RAM, ROM, EEPROM, one or more flash drives, one or more hard disks, one or more solid state drives, one or more optical drives, and so on. In some embodiments, memory 210 can have encoded thereon, or otherwise stored therein, a computer program for controlling operation of computing device 150. In such embodiments, processor 202 can execute at least a portion of the computer program to present content (e.g., images, user interfaces, graphics, tables), receive content from server 152, transmit information to server 152, and so on.

[0047] In some embodiments, server 152 can include a processor 212, a display 214, one or more inputs 216, one or more communications systems 218, and / or memory 220. In some embodiments, processor 212 can be any suitable hardware processor or combination of processors, such as a CPU, a GPU, and so on. In some embodiments, display 214 can include any suitable display devices, such as a computer monitor, a touchscreen, a television, and so on. In some embodiments, inputs 216 can include any suitable input devices and / or sensors that can be used to receive user input, such as a keyboard, a mouse, a touchscreen, a microphone, and so on.

[0048] In some embodiments, communications systems 218 can include any suitable hardware, firmware, and / or software for communicating information over communication network 154 and / or any other suitable communication networks. For example, communications systems 218 can include one or more transceivers, one or more communication chips and / or chip sets, and so on. In a more particular example, communications systems 218 can include hardware, firmware and / or software that can be used to establish a connection, a Bluetooth connection, a cellular connection, an Ethernet connection, and so on.

[0049] In some embodiments, memory 220 can include any suitable storage device or devices that can be used to store instructions, values, data, or the like, that can be used, for example, 8   129319.01056 by processor 212 to present content using display 214, to communicate with one or more computing devices 150, and so on. Memory 220 can include any suitable volatile memory, non- volatile memory, storage, or any suitable combination thereof. For example, memory 220 can include RAM, ROM, EEPROM, one or more flash drives, one or more hard disks, one or more solid state drives, one or more optical drives, and so on. In some embodiments, memory 220 can have encoded thereon a server program for controlling operation of server 152. In such embodiments, processor 212 can execute at least a portion of the server program to transmit information and / or content (e.g., data, images, a user interface) to one or more computing devices 150, receive information and / or content from one or more computing devices 150, receive instructions from one or more devices (e.g., a personal computer, a laptop computer, a tablet computer, a smartphone), and so on.

[0050] In some embodiments, data source 102 can include a processor 222, one or more data acquisition system(s) 224, one or more communications systems 226, and / or memory 228. In some embodiments, processor 222 can be any suitable hardware processor or combination of processors, such as a CPU, a GPU, and so on.

[0051] Note that, although not shown, data source 102 can include any suitable inputs and / or outputs. For example, data source 102 can include input devices and / or sensors that can be used to receive user input, such as a keyboard, a mouse, a touchscreen, a microphone, a trackpad, a trackball, and so on. As another example, data source 102 can include any suitable display devices, such as a computer monitor, a touchscreen, a television, etc., one or more speakers, and so on.

[0052] In some embodiments, communications systems 226 can include any suitable hardware, firmware, and / or software for communicating information to computing device 150 (and, in some embodiments, over communication network 154 and / or any other suitable communication networks). For example, communications systems 226 can include one or more transceivers, one or more communication chips and / or chip sets, and so on. In a more particular example, communications systems 226 can include hardware, firmware and / or software that can be used to establish a wired connection using any suitable port and / or communication standard (e.g., VGA, DVI video, USB, RS-232, etc.), Wi-Fi connection, a Bluetooth connection, a cellular connection, an Ethernet connection, and so on.

[0053] In some embodiments, memory 228 can include any suitable storage device or devices that can be used to store instructions, values, data, or the like, that can be used, for example, 9   129319.01056 by processor 222 to control the one or more data acquisition system(s) 224, and / or receive data from the one or more data acquisition system(s) 224; to images from data; present content (e.g., images, a user interface) using a display; communicate with one or more computing devices 150; and so on. Memory 228 can include any suitable volatile memory, non-volatile memory, storage, or any suitable combination thereof. For example, memory 228 can include RAM, ROM, EEPROM, one or more flash drives, one or more hard disks, one or more solid state drives, one or more optical drives, and so on. In some embodiments, memory 228 can have encoded thereon, or otherwise stored therein, a program for controlling operation of data source 102. In such embodiments, processor 222 can execute at least a portion of the program to generate images, transmit information and / or content (e.g., data, images) to one or more computing devices 150, receive information and / or content from one or more computing devices 150, receive instructions from one or more devices (e.g., a personal computer, a laptop computer, a tablet computer, a smartphone, etc.), and so on.

[0054] Referring now to FIG.3, a non-limiting method of determining a biological age of a patient 300 is described. At step 302, metabolomic and proteomic data related to a patient is assembled. As used herein, “assembled” refers to, for example, that the data may be accessed from patient records, databases, or storage mediums, or may be acquired or derived. Additionally or alternatively, assembled may refer to the data being synthesized, processed, or pre-processed or formatted for use. In one non-limiting example, the data may be normalized or aligned, or may be otherwise pre-processed to be synthesized for processing, despite originating from different or multiple sources. The metabolomic and proteomic data may be derived from any of a variety of sources. As just one example, these may include a clinical sample of the patient and / or database or records systems. For example, the clinical sample may be blood, urine, fecal, or tissue biopsy sample, or records about the same. In a non-limiting example, untargeted global plasma metabolomics profiling may be performed using the Metabolon platform (Morrisville, NC). In a non-limiting example to generate unbiased proteomic data, proteins may be isolated from the clinical sample, enzymatically digested, separated using high pressure liquid chromatography (HPLC) and subsequently analyzed with tandem mass spectrometry. In another example, unbiased proteomic data may be generated using the Seer SP100 platform (Redwood City, CA) combined with liquid chromatography tandem mass spectrometry.   10   129319.01056

[0055] At step 304, the metabolomic data and proteomic data is delivered to a system for biological age analysis. The system may include the morbidity prediction system 104 as discussed above in FIGS. 1 and 2. In a non-limiting example, the morbidity prediction system 104 can include a biological age model. The biological model will be described in further detail in the example below.

[0056] Referring again to FIG.3, the process may then include receiving a biological age of the patient at step 306. In this way, as will be described, a system and / or model may be used to determine the biological age of the patient, which is then received at step 306. In a non-limiting example, the method further includes generating a report including a biological age of the patient at step 308. For example, the biological age is a morbidity parameter that is an estimate of time to death. In a non-limiting example, the morbidity parameter includes a comparison of the biological age of the patient (based on the metabolomic and proteomic data) with the chronological age of the patient. Here, the chronological age refers to the amount of time that has passed since a patient’s birth. In a further non-limiting example, the report may further include a treatment plan based on the comparison. For example, if the biological age of the patient is lower than their chronological age, no treatments are suggested. On the other hand, if the biological age of the patient is the same as their chronological age, the report may suggest no additional treatment or may suggest medication interventions or lifestyle changes to lower the biological age relative to the chronological age. In a further example, if the biological age of the patient is greater than their chronological age, the report may recommend at least one medical intervention and / or lifestyle change. In a non-limiting example, a lifestyle change may include increasing physical activity, dietary changes, and reducing smoking or use of other harmful substances. The example below provides additional detail of factors that have beneficial or adverse effects on the biological age of an individual.

[0057] In a non-limiting example, the method 300 may further include delivering at least one additional clinical parameter or genomic DNA methylation data to the system, at step 310. For example, the at least one additional clinical parameter may include demographic information, risk indices, substance use information, vital sign measurements, metabolic measurements, hepatic function measurements, lipid measurements, complete blood count, and / or differential blood count. In a further non-limiting example, the at least one additional clinical parameter may include at least one of gender, race, age, Charlson Comorbidity Index, smoking status, alcohol drinking 11   129319.01056 status, height, body mass index (BMI), systolic blood pressure, diastolic blood pressure, blood urea nitrogen, creatinine, glucose, albumin, total bilirubin, alkaline phosphatase, alanine transaminase, aspartate transaminase, high-density lipoprotein, low-density lipoprotein, triglyceride, hematocrit, platelet count, red cell width distribution, white blood cell count, basophils count, eosinophils count, and neutrophils count. This information may be used to determine and / or deliver the biological age at step 306 and / or may be included (individually or in synthesized or processed form) in the report at step 308.

[0058] In another aspect, a non-limiting example system for determining a biological age of a patient is disclosure and is described. The system may be the morbidity prediction system 104 as previously described in FIG. 1. In a non-limiting example, the system comprises a processor configured to receive metabolomic data and proteomic data related to a patient as described previously for the method of FIG. 3. Further, the processor delivers the metabolomic data and proteomic data to a biological age model, receives a biological age of the patient from the biological model, and generates a report indicating the biological age of the patient, as described for FIG.3 above.

