Systems and methods assessment of neonate health

Abstract

Systems and methods for assessment of neonates and applications thereof are described. Generally, systems and methods utilize analyte measurements acquired from an umbilical cord blood, which can be used as a basis to inform health status or treat the individual.

Description

SYSTEMS AND METHODS ASSESSMENT OF NEONATE HEALTHCROSS REFERENCE TO RELATED APPLICATIONS

[0001] This application claims benefit to U.S. Provisional Appl. No. 63 / 568,997, entitled “A Platform that Identifies Diseased and Healthy Signatures from Umbilical Cord Blood Metabolomics”, filed Mar. 22, 2024, the disclosure of which is hereby incorporated by reference in its entirety.TECHNICAL FIELD

[0002] The invention is generally directed toward systems and methods to assess health of a neonate.BACKGROUND

[0003] Determining the health trajectory of an infant at birth is a cornerstone of maternal, fetal, and newborn medicine. During pregnancy, the in-utero environment supports the developing fetus in preparation for transition to postnatal life through maturation that is classically assessed by gestational age. Umbilical vessels connecting the fetus and placenta are critical conduits for transporting nutrients and molecular mediators for the mutual exchange of information between the fetus and mother that direct pregnancy, reflect pregnancy trajectory, and reflect fetal development. Measuring the vast number of molecular compounds via metabolomic and lipidomic sampling has the potential to yield rich insight into both maternal and child health, revealing endogenous small molecules and intermediate metabolites important for fetal health as well as exogenous factors, medications, environmental exposures, and their effects on physiologic status.SUMMARY

[0004] Several embodiments of the disclosure are directed towards systems and methods for assessing health of neonates and infants. In many embodiments, analytes are collected from arterial and / or venous umbilical cord blood and / or placenta andmeasured; the measurements are utilized within a computational model to assess baby health. In some embodiments, an assessment can predict good health, a likelihood a poor health event, jaundice, hypoxia, respiratory distress syndrome, bronchopulmonary dysplasia, retinopathy, and drug exposure during pregnancy.BRIEF DESCRIPTION OF THE DRAWINGS

[0005] The description and claims will be more fully understood with reference to the following figures and data graphs, which are presented as exemplary embodiments of the invention and should not be construed as a complete recitation of the scope of the invention.

[0006] Fig. 1 provides a flow chart of an exemplary method to assess neonatal health.

[0007] Fig. 2 provides a schematic of an example of a processing system.

[0008] Fig. 3 provides an overview of the study to assess analytes from arterial and venous umbilical cord plasma.

[0009] Figs. 4A-4D provide a data showing endogenous metabolic associations, including arterial minus venous gradient and associations with clinical labs. Fig 4A: Metabolomic differences between arterial and venous vessels for energy molecules with example metabolites. P-values are FDR adjusted Wilcoxon rank sum tests between arterial and venous. The LogFC is arterial - venous, meaning analytes higher in veins have a negative value. Fig 4B: Metabolomic differences between arterial and venous veins for a molecule, hypoxanthine, with a primary function other than supplying energy, but still having a biological impact on the newborn. Fig 4C: Correlation network between the clinical lab values measuring cord pH and acylcarnitines. Orange nodes are acylcarnitines, and blue nodes are the clinical labs obtained through blood gas analysis. Red edges are negative correlations and blue edges are positive correlations, all significant (<0.05) after FDR correction. Fig 4D: Metabolomic correlations with individual clinical measurements. Larger dots are more significant after FDR, plotted as -Iog10.

[0010] Figs. 5A-5F provide data showing natural exogenous metabolites, diet modifiable metabolites, and microbial interactions with metabolomics. Fig. 5A: Dot plot of microbial metabolites and exogenous metabolites associated with diagnoses. Larger dotsare more significant after FDR, plotted as -Iog10. Odds ratio determined as the odds of disease based on the highest and lowest tertile of metabolite expression. Fig. 5B: Dot plot of microbial metabolites associated with newborn medications as another indication of 488 newborn phenotype. Fig. 5C: Dot plot of taurine and related metabolites associated with diagnoses. Fig. 5D: Lipid differences between arterial and venous data, with the most outlying lipid, linoleic acid, plotted, and other essential free fatty acids labeled. The LogFC is arterial - venous, meaning the analytes higher in the venous have a negative value. Fig. 5E: Newborn affected by chorioamnionitis. P-values are FDR adjusted Wilcoxon rank sum tests between diagnosed and undiagnosed newborns. A selected individual metabolite, heptanoylcarnitine, plotted. Fig. 5F: Metabolite and lipid classes Streptococcus carrier status affecting childbirth. P-values are FDR adjusted Wilcoxon rank sum tests between diagnosed and undiagnosed newborns.

[0011] Fig. 6 provides a table listing average predictive performance and standard deviation of predictive models across cross validation. (AUROC- area under the receiver operator (sensitivity and specificity) curve; AUPRC- area under the precision recall curve).

[0012] Figs. 7A-7G provide data showing the relationship between cord blood metabolomics and maternal medication usage in pregnancy. Fig. 7A) Performance of machine learning models to predict maternal medication use during pregnancy using venous cord blood metabolites. Fig. 7B: Pathway enrichment analysis showing differential activity in metabolic pathways between mothers who were given medications during pregnancy and those who were not. Larger dots are more significant after FDR, plotted as -Iog10. Fig. 7C: Volcano plot showing differences in individual metabolites associated with betamethasone use during pregnancy. Fig. 7D: SHAP plot showing associations between caffeine metabolites and predicted risk of BPD. Fig. 7E: Errors in our gestational age predictive model based on medication use in the last 14 days of pregnancy. Fig. 7F: The associations between diagnoses and the medications that are measured by our metabolomics platform. Larger dots are more significant after FDR, plotted as -Iog10. Fig. 7G: There was a significant difference between bupivacaine levels in the arterial and venous data sets. For births where bupivacaine was administered as determined by health records, there was a better uptake of DHA, an essential fatty acidfor infant development. Negative values indicate the molecule is higher in the venous dataset.

[0013] Figs. 8A-8E provide data showing associations with gestational age diagnoses and cord blood metabolomics. Fig. 8A) Preterm Birth, 28-36 weeks of gestation and neonatal jaundice associated with preterm delivery. P-values are FDR adjusted Wilcoxon rank sum tests between preterm and term newborns. Selected individual metabolites and lipids form the significant ones shown. Higher log fold means more 904 of that metabolite in the respective diagnosis. Estriol and cortisol are from preterm birth, while the ceramide (CER) was from neonatal jaundice. Fig. 8B: Post-term pregnancy, volcano plot of post term pregnancy vs non-post term with Iog2 fold change. Fig. 8C: Associations between selected sex hormones and glucocorticoids with diagnoses related to gestational age. Larger dots are more significant after FDR, plotted as -Iog10. Fig. 8D: Creating a gestational clock using the metabolomics data. Actual gestational age in weeks on the y- axis, which the predicted gestational age on the x-axis. Fig. 8E: Associations between lipids that are associated with lung surfactants and selected other medications that the newborn received.

[0014] Figs. 9A-9B provide data depicting associations between cord blood analytes and good health, and C-section delivery. Fig. 9A: Encounter for routine child health examination without abnormal findings. P-values are FDR adjusted Wilcoxon rank sum tests between diagnosed and undiagnosed newborns. Selected individual metabolites and lipids form the significant ones shown. Fig. 9B: Single liveborn infant, delivered by cesarean. P-values are FDR adjusted Wilcoxon rank sum tests between diagnosed and undiagnosed newborns. Selected individual metabolites and lipids form the significant ones shown. Red is from the venous data set and the purple indicates a different A-V gradient during C-section births for lidocaine.DETAILED DESCRIPTION

[0015] The various systems and methods of the application are related to assessment of neonate health, in accordance with various embodiments of the disclosure. An analyte sample derived from arterial and / or venous cord blood / plasma are assessed for one ormore analytes therein. It has been discovered that many analytes have an association with a health-related outcome. In some embodiments, analytes are measured and the level of such analyte is utilized to determine a health status of the neonate. In some embodiments, a computational model is trained to utilize analyte measurements to predict health of a neonate. Neonate health status can be assessed soon after birth, which can be performed as a routine assessment in which arterial and / or venous cord blood / plasma is collected. Various analytes associated with health status are also described.

[0016] The term neonate is used throughout, referring to when a cord blood sample is collected. It is to be understood that a health assessments of a cord blood sample can be a prediction of likelihood that a future health status will arise. Accordingly, assessment of a neonate is not limited to assessments of health status solely within the neonatal period, but also beyond the neonatal period.Neonate health status

[0017] An example of a method for assessing neonate health using analyte measurements derived from umbilical cord blood is shown in Fig. 1A. This example is directed to determining a health status, such as (for example) healthy or at risk for a disease. The method can be performed following birth, retrieving blood from the umbilical cord or placenta. Because umbilical cord and placenta are “waste products” of birth, utilizing the blood therein to perform an assessment of health provides an opportunity to better identify potential concerns in neonates. If a health concern is indicated, clinical diagnostics, monitoring, and / or treatment can be performed or administered.

[0018] In a number of embodiments, analytes and analyte measurements can include biological and chemical species able to be detected within a sample derived from blood. In some embodiments, analytes include metabolites derived from blood. In some embodiments, analytes include lipids derived from blood. In some embodiments, clinical data or personal can also be added to an assessment. In some embodiments, metabolites can include precursors, intermediates, and products of metabolism such as (for example) sugars, amino acids, nucleotides, antioxidants, organic acids, polyols, vitamins, hormones, steroids, microbial-derived metabolites, ingested compounds, medications,and the like. Lipids can include (but are not limited to) fatty acid molecules, fat soluble vitamins, glycerolipids, phospholipids, sterols, sphingolipids, prenols, saccharolipids, polyketides, and the like.

[0019] Based on experiments performed, such as those described herein, it is now known that a number of analytes derived from cord blood have an indication of outcome of various health status. Analyte measurements can be used to provide an indication of: a likelihood of a health complication, exposure to drug, and healthy.

[0020] Process 100 begins with measuring (101 ) analytes in a sample derived from umbilical cord blood. In many instances, a sample to be tested is provided that has been purified, enriched, and / or sampled from a cord blood sample. In some embodiments, a clinician collects an umbilical cord and / or placenta and collects blood for analysis.

[0021] The sample can be collected from any recent birth, whether there were complications or not. In some embodiments, a birth was vaginal. In some embodiments, a birth was surgical (e.g., C-section). In some embodiments, a birth was full term. In some embodiments, a birth was preterm. In some embodiments, a birth was full post-term

[0022] A number of analytes can be used to assess health status, including (but not limited to) metabolites and lipids. Analytes can be detected and measured by a number of methods, including mass spectrometry, colorimetric analysis, immunodetection, and the like.

[0023] Using measurements of analytes, Method 100 predict (103) a health status using analyte measurements. Analytes can be assessed based on one or more of the following: presence, measurement level, relative measurement to another analyte, threshold analysis, correlation analysis, statistical analysis, and machine-learning models.

[0024] Examples of health status that can be assessed include (but are not limited to): healthy (e.g., has a healthy signature), risk of poor health outcome, risk of hypoxia, risk of bronchopulmonary dysplasia, risk of jaundice, risk of respiratory failure, and drug exposure. Each of these status examples have been found to associate with one or more analytes derived from cord blood.

[0025] Correlations, statistical models, and machine-learning models can be utilized to identify which analytes are useful for assessing health status. Examples of correlation analysis include (but are not limited to) correlation strength (e.g., correlation coefficient), including linear association (Pearson correlation coefficient), Kendall rank correlation coefficient, and Spearman rank correlation coefficient. Examples of statistical analysis include (but not limited to) Wilcoxon rank-sum test, Kruskal-Wallis test, Chi-squared test, and student’s t-test.

[0026] In some embodiments, a machine learning model can be trained to use analyte measurements as features to predict a health status. A number of various machine learning architectures to yield a likelihood and / or classification of a health status. Ridge regression technique can reduce variances to better reach the true value. It should be understood, however, that other models can also be used, including (but not limited to), decision trees, ridge regression, kernalized ridge regression, K-nearest neighbors, logistic regression, LASSO, elastic net, least angle regression (LAR), random forest, neural network, and principal components analysis. Machine-learning models can be supervised, partially supervised, or unsupervised.

[0027] A machine-learning model can be trained using cohorts of cord blood samples associated with a health status. For supervised (or partially supervised) training, analyte measurements from the cord blood samples can labeled with their known associated health status and entered into the model such that it can learn to differentiate samples having the health status from samples that do not. Training a model can identify analytes that provide high prediction power. In some embodiments, analytes identified as highly correlative, statistically significant, and / or having high prediction power are utilized to train a computational model.

[0028] Based on experimental data, it has been found that several analyte measurements are correlated, statistically significant, and / or highly predictive of a health status. Various analytes have been found to be useful for the following: good health, a likelihood a poor health event, jaundice, hypoxia, respiratory distress syndrome, bronchopulmonary dysplasia, retinopathy, and drug exposure during pregnancy.

[0029] In some embodiments, an assessment of whether a neonate is of good health. By experiments performed and described herein, a neonate healthy signature has been identified (Fig. 9A). Accordingly, an assessment of good health can be performed by assessing one or more of the following (and any combination of the following): vitamin B5, tryptophan betaine, piperine, LysoPE 22:5, TAG60:12-FA22:6, 16:0 / 22:6, 18:1 / 22:6, 18:0 / 22:6, and 16:0.

[0030] In some embodiments, an assessment of whether a neonate has a chance to experience a poor health event is performed. By experiments performed and described herein, effects of various compounds were found to associate with a variety of poor health events. 3-indolepropionic acid was found to positively associate with premature birth, atelectasis (lung collapse), and abdominal distension. Gamma-glutamyl-L-putrescine was found to positively associated with jaundice and newborns affected by maternal hypertensive disorders. Hydroxyhippurate was found to negatively associate with preterm-associated jaundice and respiratory distress syndrome. Sulfolithocholylglycine was found to negatively associate with preterm -associated jaundice. Vanillin 4-sulfate was found to negatively associated with retinopathy.

[0031] In some embodiments, an assessment of whether a preterm neonate is at risk of jaundice. By experiments performed and described herein, effects of various compounds were found to associate with preterm -associated jaundice. Taurine and metabolites in similar pathways were found to associate with jaundice.

[0032] In some embodiments, an assessment of whether a neonate experienced clinically relevant hypoxic ischemia (hypoxia) is performed. By experiments performed and described herein, effects of various compounds were found to associate with hypoxia. Acylcarnitines was found to associate with hypoxia (Fig. 4D).

[0033] In some embodiments, an assessment of whether a preterm neonate is at risk of bronchopulmonary dysplasia. By experiments performed and described herein, effects of various compounds were found to associate with preterm-associated bronchopulmonary dysplasia. Caffeine and metabolites thereof (including xanthine, theophylline, 5-acetylamino-6-amino-3-methyluracil, theobromine, methylxanthine, andhypoxanthine) were found to associate with bronchopulmonary dysplasia (Fig. 7D). In some implementations, a pregnant individual is administered caffeine.

[0034] In some embodiments, an assessment of whether a mother took a drug or medication during pregnancy. By experiments performed and described herein, effects of various compounds were found to be predictive of drug or medication intake during pregnancy (Figs. 6 and 7A). Examples of compounds predictive with of drug or medication intake during pregnancy include metabolites related to steroid hormone biosynthesis and histidine metabolism (Fig. 7B).

[0035] In some embodiments, two or less features described are utilized in a machinelearning model. In some embodiments, three or less features described are utilized in a machine-learning model. In some embodiments, four or less features described are utilized in a machine-learning model. In some embodiments, five or less features described are utilized in a machine-learning model. In some embodiments, six or less features described are utilized in a machine-learning model. In some embodiments, seven or less features described are utilized in a machine-learning model. In some embodiments, eight or less features described are utilized in a machine-learning model. In some embodiments, nine or less features described are utilized in a machine-learning model. In some embodiments, ten or less features described are utilized in a machinelearning model. In some embodiments, fifteen or less features described are utilized in a machine-learning model. In some embodiments, twenty or less features described are utilized in a machine-learning model. In some embodiments, fifty or less features described are utilized in a machine-learning model.