[0059] As will be described in further detail in the example section, the biological age model is a weighted linear combination of a plurality of clinical variables (EMRAge). In another aspect, the biological age model further includes DNA methylation values associated with the plurality clinical variables, wherein the DNA methylation values are CpG sites predictive of the biological age model based on the plurality of clinical variables (DNAm EMRAge).

[0060] In another aspect, the processor is further configured to determine a plurality of Epigenetic Biomarker Proxy (EBP) scores from the metabolomic data and proteomic data (OMICm Age). In a non-limiting example, the processor is configured to prove the biological age model with the plurality of EBP scores. The determination and use of the EBP scores is described in further detail in the example below.

[0061] As described previously for the method of FIG. 3, the report may include a morbidity parameter, wherein the morbidity parameter is an estimate of time until death.

[0062] In another aspect, a non-transitory CRM is disclosed. In a non-limiting example, the instructions stored on the CRM constitute the method described for FIG. 3 and the computer processor is the system and processor described above.   12   129319.01056

[0063] In a non-limiting example, the CRM may further update the biological age model further clinical variables, methylation profiles, and omic data.

[0064] As used in this specification and the claims, the singular forms “a,” “an,” and “the” include plural forms unless the context clearly dictates otherwise.

[0065] As used herein, “about”, “approximately,” “substantially,” and “significantly” will be understood by persons of ordinary skill in the art and will vary to some extent on the context in which they are used. If there are uses of the term which are not clear to persons of ordinary skill in the art given the context in which it is used, “about” and “approximately” will mean up to plus or minus 10% of the particular term and “substantially” and “significantly” will mean more than plus or minus 10% of the particular term.

[0066] As used herein, the terms “include” and “including” have the same meaning as the terms “comprise” and “comprising.” The terms “comprise” and “comprising” should be interpreted as being “open” transitional terms that permit the inclusion of additional components further to those components recited in the claims. The terms “consist” and “consisting of” should be interpreted as being “closed” transitional terms that do not permit the inclusion of additional components other than the components recited in the claims. The term “consisting essentially of” should be interpreted to be partially closed and allowing the inclusion only of additional components that do not fundamentally alter the nature of the claimed subject matter.

[0067] The phrase “such as” should be interpreted as “for example, including.” Moreover, the use of any and all exemplary language, including but not limited to “such as”, is intended merely to better illuminate the invention and does not pose a limitation on the scope of the invention unless otherwise claimed.

[0068] Furthermore, in those instances where a convention analogous to “at least one of A, B and C, etc.” is used, in general such a construction is intended in the sense of one having ordinary skill in the art would understand the convention (e.g., “a system having at least one of A, B and C” would include but not be limited to systems that have A alone, B alone, C alone, A and B together, A and C together, B and C together, and / or A, B, and C together.). It will be further understood by those within the art that virtually any disjunctive word and / or phrase presenting two or more alternative terms, whether in the description or figures, should be understood to contemplate the possibilities of including one of the terms, either of the terms, or both terms. For example, the phrase “A or B” will be understood to include the possibilities of “A” or “B” or “A and B.”  13   129319.01056

[0069] All language such as “up to,” “at least,” “greater than,” “less than,” and the like, include the number recited and refer to ranges which can subsequently be broken down into ranges and subranges. A range includes each individual member. Thus, for example, a group having 1-3 members refers to groups having 1, 2, or 3 members. Similarly, a group having 6 members refers to groups having 1, 2, 3, 4, or 6 members, and so forth.

[0070] The modal verb “may” refers to the preferred use or selection of one or more options or choices among the several described embodiments or features contained within the same. Where no options or choices are disclosed regarding a particular embodiment or feature contained in the same, the modal verb “may” refers to an affirmative act regarding how to make or use an aspect of a described embodiment or feature contained in the same, or a definitive decision to use a specific skill regarding a described embodiment or feature contained in the same. In this latter context, the modal verb “may” has the same meaning and connotation as the auxiliary verb “can.”

[0071] The example below provides a non-limiting example of just one development of a system and method implementing that which is described herein. Example

[0072] To develop and validate EMRAge, DNAm EMRAge, and OMICm Age, data was used from different cohorts (Table 1).14   129319.01056 Table 1

[0073] First, all the participants in the Massachusetts General Brigham (MGB) Biobank were used to develop the clinical clock EMRAge. Second, a subset of the MGB Biobank called MGB Aging Biobank Cohort (MGB-ABC) was used to develop DNAm EMRAge and OMICm Age. Finally, DNAm EMRAge and OMICm Age were validated in an independent cohort, the TruDiagnostic Biobank (FIG.4).

[0074] Development of EMRAge

[0075] For the development of EMRAge, 30,884 participants were used from the MGB Biobank from which all clinical phenotypes from electronic medical records (EMRs) were available (FIG. 5 and Table 2). Briefly, FIG.5 shows a flowchart 500 for the participant selection. At step 502, all MGB Biobank participants with plasma samples were considered (N = 60,370). At steps 504 and 506, participants under 18 years old and those missing vital status were excluded, respectively, resulting in 59,213 adult participants with verified vital status (step 508). At step 510, and additional 30,480 participants were excluded due to missing data on any clinical phenotypes, resulting in 28,733 participants with complete data on all clinical phenotypes (step 512). This cohort was used for selection of clinical phenotypes. At step 514, 2,151 participants with complete data on the selected clinical phenotypes were add back, resulting in a final set of 30,884 participants with complete data on the selected clinical phenotypes (step 516). Details of the selection and exclusion steps are described in further detail below. EMRAge DNAmEMRAge Biomarkers validation OMICmAge Study population MGB Biobank MGB - Aging TruDiagnostic Biobank Cohort Biobank N 30,884 3,451 12,666 Sex, male, N (%) 13,759 (44.6%) 1,250 (36.2%) 7486 (59.1%) Age in years, mean (range) 56.6 (18.0 - 98.6) 58.5 (18.3 - 53.2 (12.9-100) 96.3) Race / Ethnicity, N (%) White 25,940 (84.0%) 2,802 (81.2%) 9718 (76.7%) African American 1,785 (5.8%) 250 (7.2%) 184 (1.5%) Asian 710 (2.3%) 59 (1.7%) 744 (5.9%) 15   129319.01056 Other 1,700 (5.5%) 263 (7.6%) 2020 (15.9%) Unknown / Missing 749 (2.4%) 77 (2.2%) 0 (0%) BMI* (kg / m2), mean (range) 28.7 (12.6 - 83.0) 30.0 (15.5, 69.0) 25.4 (6.9-60.3) Tobacco Use, N (%) Yes 1,752 (5.7%) 212 (6.1%) 537 (4.2%) No* 29,132 (94.3%) 3,239 (93.9%) 11918 (94.1%) Missing 0 (0%) 0 (0%) 214 (1.7%) Alcohol Drink, N (%) Yes 17,941 (58.1%) 1,619 (46.9%) 10,086 (79.6%) No 11,516 (37.3%) 1,761 (51.0%) 2,366 (18.7%) Missing 1,427 (4.6%) 71 (2.1%) 214 (1.7%) Prevalence of comorbidity at baseline, N (%) Stroke 2,073 (6.7%) 286 (8.3%) 100 (0.8%) Type-2 Diabetes 5,582 (18.1%) 1,139 (33.0%) 258 (2.0%) COPD 2,391 (7.7%) 865 (25.1%) 59 (0.5%) Depression 8,572 (27.8%) 1,570 (45.5%) 44 (0.3%) Cancer 5,539 (17.9%) 691 (20.0%) 1301 (10.3%) CVD (excl. stroke) 10,501 (34.0%) 1,892 (54.8%) 4301 (33.9%) Vital Status, as of Jul 282023, N (%) Deceased 2,405 (7.8%) 418 (12.1%) - Alive 28,479 (92.2%) 2,886 (83.6%) - Unknown - 147 (4.3%) - Table 2 *BMI: Body Mass Index. BMI is based on median values of BMI records within 5 years around the first plasma collection for the MGB Biobank samples. *For tobacco use, the “No” level combines the non-smokers and the former smokers.