[0036] Based on the health assessment, a neonate can optionally be further examined or treated (105). In several embodiments, a predict a health status can be indicative of whether further action is needed. In some embodiments, analyte measurement data is used to compute a health status as a general or routine neonate health assessment. In some embodiments, analyte measurement data is used to a health status yields a determination that another diagnostic test or further clinical assessment is to be performed. In some embodiments, analyte measurement data is used to a health status yields a determination that a treatment is be performed. In some embodiments, In someembodiments, analyte measurement data is used to a health status yields a determination that a neonate should be monitored for potential development of a health complication. In some embodiments, analyte measurement data is used to a health status yields a determination that a neonate is healthy, which may signify no further assessment or monitoring is to be performed.

[0037] In a number of embodiments, a treatment entails a medication, a vitamin, a dietary supplement, dietary food, or any combination thereof. Treatments can be administered directly to a neonate (e.g., in a formulation) or ingested by nursing mothers to feed to the neonate (e.g., dietary foods high in compound for treatment). In some embodiments, a further diagnostic test or assessment is performed. For example,

[0038] While specific examples of determining an individual’s exercise capacity are described above, one of ordinary skill in the art can appreciate that various steps of the process can be performed in different orders and that certain steps may be optional according to some embodiments of the invention. As such, it should be clear that the various steps of the process could be used as appropriate to the requirements of specific applications. Furthermore, any of a variety of processes for determining an individual’s VO2 max appropriate to the requirements of a given application can be utilized in accordance with various embodiments of the invention.Analyte selection for assessment

[0039] As explained in the previous sections, analyte measurements derived from cord blood can be utilized to asses a health status of a neonate. Analytes can be assessed based on one or more of the following: presence, measurement level, relative measurement to another analyte, threshold analysis, correlation analysis, statistical analysis, and machine-learning models. In some embodiments, analyte measurements are assessed by comparison to population norm or threshold analysis. In some embodiments, analyte measurements are assessed by a correlation assessment. In some embodiments, analyte measurements are assessed by statistical analysis. In some embodiments, analyte measurements are utilized as features and assessed by machinelearning model. Various methodologies can be utilized to select analytes, including (butnot limited to) correlation analysis, statistical analysis, and machine-learning analysis. In some embodiments, practical factors, such as (for example) the ease and / or cost of obtaining the analyte measurement and current clinical protocols are also considered when selecting analytes for assessment.

[0040] Correlation analysis utilizes statistical methods to determine the strength of relationships between two measurements. Accordingly, a strength of relationship between an analyte measurement and VO2 max test measurement can be determined. Many statistical methods are known to determine correlation strength (e.g., correlation coefficient), including linear association (Pearson correlation coefficient), Kendall rank correlation coefficient, and Spearman rank correlation coefficient. Analyte measurements that correlate strongly with a glycemic regulation can then be used as features to construct a computational model to determine an individual’s glycemic regulation.

[0041] In a number of embodiments, analyte measurement features are identified by a machine-learning model, including (but not limited to) a Bayesian network model, LASSO, logistic regression, decision trees, ridge regression, and elastic net. Predictive ability of features can be identified. One or more of the top predictive features can then be utilized for any type of assessment described, or can be utilized to train a model with a reduced number of features.Measuring analytes

[0042] In several embodiments, analytes are detected and measured, and based on the ability to be detected and / or level of the analyte, a health can be predicted. Analytes that can include (but are not limited to) metabolites, lipids, or any other compounds identifiable within cord blood. In some embodiments, analytes comprise metabolites and lipids. In some embodiments, analytes consist of metabolites and lipids.Detecting and Measuring Levels of Analytes

[0043] Analytes in a sample (e.g., a sample derived from cord blood) can be determined by a number of suitable methods. Suitable methods include liquid chromatography (LC), high-performance liquid chromatography (HPLC), hydrophilicinteraction chromatography (HILIC), reversed-phase liquid chromatography (RPLC), gas chromatography (GC), mass spectrometry (e.g., MS, MS-MS), NMR, enzymatic or biochemical reactions, immunoassay, and combinations thereof. For example, mass spectrometry can be combined with chromatographic methods, such as liquid chromatography (LC), gas chromatography (GC), or electrophoresis to separate the metabolite being measured from other components in the biological sample. See, e.g., Hyotylainen (2012) Expert Rev. Mol. Diagn. 12(5):527 -538; Beckonert et al. (2007) Nat. Protoc. 2(11 ):2692-2703; O’Connell (2012) Bioanalysis 4(4):431 -451 ; and Eckhart et al. (2012) Clin. Transl. Sci. 5(3):285-288; the disclosures of which are hereby incorporated by reference. Alternatively, analytes can be measured with biochemical or enzymatic assays. For example, serum glucose can be measured with a hexokinase-glucose-6- phosphate dehydrogenase coupled enzyme assay. In another example, analytes can be separated by chromatography and relative levels of an analyte can be determined from analysis of a chromatogram by integration of the peak area for the eluted analyte.

[0044] Immunoassays based on the use of antibodies that specifically recognize an analyte may be used for enrichment and / or measurement of analyte levels. Such assays include (but are not limited to) enzyme-linked immunosorbent assay (ELISA), radioimmunoassays (RIA), "sandwich" immunoassays, fluorescent immunoassays, enzyme multiplied immunoassay technique (EMIT), capillary electrophoresis immunoassays (CEIA), immunoprecipitation assays, western blotting, immunohistochemistry (IHC), flow cytometry, and cytometry by time of flight (CyTOF).

[0045] Antibodies may be used in diagnostic assays to detect the presence or for quantification of the analytes in a biological sample. Such a diagnostic assay may comprise at least two steps; (i) contacting a biological sample with the antibody, wherein the sample is or derived from blood or plasma, a microchip (e.g., See Kraly et al. (2009) Anal Chim Acta 653(1 ):23-35), or a chromatography column with bound analytes, etc.; and (ii) quantifying the antibody bound to the substrate. The method may additionally involve a preliminary step of attaching the antibody, either covalently, electrostatically, or reversibly, to a solid support, before subjecting the bound antibody to the sample, as defined above and elsewhere herein.

[0046] Various diagnostic assay techniques are known in the art, such as competitive binding assays, direct or indirect sandwich assays and immunoprecipitation assays conducted in either heterogeneous or homogenous phases (Zola, Monoclonal Antibodies: A Manual of Techniques, CRC Press, Inc., (1987), pp 147-158). The antibodies used in the diagnostic assays can be labeled with a detectable moiety. The detectable moiety should be capable of producing, either directly or indirectly, a detectable signal. For example, the detectable moiety may be a radioisotope, such as 2H, 14C, 32P, or 1251, a fluorescent or chemiluminescent compound, such as fluorescein isothiocyanate, rhodamine, or luciferin, or an enzyme, such as alkaline phosphatase, beta-galactosidase, green fluorescent protein, or horseradish peroxidase. Any method known in the art for conjugating the antibody to the detectable moiety may be employed, including those methods described by Hunter et al., Nature, 144:945 (1962); David et al., Biochem. 13:1014 (1974); Pain et al., J. Immunol. Methods 40:219 (1981 ); and Nygren, J. Histochem. and Cytochem. 30:407 (1982).

[0047] In various embodiments, analytes in a sample can be separated by high- resolution electrophoresis, e.g., one or two-dimensional gel electrophoresis. A fraction containing an analyte can be isolated and further analyzed by gas phase ion spectrometry. Preferably, two-dimensional gel electrophoresis is used to generate a two- dimensional array of spots for the analytes. See, e.g., Jungblut and Thiede, Mass Spectr. Rev. 16:145-162 (1997).

[0048] In a number of embodiments, LC (inclusive of HPLC, HILIC, and RPLC) can be used to separate a mixture of analytes in a sample based on their different physical properties, such as polarity, charge and size. HPLC instruments typically consist of a reservoir, the mobile phase, a pump, an injector, a separation column, and a detector. Analytes in a sample are separated by injecting an aliquot of the sample onto the column. Different analytes in the mixture pass through the column at different rates due to differences in their partitioning behavior between the mobile liquid phase and the stationary phase. A fraction that corresponds to the molecular weight and / or physical properties of one or more analytes can be collected. The fraction can then be analyzed by gas phase ion spectrometry to detect analytes.