[0076] Among them, 56% were female and 84% were white. The clinical characteristics of these individuals are described in Table 1. The total amount of 30,884 individuals were divided into 16   129319.01056 training and testing sets using a 70:30 ratio, and then fitted a Cox proportional hazards (Cox-PH) model in the training set to estimate the weightings (i.e., coefficients) of the 19 selected features, as shown in Table 3. Clinical Variable Estimated Weighting Age 0.0374 BMI -0.0148 Blood Urea Nitrogen 0.0113 Charlson Comorbidity Index - 6 0.8261 Charlson Comorbidity Index - 7+ 1.2362 Diastolic Blood Pressure 1.1341 Eosinophil count -0.0178 Female -0.0064 Glucose -0.3938 Hematocrit 0.0047 Albumin -0.0405 Alkaline phosphatase -0.7805 Alanine transaminase 0.0021 Aspartate transaminase -0.0073 Neutrophil count 0.0051 Platelet count 0.0004 Red Cell Distribution Width -0.0015 Current-smoker 0.2021 Triglyceride 0.2355 Table 3

[0077] Then, the risk estimate was calculated for each individual in the training and test sets by linearly combining the estimated weightings and predictor values. This estimate was transformed into the EMRAge to have the same mean and variance as the chronological age of the training population, as done before. EMRAge was further estimated in both training and testing sets, in which both showed > 0.75 Pearson correlation to chronological age (Training correlation = 0.75, p < 0.001; Test correlation = 0.76, p < 0.001, see FIG.6). 17   129319.01056

[0078] We also tested the association between the EMRAge value and various aging-related health outcomes, including all-cause mortality, stroke, type-2 diabetes, COPD, depression, other cardiovascular diseases (CVD), and any type of cancer. FIG.7A shows that all the health outcomes are significantly associated with higher hazard and odds ratios based on EMRAge. The highest hazard ratio was seen for all-cause mortality, followed by stroke and COPD. For the odds ratio, type-2 diabetes showed the highest values, followed by CVD and COPD. Train and test sets showed very similar results for all the health outcomes.

[0079] Furthermore, the EMRAge predictors were validated by re-training the algorithm at four time points: Jan 1st of 2008, 2010, 2012, and 2014 using imputed data. During each phase, the chosen phenotype values were adjusted to median observations within a year, excluding invariant CCI indicators. The four derived equations were then applied to participants (N = 11,673) on Jan 1st of 2016. Pearson's correlations among estimates were nearly 1, affirming the robustness of the predictors (FIG.7B).

[0080] EMRAge vs. PhenoAge

[0081] We calculated PhenoAge for the MGB Biobank cohort using the established toolkit, “prcPhenoAge” described in Levine ME, Lu AT, Quach A, Chen BH, Assimes TL, Bandinelli S, Hou L, Baccarelli AA, Stewart JD, Li Y, Whitsel EA, Wilson JG, Reiner AP, Aviv A, Lohman K, Liu Y, Ferrucci L, Horvath S. An epigenetic biomarker of aging for lifespan and healthspan. Aging (Albany NY). 2018 Apr 18;10(4):573-591. doi: 10.18632 / aging.101414. PMID: 29676998; PMCID: PMC5940111, which is incorporated herein by reference. Although PhenoAge is derived from eight clinical lab metrics, one specific parameter, C-reactive protein, is not frequently ordered in routine clinical settings. To maximize sample retention, the same strategy was employed to retain the median value of observations over a 5-year window centered on the plasma collection date. Following this imputation, the sub-cohort consisted of 17,093 participants, with 11,945 samples in the training set and 5,148 in the testing set.

[0082] As illustrated in FIG.7C, participants with older EMRAge or PhenoAge values exhibit a markedly higher mortality risk than their younger counterparts. However, EMRAge more effectively discriminates between populations with high versus low survival probabilities, particularly among the oldest demographic. Furthermore, after adjusting for chronological age, sex, race, smoking status, alcohol consumption, and BMI, EMRAge presents a stronger or comparable association with the onset of aging-related diseases, as detailed in FIG.7D. 18   129319.01056

[0083] Development of DNAm EMRAge

[0084] After developing the EMRAge measure, a DNAm surrogate predictor of EMRAge was next created using matched EPIC array data (DNAm EMRAge). To this end, the MGB-ABC cohort was used, which is a subset of the MGB Biobank that was created with the aim of possessing a proportionate aging biobank population. A total of Methylation values from a total of 4,803 samples were generated using the EPICv1 array. To allow for training, samples were then selected for having EMRAge quantified and the availability of chronological age and sex information, which retained 3,451 samples. An elastic net regression model (alpha=0.1) was applied to select the variants in the predictive model for EMRAge. The model for DNAm EMRAge included 1,097 CpG sites and age as predictors. A 25-fold cross validation showed an R2= 0.827, suggesting good concordance in prediction. However, to further assess the agreement between DNAm EMRAge and the EMRAge, the data was resampled to identify a new training data set composed of samples used to generate the model and samples not in the model (N = 2,762). Within the training data, the resulting DNAm EMRAge and EMRAge values showed high correlation (N = 2,762, R2= 0.82, p<2.2e-16, Rho = 0.91, p<2.2e-16, see Figure 3). Finally, a smaller test dataset was also used which was not used for training to assess concordance; comparable correlations were found within this test set (N = 689, R2= 0.83, p<2.2e-16, Rho = 0.91, p<2.2e-16, see Figure 3). Finally, the MAE between DNAm EMRAge to EMRAge is 8.33 years in the training set and 8.50 years in the testing set. The intraclass correlation coefficient (ICC) was 0.995 (FIGS.8A-8C).

[0085] Development of OMICm Age

[0086] The last step consisted in incorporating multiple layers in a clock which predicts EMRAge. To this end, Epigenetic Biomarker Proxy Scores (EBPs) were developed for metabolites, proteins, and clinical variables.

[0087] Development of Epigenetic Biomarker Proxy (EBP) scores

[0088] We elected to utilize high throughput and unbiased methods to quantify metabolite and proteomic data. To this end, untargeted global plasma metabolomics profiling was utilized using the Metabolon platform. After preprocessing and scaling, the final dataset consists of 1,459 metabolites across a sample size of 1,986 samples, among which 1,691 were matched to unique methylation data. Majority of these metabolites belonged to the Lipid (N = 561) and Xenobiotic (N = 400) super pathways, while the least number belonged to the Energy (N = 11), Carbohydrate (N = 28), and partially characterized Molecules (N = 29). Further analysis of the subpathway 19   129319.01056 categories of the metabolites further characterized the majority of metabolites belonging to food or plant components (N = 96), analgesics and aesthetic drugs (N = 37), chemicals (N = 39), lysophospholipids (N = 49) and fatty acid dicarboxylate (N = 35) (FIG.9A). To generate unbiased proteomic data, the Seer SP100 platform was used, combined with liquid chromatography mass spectrophotometry. The final processed dataset consists of 2,805 non-unique, or 536 unique, protein groups, across 1,789 samples, in which the majority of samples (N = 1,475) matched to methylation data (FIG.9B).

[0089] Using samples which were matched in methylation, clinical, protein, and metabolite datasets, Epigenetic Biomarker Proxy scores (EBPs) were generated next. Given the wide array of protein groups, metabolites, and clinical outputs, we focused on identifying covariates which exhibited significant correlation (r > 0.1, p < 0.05) to EMRAge. This resulted in the 25 clinical outputs, 299 metabolites, and 110 protein groups that were then used to train elastic lasso- penalized models. Based on the correlation between the observed and the predicted values, those EBPs with a significant Pearson correlation above 0.2 were retained. In total, 21 clinical EBPs, 267 metabolite EBPs, and 107 protein group EBPs were retained.

[0090] Creation of OMICm Age clock

[0091] Previous studies have shown that the incorporation of estimated values for proteins and immune cell changes can improve epigenetic clock associations to disease and overall health risk, thereby generating biologically informed metrics. Therefore, a new lasso regression model was trained, named OMICm Age, which considers methylation CpG values, relative immune percentages of 12 immune cell subsets, the 396 highly correlated EBP score estimates, and demographic information such as age and sex. Based on previous literature, performed a penalized model for all the predictors. This model retained 990 CpGs, 40 EBP (16 protein EBP, 14 metabolite, and 10 clinical EBP), and age as significant predictors of EMRAge values. The weights of a predictor can be calculated, as one non-limiting example, as described in co-pending US Application Serial No.63 / 472,428, filed June 12, 2023, which is incorporated herein by reference, and reflected as follows: weights CpG MRS Output20   129319.01056 Glucose_MRS_ 0.073043562 Glucose_MRS_ #N / A predicted predicted HbA1c MRS 1037072289 HbA1c MRS #N / A ne129319.01056 H2B1N_HUMAN; H2B2F-2_HUMAN; H2B2F HUMAN;

[0092] To assess the performance of OMICm Age, FIGS. 8A-8C show the correlation between EMRAge and OMICm Age in the training (N = 2,762, R2= 0.83, p<2.2e-16, Rho = 0.91, p<2.2e- 16) and testing sets (N = 689, R2= 0.84, p<2.2e-16, Rho = 0.92, p<2.2e-16), and the intra-class correlation coefficient (ICC) using 30 replicates (0.998). In terms of error, the MAE between OMICm Age and EMRAge was 4.96 years in the training set and 4.97 years in the testing set. It is important to note that OMICm Age showed lower MAE than DNAm EMRAge (8.33 and 8.50, respectively).

[0093] Inferring the underlying biology from selected CpG sites in the OMICm Age clock 22   129319.01056

[0094] To understand the biology behind the 990 CpG sites included as predictors in the selected OMICm Age clock, an enrichment analysis was performed using the GREAT software to identify overrepresented or enriched Gene Ontology (GO) terms. The top GO terms included anatomical structure morphogenesis (p-value=9.8 ^10-13) and development (p-value=8.7 ^10-11), single- multicellular organism process (p-value=1.1 ^10-11), and nervous system development (p- value=7.6 ^10-9), among many others. Regarding the significant molecular functions, multiple terms were identified that are involved in the regulation of transcription, such as sequence-specific DNA binding (p-value=1.4 ^10-12) and RNA polymerase II transcription (p-value=2.0 ^10-10). Interestingly, the only cellular component identified as enriched was synapse (p-value=6.7 ^10-6).

[0095] Second, the CpG sites used for estimating DNAm EMRAge and OMICm Age were compared to the CpG sites used in previous clocks. FIG. 10 shows that DNAm EMRAge is the clock with the highest number of CpG sites, from which 660 are unique. OMICm Age contains 657 that are not contained in any other clock. Interestingly, 411 CpG sites are shared between these two clocks. While PhenoAge and Horvath clock share 50 CpG sites and Horvath and Hannum share 29, the maximum number of probes shared between OMICm Age and any previous clock is 3.

[0096] Inferring the underlying biology from selected EBPs in the OMICm Age clock

[0097] Previously, epigenetic clocks like GrimAge2, have shown that incorporation of additional EBPs for covariate measures can improve hazard ratios to disease. This might be due to incorporating more of the biological signals of age related disease. In OMICm Age, 40 EBPs were selected for inclusion. This included methylation based prediction algorithms of 16 proteins, 14 metabolites, and 10 clinical values, distributed along 8 biological systems (FIG. 11). The description of these EBPs can be summarized as predictors of the measured metabolites, proteins, and clinical values which were quantified by traditional methods and each has varying degrees of accuracy with Pearson correlations from .25-.80. The EBPs were highly accurate compared to previous versions of these predictors found in prior clocks. For instance, GrimAge 2 incorporated DNAm logCRP (r=0.48) and DNAm logA1C (r=0.34). EBPs developed for OMICm age had Pearson correlations of r=0.66 for CRP and r = 0.63 for HbA1C, respectively. Additionally, OMICm Age is the only epigenetic clock to have metabolite trained and incorporated EBPs.

[0098] Many of the metabolites and proteins EBPs which were selected for inclusion into OMICm Age have very few current clinical testing applications. In this analysis, the most well represented 23   129319.01056 systems include cardiovascular related biomarkers, inflammation biomarkers and endocrine associated biomarkers. This is not surprising as the prevalence of heart disease and metabolic dysfunction were higher than other types of disease.

[0099] In order to help understand the relationship between these EBPs which were selected in OMICm Age, an analysis of each EBPs to diseases was performed. An independent cohort as well as the cohort where the clock was developed (MGB-ABC), the TruDiagnostic Biobank, were used. The TruDiagnositc Biobank is a collection of 13,109 individuals who took the commercial TruDiagnostic TruAge test and had their methylation data generated. As this testing is priced to the consumer at approximately $475, this study cohort is relatively more affluent than random sampling or traditional banked cohorts. The majority of these samples were performed under a healthcare provider’s recommendation and guidance while less than 5% were in a direct-to- consumer setting. As a result, these individuals may experience a self selection bias whereby they seek preventative medicine and have fewer comorbidities than normal patient populations. In total, 12666 participants met the quality control criteria. These participants were aged between 12 and 100 years old and 59.1% were male (Table 1). During the recruitment of participants, they were asked to complete a survey that included questions about personal information, medical history, social history, lifestyle, and family history.

[0100] For calculating the associations between the EBPs and diseases, all the EBPs were standardized and linear regression models evaluated adjusting by age, sex, ethnicity, BMI, and tobacco smoking. Significant links were found between classical biomarkers of disease and the EBPs surrogates. This includes a positive relationship between DNAmGlucose and DNAmHbA1c with type 2 diabetes. inverse associations between FEV1 and COPD were also seen. Together, the directionality of these EBPs surrogates and their associated disease states are consistent with classical biomarkers of disease. This is true in both cohorts, however, the number of significant disease associations to the EBPs surrogates were more abundant in the MGB-ABC (FIG. 12A) than in the TruDiagnostic Biobank (FIG.12B).

[0101] Given that changes in these EBP scores can also change the OMICm Age output, an analysis was also performed of which lifestyle and demographic features were associated with these EBP scores in the MGB-ABC and TruDiagnostic Cohorts (FIGS.12A-12B). Here links were found which also match traditional associations to lifestyle and associated biomarkers. For instance, it was found that exercise is inversely correlated with DNAm EBPs for Glucose, Alkaline 24   129319.01056 Phosphatase, RDW, HbA1C, but positively correlated with Albumin and hematocrit. Additionally, the association of BMI to DNAm Surrogate EBPs showed higher predicted levels of glucose, HbA1c, Triglycerides, RDW, and negative associations to IGF binding protein 2 and Serum paraoxonase and arylesterase 1 (PON1). This is consistent with literature on each biomarker as IGF binding protein 2 and have been shown to be decreased in obesity. Surprisingly, some of the largest effect sizes for these DNAm EBPs surrogates were observed in connection with male sex. With Triglycerides, BUN, and Glucose being highly associated with males.

[0102] Together, the independent associations of the DNAm EBPs surrogates suggest congruence with traditional associations of disease and lifestyle features to the metabolite, protein, and clinical measurements of interest. Additionally, this suggests that the EBPs are effective at capturing the underlying biological relationship of the original biomarkers.

[0103] Comparison of OMICm Age to previous clocks

[0104] The first step to compare the performance of the DNAm EMRAge and OMICm Age clocks against previous clocks was to calculate the hazard and odd ratios to diseases. As previous clocks, PC versions for GrimAge, PhenoAge, Hannum, Horvath pant tissue, and Horvath blood & skin were included. FIG.13A summarizes the odds ratio and hazard ratio measured for all the clocks in the MGB-ABC testing set). The highest odd ratio for Type-2 Diabetes and stroke was OMICm Age, followed by DNAm EMRAge. For CVD and depression, the highest odd ratio was shown by DNAm EMRAge, followed by OMICm Age. PC GrimAge had the highest odd ratio for COPD and first-generation clocks (PC Horvath pan tissue, PC Horvath skin and blood, and PC Hannum) had the highest values for cancer. Regarding hazard ratios, OMICm Age showed the highest values for Type-2 Diabetes, Stroke, and CVD. All-cause mortality, depression, and COPD were better predicted by DNAm EMRAge but the second clock was OMICm Age. The unique disease were OMICm Age and DNAm EMRAge were not on the top clocks was cancer, which was led by PC GrimAge and PC Horvath pan tissue.

[0105] Second, the Area Under the Curve (AUC) was calculated in 5-year and 10-year survival prediction classifiers for OMICm Age, DNAm EMRAge, PC GrimAge, and chronological age (FIG.13B). In both classifiers, DNAm EMRAge showed the highest AUC values (5-year: 0.894, 10-year: 0.889), followed by OMICm Age with very similar values (5-year: 0.889, 10-year: 0.874). PC GrimAge and chronological age had much lower AUC values. 25   129319.01056

[0106] Finally, the correlation between all the epigenetic clocks with the immune subsets were tested using the TruDiagnostic Biobank. As seen in FIG. 14, all the clocks were negatively correlated to CD4T naive cells (R between -0.36 and -0.22) and CD8T naive cells (R between - 0.58 -0.30), except PC DNAm telomere length which was positively correlated (R=0.43 for CD4T naive cells and R=0.47 for CD8T naive cells). Moreover, OMICm Age shows positive correlation to CD4T memory cells (R=0.14, p-value<2.2 ^10-16) and monocytes (R=0.15, p-value<2.2 ^10-16), and negative correlation to T regulatory cells (R=-0.11, p-value<2.2 ^10-16). In the same figure, it is also displayed the correlation between the clocks with sex (female and male). Interestingly, OMICm Age and DNAm EMRAge present the highest correlations to sex (R=0.28 and R=0.36, respectively).