[0049] After preparation, analytes in a sample are typically captured on a substrate for detection. Traditional substrates include antibody-coated 96-well plates or nitrocellulose membranes that are subsequently probed for the presence of analytes. Alternatively, metabolite-binding molecules attached to microspheres, microparticles, microbeads, beads, or other particles can be used for capture and detection of analytes. The metabolite-binding molecules may be antibodies, peptides, peptoids, aptamers, small molecule ligands or other metabolite-binding capture agents attached to the surface of particles. Each metabolite-binding molecule may comprise a "unique detectable label" which is uniquely coded such that it may be distinguished from other detectable labels attached to other metabolite-binding molecules to allow detection of analytes in multiplex assays. Examples include, but are not limited to, color-coded microspheres with known fluorescent light intensities (see e.g., microspheres with xMAP technology produced by Luminex (Austin, TX)); microspheres containing quantum dot nanocrystals, for example, having different ratios and combinations of quantum dot colors (e.g., Qdot nanocrystals produced by Life Technologies (Carlsbad, CA)); glass coated metal nanoparticles (see e.g., SERS nanotags produced by Nanoplex Technologies, Inc. (Mountain View, CA)); barcode materials (see e.g., sub-micron sized striped metallic rods such as Nanobarcodes produced by Nanoplex Technologies, Inc.), encoded microparticles with colored bar codes (see e.g., CellCard produced by Vitra Bioscience, vitrabio.com), glass microparticles with digital holographic code images (see e.g., CyVera microbeads produced by Illumina (San Diego, CA)); chemiluminescent dyes, combinations of dye compounds; and beads of detectably different sizes. See, e.g., U.S. patent No. 5,981 ,180, U.S. patent No. 7,445,844, U.S. patent No. 6,524,793, Rusling et al. (2010) Analyst 135(10): 2496-2511 ; Kingsmore (2006) Nat. Rev. Drug Discov. 5(4): 310-320, Proceedings Vol. 5705 Nanobiophotonics and Biomedical Applications II, Alexander N. Cartwright; Marek Osinski, Editors, pp.114-122; Nanobiotechnology Protocols Methods in Molecular Biology, 2005, Volume 303; herein incorporated by reference in their entireties).

[0050] Mass spectrometry is useful for detection of analytes. Laser desorption time- of-flight mass spectrometer can be used in various embodiments. In laser desorptionmass spectrometry, a substrate or a probe comprising analytes is introduced into an inlet system. The analytes are desorbed and ionized into the gas phase by laser from the ionization source. The ions generated are collected by an ion optic assembly, and then in a time-of-flight mass analyzer, ions are accelerated through a short high voltage field and let drift into a high vacuum chamber. At the far end of the high vacuum chamber, the accelerated ions strike a sensitive detector surface at a different time. Since the time-of- flight is a function of the mass of the ions, the elapsed time between ion formation and ion detector impact can be used to identify the presence or absence of markers of specific mass to charge ratio.

[0051] Matrix-assisted laser desorption / ionization mass spectrometry (MALDI-MS) can also be used for detecting analytes. MALDI-MS is a method of mass spectrometry that involves the use of an energy absorbing molecule, frequently called a matrix, for desorbing proteins intact from a probe surface. MALDI is described, for example, in U.S. Pat. No. 5,118,937 (Hillenkamp et al.) and U.S. Pat. No. 5,045,694 (Beavis and Chait). In MALDI-MS, the sample is typically mixed with a matrix material and placed on the surface of an inert probe. Exemplary energy absorbing molecules include cinnamic acid derivatives, sinapinic acid ("SPA"), cyano hydroxy cinnamic acid ("CHCA") and dihydroxybenzoic acid. Other suitable energy absorbing molecules are known to those skilled in this art. The matrix dries, forming crystals that encapsulate the analyte molecules. Then the analyte molecules are detected by laser desorption / ionization mass spectrometry.Computational Processing Systems

[0052] A computational processing system can be utilized to assess neonate health utilizing analyte measurements derived from cord blood. The computational processing system can comprise one or more processing engines to perform computational methods herein (e.g., statistical analysis, machine-learning modeling). In certain embodiments, a computational processing system is a computer, mobile phone, a tablet computer, portable computer, or any other system capable of performing assessments and / or machine learning. In some embodiments, the computational processing unit is housedwith or in direct communication with an analyte measurement device.

[0053] An example of a computational processing system is provided in Fig. 2. Computational processing system 200 includes a processor system 202, an I / O interface 204, and a memory system 206. Processor system 202, I / O interface 204, and memory system 206 can be implemented using any of a variety of components such as (for example) CPUs, GPUs, ISPs, DSPs, wireless modems (e.g., WiFi, cellular, Bluetooth modems), serial interfaces, volatile memory (e.g., DRAM) and / or non-volatile memory (e.g., SRAM, and / or NAND Flash). Memory system 206 is capable of storing applications and data, each of which are optional and / or combinable in any fashion. For example, a memory system can store statistical analysis (208), correlation analysis (210), machinelearning models and analysis (212), model parameters (214), and data (216) Data can include input data (e.g., analyte measurements), intermediary data, and output data (e.g., health status). Data can and / or commands can be communicated via I / O interface 204 (e.g., WiFi, cellular Bluetooth). When executed, the various applications are capable of configuring the processing system to implement computational processes including (but not limited to) the computational processes described above and / or combinations and / or modified versions of the computational processes described above.Applications based on neonate health

[0054] Various embodiments are directed to clinical assessments and / or treatments related to health status assessment of cord blood. As described herein, an analyte sample can be derived from an arterial and / or venous cord blood sample. One or more analytes can be measured and utilized to determine the health status, which can be utilized as a basis for performing follow-on diagnostic assessments and / or treatments of the patient.Medications and Supplements

[0055] Several embodiments are directed to the use of medications, vitamins, dietary supplements, and dietary meals to treat a neonate based on health. In some embodiments, medications, vitamins and / or dietary supplements are administered in a therapeutically effective amount as part of a course of treatment. As used in this context,to "treat" means to ameliorate at least one symptom of the disorder to be treated or to provide a beneficial physiological effect. For example, one such amelioration of a symptom could be improvement, mitigation, or prevention of a health complication.

[0056] A therapeutically effective amount can be an amount sufficient to prevent, reduce, ameliorate or eliminate the symptoms of diseases or pathological conditions susceptible to such treatment, such as, for example, diabetes, heart disease, or other diseases that are affected by exercise capacity. In some embodiments, a therapeutically effective amount is an amount sufficient to improve a neonate’s health.

[0057] When jaundice is indicated, clinical actions that performed include assessment of skin and eyes, bilirubin levels, and liver function tests. Treatments for jaundice include phototherapy, intravenous fluids, and intravenous immunoglobin.

[0058] When hypoxic ischemia is indicated, clinical actions that performed include assessment of oxygen levels in blood. Treatments for hypoxic ischemia include therapeutic hypothermia (cooling therapy) and erythropoietin.

[0059] When respiratory distress syndrome is indicated, clinical actions that performed include assessment by chest X-ray and blood gas analysis. Treatments for respiratory distress syndrome include therapeutic oxygen (e.g., via ventilator).

[0060] When bronchopulmonary dysplasia is indicated, clinical actions that performed include a dilated eye exam. Treatments for bronchopulmonary dysplasia include laser therapy, anti-VEGF injections, and surgical intervention.EXAMPLES AND DATA

[0061] The various embodiments of the disclosure will be better understood with the several examples data results performed to understand components of cord blood and relationships to neonate and infant health. Many exemplary results of analytes correlative with and / or indicative of health and disease are described. In addition, computational models to predict neonate and infant health utilizing analytes extracted from cord blood are described. Validation results are also provided.Deep Metabolic Phenotyping of Newborn Cord Blood Reveals Maternal-Fetal Interactions and Disease Risk

[0062] Deep metabolic profiling associated with maternal or neonatal clinical phenotypes at birth through a comprehensive analysis of umbilical cord blood during the critical transition window between intrauterine and extrauterine life has the potential to uncover novel molecular insights, stratify patients based on underlying biological risk driving clinical phenotypes, establish risk trajectories from birth, and generate hypotheses regarding established and novel therapeutic interventions. To date umbilical cord metabolomic investigations have been utilized to: characterize the impact of a narrow range of maternal conditions such as diabetes and hypertensive disorders on the metabolome during pregnancy, understand the relation between fetal and newborn anthropometric indices and specific metabolites, examine metabolic perturbations influenced by the events of delivery such as hypoxic ischemic encephalopathy, and to predict disorders of the newborn that may occur in the neonatal, infant, or childhood periods such as bronchopulmonary dysplasia, hypertension, and others. However, these studies have been limited to targeted metabolite or 87 finite panel analyses or have not considered both arterial and venous profiles reflecting different fetal and placental biology and therefore have not comprehensively considered maternal exposures on metabolic profiles and their effects on newborn health and disease risk. There are many common actionable newborn diagnoses such as respiratory distress syndrome, bronchopulmonary dysplasia, hyperbilirubinemia, retinopathy of prematurity and necrotizing enterocolitis for which the relationship between underlying pathobiology and cord blood metabolomics has been incompletely characterized.