[0107] Association of lifestyle factors to OMICm Age

[0108] To test the adaptability of OMICm Age clock to lifestyle factors, the estimated effect of multiple factors in the MBG-ABC and the TruDiagnostic Biobank were analyzed. As seen in FIG. 15A, being male, smoking tobacco, consuming recreational drugs, being chronologically older, and having higher Body Mass Index (BMI) were associated with a higher OMICm Age in the TruDiagnostic Biobank. Besides, doing physical activity, a high sexual activity, and supplementing with fish oil omega and antioxidants were associated with a lower OMICm Age. Finally, the association with exercise type, level of education, main diet, stress level, hours of sleep, caffeine use, marital status, given birth, digestive support, workout, and herbal, minerals or vitamin supplements were also tested but the associations were not significant after adjusting by multiple comparisons (FIGS.15A and 15B). To see the consistency between cohorts, the effect of the available factors on OMICm Age in the MGB-ABC was also evaluated. In this case, very similar estimated effects were observed for sex, tobacco smoking, chronological age, BMI, and physical activity (FIG.15B). However, while the level of education was not associated with any change in OMICm Age in the TruDiagnostic biobank, a higher education was significantly associated with a reduction of OMICm Age in the MGB-ABC.

[0109] DISCUSSION

[0110] Since identifying the strong correlation between DNA methylation and age, the development of epigenetic clocks has been attempted for the understanding multiple measures of aging. To date, the overall utility of these clocks has continued to expand with the consideration of biological data extending beyond the epigenome alone, whether it be clinical, laboratory, or 26   129319.01056 system-specific measures. The contemporaneous measurement and analysis of dense molecular (i.e., “omic”), clinical, and physiologic data has been shown to be instrumental in understanding multiple disease processes; however, to date, the use or multiple omic data types in the development and / or interpretation of aging clocks has been limited. This work introduces several novel developments for the field of aging. First, a reproducible and highly predictive aging phenotype - EMRAge - is developed that distills measures of both health and wellness and mortality into a single measure. EMRAge demonstrates stronger incident associations with chronic disease outcomes and mortality than previous aging phenotypes. Second, epigenetic, proteomic, metabolomic, and clinical data is integrated into the development of a multi-omic aging clock measured, OMICm age, that can be calculated through DNAm alone. Through the use of epigenetic biomarker proxies (EBP), complex multi-omic relationships are simplified into a measure generated with methylation data alone, making OMICm age versatile and more easily translated into a clinical setting. The clinical relevance of OMICm age is further demonstrated in its association with chronic disease outcomes and mortality, outperforming current DNAm-based aging clocks. Finally, through the availability of multi-omic data, the molecular interrelationships are examined between these omics to further elucidate important mechanisms of aging that, to date, have yet not been explored.

[0111] Previous aging phenotypes have been developed using either clinical data or mortality prediction models alone. Recognizing that both clinical measures that reflect overall health and the overall risk of death are critical yet distinct attributes of aging, EMRAge was developed, a hybrid aging phenotype that distills measures of both health and wellness and mortality into a single measure. A predictive model of time until death was created using ~ 60 clinical EMR measures and mortality data from ~30,000 individuals in the MBG Biobank, spanning up to 30 years. When assessing mortality, it was found that EMRAge demonstrates more accurate mortality risk prediction than either chronologic age or PhenoAge over a 10-year period. By creating multiple predictive models of EMRAge using EMR data from different points in time, strong robustness of this was demonstrated as a reproducible aging phenotype trained with common laboratory measures.

[0112] To date, aging clocks have predominantly been generated using a singular omic data type with varying degrees of accuracy. While it is clear that DNA methylation is one of the strongest predictors of aging phenotypes, other omic data, in particular proteomics and metabolomics, have 27   129319.01056 demonstrated strong predictive accuracy, with the distinct advantage of providing more tangible biological insights into the aging process. While the epigenome is central to the aging process, the functionality of epigenetic perturbations is often less clear. In contrast, proteins provide important information about immune function and inflammatory processes that are important contributors to the aging process. Changes in oxidative stress, hormones, and lipid profiles are just a few examples of the metabolic processes associated with aging. By including age-associated proteins and metabolites into the predictive model for OMICm age, the predictive space is therefore expanded to include relevant biological processes on multiple levels of systems biology that may further elucidate biological function.

[0113] One of the ongoing challenges with multiomic approaches is the complexity of integrating different omic data together and the subsequent interpretation of the findings. Moreover, final biological models that include multiple omic data types are impractical from a clinical perspective due to high costs, logistical difficulties, and impractical time delay. Therefore, while the development of OMICm age incorporates metabolomics, epigenetics, proteomics, and clinical data, OMICm age was distilled into a single DNAm based aging algorithm through the creation and inclusion of EBP for the other omic variants. The resulting algorithm, OMICm Age, showed significant improvements over previous epigenetic clocks. First is the relationship to overall mortality, as OMICm Age exhibits higher association to all-cause mortality events relative to the other epigenetic based clocks tested, including higher sensitivity death prediction with a 6.5% increase for 10-year death prediction compared to GrimAge. OMICm Age also shows high precision with an intra-class correlation of 0.998. High level of reproducibility has previously been an issue with epigenetic clocks which has previously been improved by features such as principal component algorithms. However, this high ICC value was attained without principal component algorithm feature inclusion. The inclusion of a highly accurate 12 cell immune deconvolution was also added to solve the issue of immune cell type confounding effects for these clocks. However, this was not selected as a penalized feature in the final algorithm. One potential reason why the algorithm didn’t select immune cells as an important variable as the EBP values may provide more correlative information to the prediction of EMRAge since it was developed on just clinical biomarkers. Furthermore, the inclusion of immune cell types such as neutrophil and eosinophil counts in EMRAge may have also provided enough immune based information to consider relative immune changes. 28   129319.01056

[0114] The inclusion of EBPs surrogates also adds a powerful feature for the development of novel age clocks. Previous age clocks have had difficulty explaining why aging clocks might be accelerated or decelerated in an individual. This has limited the utility of aging clocks in clinical practice, as the large heterogeneity in accelerated aging can manifest through many different biological mechanisms. OMICm Age is able to use these EBPs to show what factors might be responsible for increased aging and their relative effect size in the overall contribution to OMICm Age.

[0115] These EBPs have also independently shown association to disease and lifestyle factors which are consistent with the biological measures they are predicting. Large correlations were observed between HbA1C EBPs and diabetes risk. This is obviously expected with the classical biomarker of HbA1c, however, this congruency between expected EBPs associated to disease specific biomarkers suggests that they are able to capture the underlying biology of the classical biomarkers through DNAm alone. IGF-1 binding protein 2 is also a good example as the EBPs show differential response to overweight versus underweight individuals as has been described in previous studies. It is important to note that higher levels of significant associations were seen in the MBG-ABC than in the TruDiagnostic cohort because higher levels of comorbidities are seen in MBG-ABC.

[0116] We also see that the OMICm Age selected proteins and metabolites EBPs with classical association to aging phenotypes. Albumin for instance was a selected penalized EBP in OMICm Age. Albumin has been a well-studied protein which decreases with age. In previous analysis of centenarians, only low albumin was associated with high mortality across age groups. In OMICm Age, albumin is the most highly weighted EBPs. OMICm Age also included classical metabolites which are highly linked to aging such as androsterone sulfate. Androsterone sulfate declines with age in men and women due to andropause and menopause.

[0117] However, OMICm Age also selected many lesser-known proteins and metabolites for inclusion. This includes metabolites such as ribitol which only has been described in previous literature as metabolites which are predictive of mortality but have very little mechanistic information. Thus, through this algorithmic training method, search space for the prediction of biological aging processes has been expanded. Through the inclusion of methods to quantify multiple systems of aging biology DNAm predictors can be made more biologically relevant. 29   129319.01056

[0118] It is important to mention that the proteomic and metabolomic EBPs are not predicting quantified measures but qualitative measures. Thus, they are not direct surrogates and instead are informing us about genetic processes which are correlated to each of these measures omic values. However, the high Pearson correlations of these predicted EBPs to measured biomarkers also suggest that these are effective predictors of classical biomarkers and their relevant biology. As a result, these EBPs might improve complex disease methylation risk scores in other predictive models outside of aging as well.

[0119] It is for this reason that OMICm Age was able to show better hazard ratios than EMRAge in the testing dataset. DNAm EMRAge was created only through using penalized regression to predict the score developed from 19 clinical biomarkers. While this algorithm might have selected disease specific features associated with the 19 biomarkers through penalized regression training, it did not specifically incorporate any metabolomic or proteomic features. By comparison, OMICm Age training selected the inclusion of 40 independent EBPs surrogates, with 16 of these being proteomic features and 14 being metabolomic features. When comparing DNAm EMRAge to OMICm Age, OMICm Age was the clock with highest hazard ratios in three out of seven diseases, including stroke, type-2 diabetes, and CVD. This is because of the high representation of EBPs which are correlated to these disease specific phenotypic processes such as endocrine, inflammation, and cardiovascular related EBPs biomarkers.