[0063] To better explore the potential of cord blood profiling, to ascertain newborn health and physiology, and to understand the impact of maternal conditions and medications on newborn metabolic profiles and risk, we performed a comprehensive metabolic analysis of venous and arterial plasma from cord blood samples collected from a large number of women (536) who were healthy as well as those who had a variety of common conditions of pregnancy (i.e. hypertension, chorioamnionitis, respiratory distress). Pregnancies from both preterm and term neonates were included. Wecombined metabolomic data with detailed clinical phenotyping obtained from the electronic health records of mothers and neonates that included demographic information, detailed clinical diagnoses, laboratory values, as well as therapeutics, including medications. Through this combination we revealed many new discoveries about the relationship between maternal medications and conditions, umbilical cord metabolomics and newborn health.ResultsExperimental Design and Cohort

[0064] We metabolically profiled the plasma from cord blood collected from 536 women who delivered at Lucile Packard Children’s Hospital (LPCH). In general, these were high risk pregnancies (e.g. preeclampsia, chorioamnionitis, preterm delivery, gestational diabetes) or had potential complications of labor such that samples were sent to the hospital clinical lab for cord blood gas analysis. Indications for cord blood gas analysis often include evidence of concern for fetal or newborn health such as: meconium-stained amniotic fluid, category II or III fetal heart tracings, fetal-maternal hemorrhage, cord avulsion, placenta previa, placenta accreta, concern for fetal anemia, hypoxic-ischemic encephalopathy, and preterm delivery. All births were inborn and occurred in the Labor and Delivery unit of SCH between June 2018 and August 2021 .

[0065] The characteristics of the cohort indicate that a relatively diverse racial and ethnic population was studied. The generally high-risk delivery status of the study cohort is evident in the high chorioamnionitis (12.3%) and cesarean section rates (46%) compared to the general population. As expected, the incidence of preterm births (26%) in this cohort is also elevated compared to the general population (10-12%).

[0066] Cord blood was collected from umbilical veins and arteries, and plasma was stored at -80°C immediately following clinical testing. Sample preparation was performed using a biphasic separation and extraction protocol followed by metabolomic analysis using HILIC MS / MS and RPLC MS / MS, each run in both positive and negative mode, as well as a targeted lipid analysis using the Lipidyzer platform (Figure 3). A minimum of (1100), annotated metabolites and (559) lipids were identified in each sample.

[0067] Plasma samples were analyzed in two independent batches: 266 cords in the first and 294 in the second. A total of 394 arterial and 536 venous samples were analyzed with 375 derived from the same pregnancy, enabling a venous and arterial comparison in these patient samples. The same statistical analyses were performed on the two batches. To ensure reproducible association results, only metabolites significant in both batches (874 metabolites and 559 lipids) are reported.Analysis of Venous and Arterial Cord Blood Metabolomics Provides Insights into Fetal Metabolism

[0068] The metabolites entering and exiting the fetus at different amounts are indications of fetal metabolic priorities, which may yield insights into fetal biology and potential interventions. The collection of venous and arterial samples from the same cord enabled identification of differentially expressed metabolites in this context. Differential compounds were identified from the 375 arterial and venous sample pairs by Wilcoxon rank sum test (False Discovery Rate, FDR, in both batches <0.05). In our cohort, 185 compounds had a significantly higher concentration in the venous blood (flow from placenta to fetus) compared to arterial blood (flow from fetus to placenta), indicating utilization by the fetus. One of the most significant classes of these metabolites is energy molecules. This is consistent with hyperlipidemia of pregnancy in which all lipid classes rise to facilitate fetal delivery and support growth and development, particularly of the brain. (31 ) Venous analytes were subtracted from arterial such that negative values indicate venous enrichment, and positive values indicate arterial enrichment (A-V gradient). Significant energy molecules included glucose (LogFC = -0.36, -0.43), and many fatty acids (Figure 4A). One of the most significant fatty acids from the metabolomics was linoleic acid (LogFC = -0.90, -0.32). Additional fatty acids enriched in venous blood included hexadecadienoic acid (LogFC = -0.84, -0.39), dihydo- docosatetraenoic acid (LogFC = -0.61 , -0.52), oleic acid (LogFC = -0.60, -0.49), docosapentaenoic acid (LogFC = -0.40, -0.53), eicosadienoic acid (LogFC = -0.53, -0.37), and eicosatrienoic acid (LogFC = -0.41 , -0.34) all potential sources of energy and carbon for the fetus. The last two are eicosanoids, which are lipid based signaling molecules thatplay a role in innate immunity. Several fatty acids, like myristoleic acid, were also significant in the lipidomic data, demonstrating correlation between the metabolomic and lipidomic data. (Figure 4A) Overall, although the role of fatty acids in the developing fetus has been described previously, our high-resolution analysis reveals new species, a quantifiable preference for each species, and other energy molecules involved in this process including potentially important signaling molecules.

[0069] Molecules with other purposes than energy sources that were higher in venous than arterial blood include phenylalanyl tryptophan (LogFC = -0.55, -0.66), involved in peptide metabolism, and acetyldimethyllysine (LogFC = -0.76, -0.77), potentially a marker of epigenetic regulation. These highlight the diverse nature of the molecules that are higher in the venous blood than the arterial blood. There are too many to individually name, but the pathway enrichments below give a high level view.

[0070] A further diverse set of compounds were higher in the arterial dataset (A-V gradient) and thus potentially generated by the fetus, included L-lactic acid (LogFC = 0.23, 0.21 ), methylmalonic acid / succinic acid (LogFC = 0.39, 0.48), myo-inositol (LogFC = 0.48, 0.46), prolyl-arginine (LogFC = 1.29, 2.10), inosine (LogFC = 1.31 , 0.46), uracil (LogFC = 0.44, 0.35), and hypoxanthine (Figure 4B; LogFC = 0.40, 0.35). The purine inosine is a metabolite of hypoxanthine, and xanthines are important for preterm newborn health. (33) There are also several molecules, like prolyl-arginine, that are involved with amino acid and protein metabolism. Together, these provide a broad picture of both metabolic and catabolic processes of the fetus in utero, revealing important features of fetal metabolism not described previously.Many Differential Pathways are Evident in the Arterial and Venous Metabolites

[0071] The differential A-V metabolites represented 40 pathways common to both batches reflecting the large number of metabolic functions differentially performed by the fetus or maternal placental dyad (all FDR < 0.05). Any significant molecule with an A-V gradient positive or negative was included. Lipid pathways, representative of both an energy and structural carbon source for the fetus, are reflected in the fetal gradient. Key differences included arachidonic acid metabolism (involved in inflammation, enrichmentratio = 28.7, 8.91 ), biosynthesis of unsaturated fatty acids (ER = 50.1 , 18.5), fatty acid biosynthesis (ER = 19.1 , 3.93), glycerophospholipid metabolism (ER = 6.48, 14.3), phosphatidylinositol signaling (ER = 55.3, 81 .3), and sphingolipid metabolism (ER = 4.92, 7.41 ). Amino acid metabolic pathways were also prominent, including alanine, aspartate and glutamate metabolism (ER = 6.98, 8.61 ); arginine and proline metabolism (ER = 8.14, 11 .6); arginine biosynthesis (ER = 17.5, 17.8); beta-alanine metabolism (ER = 21.5, 29.3); D-glutamine and D-glutamate metabolism (ER = 20.1 , 20.7); histidine metabolism (ER = 8.88, 16.9); lysine degradation (ER = 3.15, 5.51 ); phenylalanine metabolism (ER = 3.51 , 5.99); tryptophan metabolism (ER = 3.88, 2.69); and valine, leucine, and isoleucine biosynthesis / degradation (ER = 21.7, 18.5 / 17.2, 11.9). Another group of metabolic pathways significantly enriched are sugar metabolism pathways, undoubtedly also an important energy source for the fetus. These include amino sugar and nucleotide sugar metabolism (ER = 16.0, 22.0); galactose metabolism (ER = 24.1 , 28.3); pentose phosphate pathway (ER = 6.69, 7.55); starch and sucrose metabolism (ER = 26.6, 29.5); and glycolysis / gluconeogenesis (ER = 21.5, 30.2). Together these pathways show that energy metabolism, embodied by fatty acid synthesis and sugar metabolism, and the building blocks of proteins, amino acids, are important metabolic pathways in the development of the fetus.Clinical Laboratory Tests Associated with Metabolites Reveals Insights into Fetal and Newborn Biology

[0072] To understand how metabolites in the cord blood associate with newborn clinical parameters, we examined whether clinical laboratory tests performed in the neonate correlated with the profiled metabolites (Spearman correlations, all FDR <0.05 in both batches). To obtain the most robust results, we performed this analysis using the venous samples, which are more numerous (536). To be considered for individual correlations, the clinical tests needed to be performed on at least 25 births and fewer than 200 (at which point they would qualify for the correlation network of all the labs together, as described below). Of the 190 clinical laboratory tests in either the mother or infant that fell into this category, five clinical tests correlated with metabolites in both batches, twotests from the mother and three from the newborn. Metabolic correlations with mother tests are described below. Additional health measurements such as weight, height, pH, and gestational age also correlated with cord metabolite signatures, and were included in a correlation network all together.

[0073] Bilirubin is a test used postnatally to identify newborns at risk for hyperbilirubinemia, which causes jaundice. During pregnancy the placenta excretes bilirubin for degradation by the maternal liver, postnatally, the immature liver assumes this function. Severe cases of hyperbilirubinemia are important to predict and treat due to the risk of brain toxicity. We sought to identify early cord blood predictors of hyperbilirubinemia, which is typically screened for in the first 3 days of life. Clinical labs for bilirubin correlated with 10 metabolic measurements from the venous dataset including 2 bilirubin degradation products (C17H20N2O5 and C17H18N2O4, average R2= 0.32 and 0.31 ), and 7 unknown metabolites. These measurements show that cord blood bilirubin degradation products correlate with postnatal infant bilirubin levels and might be useful as an early marker for bilirubin (Figure 4C).

[0074] Low serum albumin is a clinical test that predicts mortality in premature newborns. In our cohort, newborn albumin correlated with two measurements, estriol sulfate glucuronide and dehydrotestosterone glucuronide (average R2= 0.38 and 0.37; Figure 4C), both steroid hormones found in pregnant women. These hormones and albumin both correlate with gestational age, explaining this relationship. Perhaps low levels of these hormones contribute to preterm newborn mortality in low-album in newborns.

[0075] Mean corpuscular hemoglobin (MCH) is a measurement of hemoglobin, and low levels can be an indicator of anemia, which is a common complication of premature newborns resulting from incomplete iron transfer from the placenta. Eleven metabolites, all either hormones or unknowns, are negatively correlated with MCH (Figure 4C). Thus, several hormones are associated with MCH, and these are also markers associated with gestational age. The association is negative meaning anemia is more likely where these markers are high.

[0076] In the health records there are several measurements of physiology that we expected to be correlated with umbilical cord metabolomics. Birth weight, length, and gestational age (GA) all correlated with each other as well as with many metabolomics measurements (from the venous data, weight: 129 correlations, length: 94, gestational age: 184; and from the arterial data, weight: 67, length: 51 , gestational age: 141). These are likely additional proxy measures for GA. Arterial cord pH significantly inversely correlates with birth weight and GA (average R2= -0.22 and R2= -0.25), showing that signs of hypoxia are associated with preterm newborns. Together these findings show that GA and its associated measurements are associated with umbilical cord metabolomics, which is expected.

[0077] An important clinical indication for cord blood collection is to assess acid base status at birth and identify clinically significant cases of hypoxic ischemia. Severe cases can be treated with therapeutic cooling for improved neurodevelopmental outcomes. However, cord blood testing is only a snapshot of systemic metabolism and is generally most useful at extremes. We sought to therefore identify additional metabolites that associate with low pH and thus may provide insights into hypoxic physiology. Of the cords tested at Stanford, only 14 from both batches had a critical pH of <7.05, which requires immediate, verbal sharing of results with the clinical team. As expected, we found that arterial cord pH correlates with arterial cord base excess, venous cord base excess, arterial cord HCO3, venous cord HCO3, arterial cord pCO2, venous cord pCO?, and venous cord pH (average R2= 0.67, 0.53, 0.41 , 0.32, -0.70, -0.38, and 0.65; Figure 4D). We found that arterial cord pH significantly correlates with 106 metabolites, 32 of which are acylcarnitines (ACs) (Figure 4D). Acylcarnitines (AC) are intermediate metabolites of FA and accumulate in the cellular cytoplasm with the loss of beta oxidation due to hypoxia. Animal models have demonstrated that ACs accumulate within minutes of the onset of hypoxia. Fatty acid oxidation is active in the brain and ACs are thought to contribute to brain energy production and neuroprotection. Thus, the correlation of arterial pH with ACs provide a mechanistic interpretation of what occurs during hypoxia as well as a useful link for therapeutic intervention during hypoxia.Maternal Clinical Tests Reveal Umbilical Cord Metabolome Associates with Maternal Kidney Function

[0078] Two maternal clinical tests correlated with the cord metabolome (FDR<0.05 in both batches), both related to kidney function: creatinine and uric acid. Clinical creatinine values correlate with both metabolomic measurements of creatinine (average R2= 0.58); uric acid correlates with 6 different metabolites: hydroxyisovaleric acid, decatrenoylcarnitine, decadienoylcarnitine, octenoylcarnitine, and L-cystine (average R2= 0.49, 0.42, 0.48, 0.42, and 0.45). Together these results indicate that kidney function in the mother correlates with metabolites that are present in the fetus, suggesting that the mother's kidneys also ultimately act in part as kidneys for the fetus thereby modifying its metabolite levels (Figure 4C).Microbial ly Derived Metabolites and other Xenobiotics are Associated with Infant Diseases

[0079] We also examined the effects of microbial and exogenous compounds in umbilical cords on newborn and mother health, showing that mother microbiome and diet are associated with infant outcomes via the metabolome. Microbially derived and exogenous metabolites are determined using our own database and the human metabolome database. These compounds were associated with specific diagnoses (the odds ratio between top tertile metabolite levels compared to bottom tertile all FDR <0.05; see methods). We found lower levels of 3-indolepropionic acid associated with several diseases: an increase in premature birth, atelectasis (lung collapse), and abdominal distension (ORs = -3.00, -2.88, -1.81 , respectively). Several microbial metabolites are also associated with medications the newborn received (which is a proxy for the infant phenotype, all FDR <0.05). Again, 3-indolepropionic acid, detected in both positive and negative mode, and indoleacetic acid are lower in newborns receiving benzodiazepines (used for neural conditions; ORs = -1.81 , -1.60, NS), proparacaine (use to relieve eye pain; ORs = -3.26, -2.18, -2.67), central nervous system drugs (ORs = -2.01 , -1 .39, -1 .96), acetaminophen (ORs = -2.23, -Inf, NS), anticoagulants (ORs = -1.28, -1.24, -1.10), caffeine (ORs = -1.66, NS, -1.81 ), cyclopentolate phenylephrine (ORs = -3.33, -2.18, -2.74), and parenteral nutrition (ORs = -1.66, -1.21 , -1.23) (Figures 5A and 5B). These findings show that these indole short chain fatty acids, which have proven health effects in adults, are lower in ill newborns. We postulate they may provide valuable markers for neurological and other treatments.