[0120] Comparing the CpGs included in OMICm Age to other DNAm clocks, FIG.10 shows that DNAm EMRAge is the clock with the highest number of CpG sites, from which 660 are unique. OMICm Age contains 657 that are not contained in any other clock. 411 CpG sites are shared between these two clocks. Interestingly, the maximum number of probes shared between OMICm Age and any previous clock is 3 (PhenoAge). This suggests that this method indeed captures a different biological signal than previous clocks.

[0121] The quality of the data was influenced by three primary factors: 1) the absence of complete clinical lab values on the plasma collection date; 2) potential selection bias due to reliance on a singular data source, which could impact the broader applicability of EMRAge; and 3) the unavailability of lab test results in the initial study phase, attributed to the limited use of sophisticated analyzers. To mitigate the first concern, the median lab test value was calculated from a five-year span surrounding the collection date, ensuring a robust sample size for EMRAge development. The validity of the chosen EMRAge predictors was further reinforced by the near- 30   129319.01056 perfect pairwise correlations among four reconstructed EMRAge estimates. To address the third limitation, commonly conducted lab tests were prioritized as EMRAge predictors. While this strategy might introduce a new selection bias, it effectively maximizes the sample size and maintains a manageable predictor pool. To circumvent the second limitation, OMCm Age and DNAm Age models were introduced, subsequently corroborating their strong associations with various aging-related diseases across two diverse cohorts.

[0122] We also believe there is room to improve the EBPs which were included into the OMICm Age algorithm. Over 3,000 metabolites, 28,000 peptide and protein groups, and 61 clinical variables which could have DNAm EBPs surrogates were quantified. However, due to the computation requirements, only 396 EBPs were included into the training dataset. The 396 EBPs which were included in the OMICm Age were selected from this group based on their high correlation to EMRAge and high correlation to observed values. In future analysis, many more EBPs will be included in the training of algorithms.

[0123] Additionally, the proteomic EBPs can also improve in resolution. While the DNAm EBPs proteomic surrogates have high Pearson correlation to the measured proteomic quantification, this could be improved by regressing out known genetic pQTL effects from the protein levels prior to generating the EBPs Scores. This should be done to preclude the signatures being driven by common SNP data that are invariant across the lifespan. The dataset from the MGB-ABC cohort does include whole exome sequencing data and these EBPs will be improved in future iterations.

[0124] The creation of EMRAge and OMICm Age represents a step forward in the evolution of improvement of epigenetic aging clocks. EMRAge is the first clinical biomarker-based clock trained using EMR data to the phenotype of time until death. This provides a unique EMR resource to quantify aging and longevity in large EMR populations. Additionally, OMICm Age is the first epigenetic clock to be trained on metabolomic, clinical, and proteomic data and the EBP scores which were included in this training show high association to the same factors which are known to be associated with the classical measurement of the measured biomarkers. With these unique EBP scores to capture multi omic features of aging, OMICm Age is shown to be the most predictive and precise epigenetic to date. METHODS

[0125] Discovery Cohort

[0126] Massachusetts General Brigham (MGB) Biobank 31   129319.01056

[0127] The Massachusetts General Brigham (MGB) Biobank is a large biorepository that provides access to research data and approximately 130,000 high-quality banked samples (plasma, serum, and DNA) from >100,000 consented patients enrolled in the MGB system. These patients can be linked to corresponding Electronic Medical Record (EMR) data, dating from the start of their medical history within the MGB network, in addition to survey data on lifestyle, environment, and family history. The original number of participants from the MGB Biobank who provided plasma samples was 60,371. Of these, 124 participants were excluded because they were younger than 18 years old at the time of plasma collection. Among the remaining adult participants, the vital status of 59,213 has been verified as either alive or deceased, with an accurate record of the date of death as of 07 / 28 / 2022. The other 1,034 participants were excluded due to missing verification of their vital status (FIG.5).

[0128] Massachusetts General Brigham Aging Biobank Cohort (MGB-ABC)

[0129] The MGB-Aging Biobank Cohort (MGB-ABC) is a cohort of 3,451 randomly selected participants from the MGB Biobank to create a proportionate aging biobank population. The cohort was selected based on even weighting in terms of age, sex and BMI, representative of the MGB Biobank. Demographics from this cohort can be found at Table 1. Comprehensive EMR, metabolomic profiling, proteomic profiling, Whole Exome Sequencing (WES), and epigenetics is available for all subjects in MGB-ABC.

[0130] Blood samples, collected either as part of clinical care or through research draws at Brigham and Women's Hospital (BWH) or Massachusetts General Hospital (MGH), were used for serum, plasma, and DNA / genomic research. Each blood draw typically involved collecting 30-50 ml of blood, which was linked to the corresponding clinical data from the Electronic Medical Record (EMR). The Biobank team also gathered additional health-related information during the blood draw process.

[0131] The administration of questionnaires for the study was carried out electronically or in written form, and participants spent approximately 10-15 minutes completing the surveys. The survey included questions related to family history, lifestyle, and environment. and utmost care was taken to ensure the confidentiality and security of the information. Participants' identities were protected, as no personally identifiable information was requested. The survey data was encrypted to ensure privacy. 32   129319.01056

[0132] The Phenotype Discovery Center (PDC) of MGB integrates various data sources, including the Research Patient Data Registry (RPDR), health information surveys, and genotype results, into the Biobank Portal. This portal combines specimen data with EMR data, creating a comprehensive SQL Server database with a user-friendly web-based application. Researchers can perform queries, visualize longitudinal data with timestamps, employ established algorithms to define phenotypes, utilize automated natural language processing (NLP) tools for analyzing EMR data using the Informatics for Integrating Biology and the Bedside (i2b2) toolkit, and request samples from cases and controls. Data in the Biobank Portal database includes narrative data from clinic notes, text reports (cardiology, pathology, radiology, operative, discharge summaries), codified data (e.g., demographics, diagnoses, procedures, labs and medications) as well as patient-reported data from the health information survey on exposures and family history. Validated phenotypes are available in the Biobank Portal user interface for genotyped Biobank participants. Other relevant measures such as lung function will be extracted using a validated algorithm incorporating NLP when necessary.

[0133] Metabolomic Profiling

[0134] Untargeted global plasma metabolomics profiling was generated by Metabolon Inc. Coefficients of variation were measured in blinded QC samples randomly distributed among study samples. Batch variation was controlled for in the analysis. Sample preparation and global metabolomics profiling was performed according to methods described previously. Metabolomic profiling was performed using four liquid chromatography tandem mass spectrometry (LC-MS) methods that measure complementary sets of metabolite classes described previously (ref2): 1) Amines and polar metabolites that ionize in the positive ion mode; 2) Central metabolites and polar metabolites that ionize in the negative ion mode; 3) Polar and non-polar lipids; 4) Free fatty acids, bile acids, and metabolites of intermediate polarity. All reagents and columns for this project will be purchased in bulk from a single lot and all instruments will be calibrated for mass resolution and mass accuracy daily.

[0135] Metabolite peaks are quantified using area-under-the-curve. Raw area counts for each metabolite in each sample are normalized to correct for variation resulting from instrument inter- day tuning differences by the median value for each run-day, therefore, setting the medians to 1.0 for each run. Metabolites are identified by automated comparison of the ion features in the experimental samples to a reference library of ~8,000 chemical standard entries that include 33   129319.01056 retention time, molecular weight (m / z), preferred adducts, and in-source fragments as well as associated MS spectra and curated by visual inspection for quality control using software developed at Metabolon, Inc. Identification of known chemical entities is based on comparison to metabolomic library entries of purified standards. Additional mass spectral entries will be created for structurally unnamed biochemicals, which are identified by virtue of their recurrent nature. These compounds have the potential to be identified by future acquisition of a matching purified standard or by classical structural analysis. Quality control (QC) and data processing was performed using an in-house method that has now been adopted by colleagues across the Boston Longwood Medical Area. Briefly, metabolite features with a signal-to-noise ratio <10 were considered unquantifiable and excluded, as were features with undetectable / missing levels for >10% of the samples. All remaining missing values were imputed with the half the minimum peak intensity for that feature across the whole population. Features with a CV in the pooled samples greater than 25% were excluded to ensure good technical reproducibility. Metabolite features were analyzed as measured LC-MS peak areas and were log-transformed to create approximately Gaussian distributions and to stabilize variance, and pareto scaled to account for the differences in the scales of measurements across the metabolome. After the QC, 1,459 metabolites across a sample size of 1,986 samples were used in the subsequent analyses.