[0080] Other microbial and exogenous metabolites were found to be higher in diseases associated with prematurity. For example, hydroxyhippurate, which is produced by Clostridium, associates with several conditions of prematurity including respiratory distress syndrome, neonatal jaundice associated with prematurity, and feeding problems of the newborn (ORs = 1.88, 1.82, 1.55). Hydroxyhippurate also associates with several drugs given to the newborn including ampicillin, anticoagulants, caffeine, CNS drugs, ferrous sulfate, parenteral nutrition, penicillin and many others (ORs = 2.38, 1.14, 1.94, 1.69, 1.48, 1.85, 1.34), again demonstrating a link between the microbiome, prematurity, and its associated diseases including neurological diseases. Other molecules that are positively associated with newborn diseases are gamma-glutamyl-L-putrescine (putrescine is a marker of vaginal microbiome dysbiosis) which associates with disorders of bilirubin metabolism and newborns affected by maternal hypertensive disorders (ORs = 3.21 , 3.99). The microbially derived vanillin 4-sulfate is positively associated with low weight newborns and retinopathy of prematurity (ORs = Inf, 4.16). Imidazoleacetic acid riboside, an exogenous molecule found in microbes and plants, is positively associated with hyperglycemics (OR = 2.88). All these microbial metabolites are associated with the diseased newborn, indicating the complex relationship between mother derived microbial metabolites and infant health. (Figures 5A and 5B). These different 357 metabolites are candidate markers for health conditions.

[0081] There were also two secondary bile acids associated with diseases. Secondary bile acids are created when the gut microbial community metabolizes primary bile acids produced by humans. They act as signaling molecules, help absorb lipids, and when altered can have dramatic effects on lipid profiles. Sulfolithocholic acid is negatively associated with ferrous sulfate treatment (OR = -1.48), which is used to treat anemia in premature babies. Sulfolithocholylglycine, the glycine conjugate of sulfolithocholic acid, is positively associated with neonatal jaundice associated with prematurity, and low birthweight newborn, 1500-1749 grams (ORs = 1.88, 4.07). Hence the conjugation of this bile acid with glycine seems to have an opposite effect on newborn health. Very little is known about how bile acids affect the fetus, and perhaps their degree of conjugation, which influences hydrophobicity, is an important component of their physiological effects. Secondary bile acids, which are tightly linked to the gut microbiome and lipid metabolism, are associated with health outcomes of the newborn.

[0082] One plant-derived diet metabolite, tryptophan betaine, is positively associated with markers of a healthy pregnancy and newborn. We used diagnoses of an infant checkup without abnormal findings and supervision of normal pregnancy as markers of a healthy newborn and pregnancy (ORs = 1 .48, 1 .47). Thus, this diet-derived metabolite shows a link between healthy newborns and diet that is mediated through the metabolome. This finding might be an important aspect for a dietary intervention targeting pregnant mothers. (Figure 5A).Taurine Metabolism is Associated with Health Outcomes in Neonates

[0083] Taurine is an amino acid excreted in bile that conjugates with bile acids and is closely related with diet and microbial interactions. It is essential in newborns. Taurine, which we detected in both negative and positive modes, is lower in several diseases of prematurity (OR of top tertile metabolite expression compared to bottom tertile, all FDR <0.05), such as anemia of prematurity, neonatal jaundice of prematurity, apnea of the newborn, low birth weight newborn, and retinopathy of prematurity (ORs = -3.26, -2.07, - 1 .77, -Inf, and -4.39). Taurocholic acid, the taurine conjugate of the primary bile acid cholic acid, is higher in two diseases of prematurity: neonatal jaundice associated with prematurity (OR = 6.20) and preterm newborn gestational age 34 completed weeks (OR = 3.02). A molecule closely related to taurine, 2-hydroxyethanesulfonate, has even broader associations and is lower in many diseases of prematurity and higher in chorioamnionitis (26 total significant associations with diseases after FDR, chorioamnionitis OR = 3.26). L-cystathionine, which is metabolically converted into taurine, is higher in the diseases of prematurity (15 diseases, i.e. respiratory distress syndrome of the newborn, OR = 4.41 ). Deficiency in this metabolic pathway is stronglyassociated with poorer neonatal outcomes in our dataset. Since taurine is essential for development of brain, eyes, and muscles of the developing newborn, it being lower in newborns who have diseases of prematurity suggests it is an important biomarker and may inform possible future therapeutic interventions (Figure 5C). Overall, it is clear that this is an important metabolic pathway for the developing fetus.Essential Fatty Acids Shown an A-V Gradient

[0084] From the lipidomic analysis six lipid species exhibited a placental difference in both lipidomic batches, all higher in venous than arterial (FDR < 0.05 both batches). The lipidomic results are unique in that half of the significant lipids were essential and therefore diet derived. Some of these fatty acids reached a very significant level with p-values at FDR = 10-30 (i.e. 18:2 FFA FDR = 1.96*10-30, 1.88*10-30). All of them were long-chain (LC) free fatty acids (FFAs): 18:1 (LogFC = -0.35, -0.44), 18:2 linoleic acid (LogFC = - 0.58, -0.69), 18:3 gamma-linoleic acid (LogFC = -0.41 , -0.45), 14:1 (LogFC = -0.15, - 0.14), 16:1 (LogFC = -0.27, -0.32), and 22:6 arachidonic acid (LogFC = -0.25, -0.19; Figure 5D). Linoleic acid is an essential polyunsaturated fatty acid (PUFA) that the fetus obtains from the maternal circulation. Linoleic acid is also a precursor to docosahexaenoic acid (DHA) which is important for neurodevelopment. These results indicated that the possible mechanism for energy and carbon supply in the late pregnancy fetus is provided by LC FFA and perhaps primarily by these six FFA species. This finding is consistent with what is known about the use of LC FA as a preferred fuel source by the third trimester fetus for growth and development. Interestingly, the 18: 1 and 18:2 FFAs, oleic and linoleic acids, respectively, are two of the most common FFAs in breast milk suggesting continuous supply of these metabolites to the newborn.CO VID Diagnosis is Associated with a Unique Metabolomic Signature

[0085] One environmental, microbial exposure in our dataset was exposure to COVID. In the second batch there were several metabolic associations with exposure to COVID- 19 (Wilcoxon rank sum, all FDR < 0.05) during pregnancy. This indicates that there was a persistent effect of COVID exposure on the metabolome of the newborns, although themetabolic signature did not make clear what the biological impact of this exposure might be. There were no associations in the first batch, which was collected before the COVID pandemic began. In the second batch three nucleosides are lower - adenosine (LogFC = -0.78), guanosine (LogFC = -0.54), and inosine (LogFC = -0.64) and several metabolites were higher, for example hydroxyphenylacetic acid (LogFC = 1 .22). Together these findings show there is a signature in the umbilical cords associated with maternal exposure to COVID, indicating a persistent effect of the viral infection that is passed to the newborn.Infections are Associated with Differences in Umbilical Cord Blood Metabolomics

[0086] Chorioamnionitis (CA) was another microbial-associated condition which demonstrated a clear metabolic signature. There were many metabolites of varying classes that associate with CA. In this condition, acetaminophen (logFC = 2.91 ; 2.81 ), ampicillin (antibiotic logFC = 2.83; 4.42), and bupivacaine, a local anesthetic (logFC = 2.66; 2.47) were higher in diagnosed newborns (Wilcoxon rank sum, all FDR < 0.05). The acetaminophen and ampicillin associations can be largely explained by the medications administered to the pregnant women upon a diagnosis of CA. Stress-related glucocorticoids were also higher in cords from births affected by CA, including corticosterone (logFC = 0.86; 0.73), cortisol (logFC = 1.40; 1.22), hydroxycorticosterone (logFC = 1.00; 1.13), and dihydrocortisol (logFC = 1.49; 1.79), indicating a CA-induced stress response. There are sex hormones that are higher in the chorioamnionitis births, such as pregnanolone sulfate (logFC = 0.94; 1.62), hydroxypregnenolone sulfate (logFC = 0.72; 1.13), pregnanediol glucuronide (logFC = 1.02; 0.99), DHEA sulfate (logFC = 0.86; 0.85), and pregnanetriol (logFC = 0.71 ; 0.51 ). Lipids were also differentially expressed in CA. Acylcarnitines (ACs, 0 short chain, 8 medium chain, and 3 long chain) were higher in the chorioamnionitis births including heptanoylcarnitine (logFC = 0.61 ; 0.51 ), dodecenoylcarnitine (logFC = 0.75; 0.59), hexadecenoyl-carnitine (logFC = 0.77; 0.76), hexenoylcarnitine (logFC = 0.60; 0.59), hydroxy-tetradecenoylcarnitine (logFC = 0.96; 0.86), tetradecenoylcamitine (logFC = 0.72; 0.64), and others. Hundreds of lipid species from the lipidomics analysis are also significantly different in chorioamnionitis (373overlapping in both batches), most of which are elevated in chorioamnionitis births (368 are increased; 357 of which are TAGs). Therefore, CA was observed to have broad metabolic effects dominated by changes in stress response as indicated by the glucocorticoids, changes in sex hormones, and 460 lipid metabolism, including ACs. Despite the fact that CA is associated with preterm birth, the metabolic patterns also have some features that more resemble that of post term birth, like elevated triglycerides (Figure 5E).