[0136] Methylation profiling

[0137] DNA methylation / epigenetic data was generated using the Illumina Infinium® MethylationEPIC 850K BeadChip. The MethylationEPIC 850K BeadChip combines comprehensive coverage and high-throughput capabilities with comprehensive genome-wide coverage (greater than 850,000 methylation sites), including CpG islands, non-CpG and differentially methylated sites, FANTOM5 enhancers, ENCODE open chromatin, ENCODE transcription factor binding sites, and miRNA promoter regions. Biobanked samples were stored in -80C prior to shipment to the TruDiagnostic Inc. (Lexington, KY) for DNA extraction and preprocessing. Briefly, 500ng of DNA was extracted from whole blood samples and bisulfite converted using the Zymo Research EZ DNA methylation kit. All manufacturer’s instructions were followed. Bisulfite-converted DNA were randomly assigned to chip wells on the Infinium HumanMethylationEPIC array. Lab proprocessing included the following: 1) DNA amplification, 2) hybridization to the EPIC array, 3) stain, washing, and imaging with the Illumina iScan SQ instrument to generate raw image intensities. 34   129319.01056

[0138] Raw data was processed using the minfi pipeline, and low quality samples were identified using the ENmix qcfilter() function, using default parameters. This included the removal of low quality probes (p < 0.05 out-of-band) that were shown across samples. Among 866,239 probes, the quality control process retained 721,802, indicating that the large portion of the methylation data was of high quality. A combinatorial normalization processing using the Funnorm procedure (minfi package), followed by RCP method (ENmix package) was performed in order to minimize sample to sample variation as noted in Foox et al.2021.

[0139] Proteomics Profiling

[0140] We used the Seer proteomic platform for its ability to discover novel proteins and peptides related to chronological and biological aging. While Seer uses mass spectrometry like other proteomic platforms, its patented and novel nanoparticles strategy uses nanoparticles with different binding capabilities to isolate and extract peptides and proteins via corona covalent attachment to its surface. This technique is unique as it gives the platform the ability to detect low abundance peptides and proteins without the need to depletion of the fraction which can be 10x more expensive. Moreover, this increases the number of quantified proteins by 3–5 fold compared to depleted and un-depleted serum proteomics. The Proteograph untargeted approach is also different from other popular platforms based on aptamer technologies which only quantifies peptides which they specifically target. This also allows for an unbiased discovery of proteomic associations.

[0141] To quantify relative protein levels, protein levels across 2,000 samples were quantified - 1,600 from the MGB-ABC and 400 process controls - using the Proteograph Product Suite (Seer, Inc.) and LC-MS. Briefly, the samples were incubated with five proprietary nanoparticles which formed protein coronas on the Seer SP100 proteograph, which allowed for the capture of proteins using physicochemical binding. The resulting proteins were digested using trypsin, and relative levels were quantified using the default DIA method provided in the Protograph Analysis Software (PAS). Protein groups were ultimately considered for downstream analyses for two reasons: peptides identified can ambiguously map to multiple proteins, and combining peptides into protein groups can improve protein quantification devoid of spurious quantification. Thus, protein group data was sent through pre-processing, control based normalization, and outlier detection. This produced estimations of a total of 28,490 peptides across blood samples (average 15,239), and 10,265 (average 4,281) across the controls. Peptides were then further consolidated into 3,695 total protein groups (average 2,587) in MGB-ABC samples, and 1,360 total protein groups among the 35   129319.01056 plate controls. Following the signal drift and batch effect correction via the Quality Control-Robust Spline Correction (QCRSC) algorithm [PMID: 23455646], Log10 Transformation was applied, Pareto Scaling, and kNN imputation based on current guidelines [PMID: 34822371]. Stringent filters, including 80% protein presence, RSD-qc < 0.20%, and D-ratio < 0.70 , were utilized to reinforce data validity and reliability [PMID: 29805336]. The final processed dataset consists of 2,805 non-unique, or 536 unique, protein groups, across 1,789 samples, in which the majority of samples (N = 1,475) matched to methylation data.

[0142] Validation cohort

[0143] TruDiagnostic Biobank

[0144] The TruDiagnostic Biobank is a collection of data from 13,109 individuals who took the commercial TruDiagnostic TruAge test and had their methylation data generated. The participants were recruited between October 2020 and April 2023 and most of them are from the United States. They range in age from 12 to 100 years old and 59.1% of them are male. The samples that met the quality control requirements (N=12,670) are described in Table 2, including their demographic characteristics. This group of participants is relatively affluent compared to randomly sampled or traditional banked cohorts, likely due to the approximately $500 cost of the testing. The study involving human participants was reviewed and approved by the IRCM IRB and the participants provided written informed consent to take part in the study. During the recruitment of participants, they were asked to complete a survey that included questions about personal information, medical history, social history, lifestyle, and family history.

[0145] Methylation profiling

[0146] Peripheral blood samples were collected using a lancet and capillary method and placed in a lysis buffer for DNA extraction. Then, 500 ng of DNA was treated with bisulfite using the EZ DNA Methylation kit from Zymo Research following the manufacturer's instructions. The bisulfite-treated DNA samples were randomly assigned to a well on the Infinium HumanMethylationEPIC BeadChip, which was then amplified, hybridized, stained, washed, and imaged with the Illumina iScan SQ instrument to obtain raw image intensities.

[0147] To pre-process the DNAm data, the Meffil R package was used. During the sample quality control, samples with sex detection mismatches or outliers were excluded, as determined by differences in median chromosome Y and chromosome X probe intensities. Samples with a predicted median methylated signal were also removed that were more than three standard 36   129319.01056 deviations from the expected value. Additionally, outliers were excluded based on deviations from mean values for control probes, including dye bias, bisulfite 1, and bisulfite 2. Samples with more than 5% undetected probes (detection P-value >0.01) or fewer than three beads were also removed. After this quality control process, 12,670 individuals were left who met the standards, representing 96.7% of the original samples. Functional normalization was applied using the first 15 principal components of the control probes, and CpG sites were annotated to genes using EPIC Illumina annotation ilm10b4.hg19. Finally, a 12 cell immune deconvolution method was used to estimate cell type proportions.

[0148] Statistical analysis

[0149] Development of EMRAge

[0150] Of the 59,213 participants from the MGB Biobank, the 27 clinical phenotypes shown in Table 4 were extracted. DemographicsAlcohol drinking (No-alcohol / Former-drinker / Current-drinker) Vital Sign37   129319.01056 Albumin (g / dL) Total Bilirubin (mg / dL) Alkaline phosphatase (U / L) Alanine transaminase (U / L) Aspartate transaminase (U / L) Lipid Panel High-densityLow-density Lipoprotein (mg / dL) Triglyceride (mg / dL) CBC and Differential Hematocrit (%) Platelet Count (10^9 / L) Red Cell Width Distribution White Blood Cell count (10^9 / L) Basophils count (K / uL) Eosinophils count (K / uL) Neutrophils count (K / uL) Table 4

[0151] To address missing values and instrumental variations, the median value of all numerical observations, except height and age, were replaced with missing median values of all median observations within 5 years around the first plasma collection. This resulted in 28,733 individuals who had complete data for all clinical phenotypes and the exclusion of 30,480 participants (FIG. 5). Selected samples were divided into training and testing sets using a 70:30 ratio, and then fitted a Cox proportional hazards (Cox-PH) model in the training set to estimate the weightings (i.e., coefficients) of the selected features. To assure the generalizability of the equation, the training data was not scaled. Using the trained model, the risk estimate was calculated for each individual in the training set by linearly combining the estimated weightings and predictor values (Xβ). This estimate was further transformed into the EMRAge value using the equation below to get the same mean and variance as chronological age: a. EMRAge = 9.95006 * X?_{train} + 52.14512 b. One-hot encoding was utilized to convert categorical variables into binary variables and standardized the numerical variables. In total, 39 clinical variables predicted 38   129319.01056 time-to-death among 28,733 individuals, using a LASSO Cox regression model. To determine the optimal model, model performance was assessed based on Harrell's C index. As a result, the optimized model selected 30 clinical variables, which were then passed to a Cox Proportional-Hazard (Cox-PH) model for further filtering. Ultimately, 19 clinical variables remained as they had an adjusted p-value ≤ 0.05 in the fitted Cox-PH model. As a result, 30,884 participants with all available electronic medical records of selected clinical variables remained in the final cohort.

[0152] Since the EMRAge predictors (i.e., clinical phenotypes) were selected based on imputed data, their robustness was validated by re-training the algorithm at four different time points: Jan 1st of 2008, 2010, 2012, and 2014. During each time point, the values of the selected phenotypes were replaced with the median values of observations within a one-year timeframe, ensuring no overlap of data scans. To make a sensible comparison, the CCI indicators were excluded because they are invariant to time in the data. As a result, 4 estimating equations trained using the same predictors but different data. Then these equations were applied to calculate the EMRAge value of the participants (N = 11,673) with complete data of predictors around 1-year centered on Jan 1st of 2016, and then checked Pearson's correlations among these four estimates.

[0153] To identify and confirm these aging-related health outcomes, the following criteria were used based on data from the electronic medical records (EMR) of patients: 1) at least 2 relevant ICD-9 / 10 codes were recorded in the patient's EMR, and 2) the first and last dates of the relevant ICD codes should be at least 1 day apart. It was estimated that the incident risk of the EMRAge for all prospective adverse events in the Cox-PH model, adjusting for age, sex, race, BMI, smoking status and alcohol consumption. The odds ratio of the EMRAge was also estimated for all prevalent morbidities in the logistic regression model, adjusting for the same covariates.

[0154] Development of DNAm EMRAge

[0155] To develop the DNA methylation surrogate of EMRAge (DNAm-EMRAge), the normalized DNA methylation dataset was transposed so that CpG sites were considered features, and trained to EMRAge values. Using an 80-20 train to test split, the glmnet R package was used to train a gaussian penalized regression model using an alpha parameter of 0.1. Using a cross validation fold number of 25, an optimal lambda was selected based on the alpha parameter. Sex was classified as Gender_M (males) and Gender_F (females) using one-hot encoding, and was 39   129319.01056 included as unpenalized features along with chronological age. Features which showed a non-0 coefficient were selected for each final model.

[0156] Development of Methylation Risk Score models

[0157] To generate EBPs elastic net models, the normalized DNA methylation dataset was transposed to which the features were considered to be CpG sites. CpGs were not pre-filtered, and all CpGs that passed filter step were used for training. Training was conducted to each significant clinical, metabolite, and protein group values. To ensure the best model associated with EMRAge was generated, the glmnet R package was used to train gaussian penalized regression models across all measures. In order to maximize the number of CpGs, an alpha value of 0.1 was utilized. However, for three clinical values, an alpha of 0.5 as they produced the highest correlation between predicted and observed values. Using a cross validation fold number of 25, an optimal lambda was selected at each alpha threshold and implemented to select the features. Sex was classified as Gender_M (males) and Gender_F (females) using one-hot encoding, and was included as penalized features along with chronological age. Features that showed a non-0 coefficient were selected for each final model. The final models which produced the highest Pearson correlation to the observed values were selected to train OMICm Age.

[0158] Development of OMICm Age

[0159] OMICm Age was developed using the same matched methylation data used to develop DNAm EMRAge (N = 4,802), and using a similar 80:20 train:test split (N = 2,762) Penalized regression models using an alpha of 0.1 and the optimal lambda identified after a 25 cross- validation were used to train a single composite model among the train samples. For training, the following values were inputted as penalized features regression model and trained to EMRAge: all CpG sites present in the first and second generation of Illumina EPIC arrays; all clinical, metabolite, and proteomic EBPs estimates of the train samples; relative estimates of 12 cell immune cell subtypes; and the demographic information (Age, sex, ethnicity, and BMI). For sex, a similar one-hot encoding was used to identify males (Gender_M) and females (Gender_F). Features which showed non-0 coefficients were kept in the final multivariate model.       40

Claims

129319.01056 Claims   1. A method of determining a biological age of a patient, the method comprising: assembling metabolomic data and proteomic data related to a patient; delivering the metabolomic data and proteomic data to a system for biological age analysis; receiving a biological age of the patient from the system; and generating a report including the biological age of the patient.

2. The method of claim 1, further delivering at least one of an additional clinical parameter or genomic DNA methylation data related to the patient to the system.

3. The method of claim 2, further comprising generating the biological age of the patient using the system for biological age analysis, wherein the system includes a biological age model.

4. The method of claim 3, wherein the biological age model includes a weighted linear combination of the metabolomic data, proteomic data, and the at least one of additional clinical parameter or genomic DNA methylation data.

5. The method of claim 2, wherein the additional clinical parameter includes at demographic information, risk indices, substance use information, vital sign measurements, metabolic measurements, hepatic function measurements, lipid measurements, complete blood count, and / or differential blood count.

6. The method of claim 5, wherein the additional clinical parameter includes at least one of gender, race, age, Charlson Comorbidity Index, smoking status, alcohol drinking status, height, body mass index (BMI), systolic blood pressure, diastolic blood pressure, blood urea nitrogen, creatinine, glucose, albumin, total bilirubin, alkaline phosphatase, alanine transaminase, aspartate transaminase, high-density lipoprotein, low-density lipoprotein, triglyceride, hematocrit, platelet count, red cell width distribution, white blood cell count, basophils count, eosinophils count, and neutrophils count. 41   129319.01056 7. The method of claim 1, wherein the report provides a morbidity parameter.

8. The method of claim 7, wherein the morbidity parameter is an estimate of time until death.

9. The method of claim 7, wherein the morbidity parameter includes a comparison of the biological age of the patient with a chronological age of the patient.

10. The method of claim 9, wherein the report further includes a treatment plan based on whether the biological age is lower than the chronological age, the same as the chronological age, or greater than the chronological age.

11. A system for determining a biological age of a patient, the system comprising: a processor configured to: receive metabolomic data and proteomic data related to a patient; deliver the metabolomic data and the proteomic data to a biological age model; receive a biological age of the patient from the biological age model; and generate a report indicating the biological age of the patient .

12. The system of claim 11, wherein the processor is further configured to assemble and deliver at least one of the following additional clinical variables to the biological age model: demographic information, risk indices, substance use information, vital sign measurement, metabolic measurements, hepatic function measurements, lipid measurements, complete blood count, and differential blood count.

13. The system of claim 12, wherein the biological age model is a weighted linear combination of the plurality of the metabolic data, proteomic data the at least one additional clinical variables.

14. The system of claim 12, wherein the processor is further configured to receive at least one of the following additional clinical variables including at least one of gender, race, age, 42   129319.01056 Charlson Comorbidity Index, smoking status, alcohol drinking status, height, body mass index (BMI), systolic blood pressure, diastolic blood pressure, blood urea nitrogen, creatinine, glucose, albumin, total bilirubin, alkaline phosphatase, alanine transaminase, aspartate transaminase, high-density lipoprotein, low-density lipoprotein, triglyceride, hematocrit, platelet count, red cell width distribution, white blood cell count, basophils count, eosinophils count, and neutrophils count and deliver the additional clinical variables to the biological age model.

15. The system of claim 12, wherein the processor is further configured to receive genomic DNA methylation data from the patient deliver the genomic DNA methylation data to the biological age model.

16. The system of claim 15, wherein the biological age model further includes DNA methylation values associated with the at least one additional clinical variables.

17. The system of claim 16, wherein the DNA methylation values are CpG sites predictive of the biological age model based on the at least one additional clinical variables.

18. The system of claim 11, wherein the processor is further configured to determine a plurality of Epigenetic Biomarker Proxy (EBP) scores from the metabolomic data and proteomic data.

19. The system of claim 18, wherein the processor is further configured to prove the biological age model with the plurality of EBP scores.

20. The system of claim 11, wherein the report includes a morbidity parameter.

21. The system of claim 20, wherein the morbidity parameter is an estimate of time until death. 43   129319.01056 22. A non-transitory computer readable storage medium having stored thereon, instructions, that when executed by a computer processor, causes the computer processor to carry out steps comprising: delivering metabolomic data and proteomic data related to a patient to a system for biological age analysis; receiving a biological age of the patient from the system; and generating a report including the biological age of the patient.

23. The medium of claim 22, further delivering at least one clinical parameter and genomic DNA methylation data to the system.

24. The medium of claim 23, wherein the system for biological age analysis includes a biological age model.

25. The medium of claim 24, wherein the biological age model includes a weighted linear combination of the metabolomic data, proteomic data, and the at least one of additional clinical parameter or genomic DNA methylation data.

26. The medium of claim 24, further comprising updating the biological age model.

27. The medium of claim 22, wherein the report provides a morbidity parameter.

28. The medium of claim 27, wherein the morbidity parameter is an estimate of time to death.

29. The medium of claim 27, wherein the morbidity parameter includes a comparison of the biological age of the patient to a chronological age of the patient.

30. The medium of claim 29, wherein the report further includes a treatment plan based on whether the biological age is lower than the chronological age, the same as the chronological age, or greater than the chronological age.   44

Images