[0087] We also found a metabolic signature in cord blood of one of the most common clinical conditions of pregnancy, Streptococcus carrier status (SCS), which has potentially profound clinical effects on the outcomes of the newborn. This condition is the cause of a microbial infection. Thirty-five metabolites significantly associated with Streptococcus carrier status in the first batch, 29 in the second batch, and 15 overlapping between the two batches (Wilcoxon rank sum, all FDR < 0.05). Two of the metabolites significantly associated with SCS were ampicillin (logFC = 2.90; 4.47) and an unknown metabolite with the formula C19H23N3O4S (logFC = 4.29; 5.25) that has the same mass as hetacillin. Ampicillin is commonly given late in pregnancy prior to delivery in order to prevent the effects of streptococcal infection in the newborn as a result of inoculation during delivery. The sex hormone pregnanolone sulfate / allopregnanolone sulfate was also elevated in these births (logFC = 0.74; 1.08). The remainder of the molecules are dominated by the sex hormones, glucocorticoids, and unknowns. This finding demonstrates how this important microbial infection might be influencing, and is certainly associated with, the metabolic response of the developing fetus. (Figure 5F).Detecting Medication Use During Pregnancy from Metabolic Profiles

[0088] To determine if there are associations between cord blood metabolic profiles and maternal medication use during pregnancy, we utilized machine learning modeling to account for the high-dimensional data and to simultaneously explore multiple relationships between various medications and venous metabolites. Briefly, the maternal medical record was analyzed to determine if a specific medication was taken during pregnancy. We then used machine learning models to examine associations by inputtingmeasurements from venous cord blood metabolomics into an XGBoost model to predict if a specific medication was taken during pregnancy (see methods for additional details). For this analysis either the medication itself or their detected derivatives. Specifically, the models to determine maternal bupivacaine (prevalence = 38.3%) and maternal ampicillin (prevalence = 14.4%) administration utilized direct measurements of bupivacaine and ampicillin as the top feature. We additionally built models to predict maternal administration of hydralazine (prevalence = 3.2%), insulin (prevalence = 9.7%), heparin (prevalence = 4.2%), and betamethasone 514 (prevalence = 6.8%). Since the discovery metabolomics platform that we used cannot directly detect these drugs, the models to predict maternal administration utilized other metabolites as top features, indicating that use of those medications during pregnancy is associated with parallel metabolic changes. The ALIROC and ALIPRC for each of the models are reported in a table in Figure 6 and visualized in Figure 7A.Pathway Enrichment Analysis

[0089] Next, after identifying several strong associations between specific common maternal medications and cord metabolites, we performed a pathway enrichment analysis based on drug usage during pregnancy. Significant changes in steroid hormone biosynthesis were associated with the use of ampicillin, betamethasone, heparin, and insulin during pregnancy. There was a significant difference in histidine metabolism associated with betamethasone use. High levels of maternal histamine have been associated with negative birth outcomes such as preeclampsia, spontaneous abortion, preterm labor and hyperemesis gravidarum in part due to the general inflammation associated with these gestational complications. These results are visualized in Figure 7B.Metabolite Changes Associated with Betamethasone Use and Other Administered Drugs

[0090] Drugs given at birth as determined by clinical records were associated with different metabolic signatures. Eleven metabolites with Iog2 fold changes of greater than 1 or less than -1 were found to be significantly associated with betamethasone use aftermultiple hypothesis test correction. Betamethasone is used because it improves outcomes for premature newborns. The associated metabolites are C9H15N3O4, C21 H24N4O11 , C24H31 NO6S2, C19H22O7S, C21 H30O5S, dehydrotestosterone glucuronide, estriol glucuronide, C25H17NO5S, C22H25NO6S2, hydroxycorticosterone, hydroxyandrosterone glucuronide. The details are visualized in Figure 7C; they indicate that betamethasone affects steroid hormone and cortisol pathways, as might be expected.

[0091] There were numerous other associations between maternal drug administration as determined by clinical records (Wilcoxon rank sum, FDR < 0.05). Although these are a less precise measurement of the metabolic exposure of the fetus, when measuring the same drug, the results can be confirmatory, and for some drugs we cannot detect them in the cords. For example, bupivacaine was associated with 189 metabolites in the first batch and 315 metabolites in the second batch, with 144 overlapping. The most significant metabolite was the bupivacaine itself that we measured from the umbilical cord (LogFC = 4.47, 3.48). The bupivacaine result indicates the robustness of our system, and the many other associations potentially show the broad impacts of bupivacaine administration. Significant pathways associated with bupivacaine are steroid hormone biosynthesis, primary bile acid biosynthesis, nicotinate and nicotinamide metabolism, pyruvate metabolism, glycolysis / gluconeogenesis, ascorbate and aldarate metabolism, and pentose and glucuronate interconversions.

[0092] Magnesium sulfate, a drug given for gestational hypertension that was not detected in the cords, was associated with 216 metabolites in the first batch and 75 in the second batch, with 48 overlapping (FDR < 0.05). These associations included carnosine (LogFC = 0.735, 0.729) and dehydrotestosterone glucuronide (Log FC = -1.46, -1.38), and both of these metabolites are also associated with gestational age. Since magnesium sulfate is used to treat preterm birth, it might be working through associated pathways which include steroid hormone biosynthesis, cysteine and methionine metabolism, nicotinate and nicotinamide metabolism, lysine degradation, [3-alanine metabolism, and histidine metabolism.

[0093] Associated with insulin use were 78 metabolites in the first batch and 173 in the second batch, with 19 overlapping (FDR < 0.05). The overlapping molecules include nonanoylcarnitine (LogFC = -0.556, -1.23) and 3-indolepropionic acid (LogFC -1.50, - 1.84). The acylcarnitine demonstrates the association with insulin and lipid profile. Interestingly the microbially derived metabolite 3-lndolepropionic acid also indicates that insulin usage is associated with unique microbial metabolites. This molecule has been reported at lower levels in people with type II diabetes, which itself is linked to lower dietary fiber intake and higher inflammation. Interestingly, steroid hormone biosynthesis is significantly associated with insulin in both batches suggesting alterations in hormone signaling.

[0094] Oxytocin, which is used to induce uterine contractions, is associated with 174 metabolites in the first batch and 304 in the second batch, with 123 overlapping, including bupivacaine (LogFC = 1.99, 2.19) and dihydrocortisol (LogFC = 2.01 , 1.44). The pathways associated with oxytocin administration are steroid hormone biosynthesis, nicotinate and nicotinamide metabolism, and primary bile acid biosynthesis. Oxytocin administration most likely is associated with the induced delivery metabolic signature, but nonetheless demonstrates the lingering metabolites that associate with the induced newborn.Caffeine Metabolites Associated with Predicted Risk of Bronchopulmonary Dysplasia

[0095] Caffeine is commonly given to newborns in the NICU to treat apnea of prematurity and in-part to prevent bronchopulmonary dysplasia (BPD). We investigated the relationship between caffeine metabolites in venous cord blood and BPD risk. First, we identified all caffeine related metabolites in the cord blood: xanthine, theophylline, 5- acetylamino-6-amino-3-methyluracil, theobromine, methylxanthine, hypoxanthine, and caffeine itself. The direct measurement of these metabolites in venous cord blood were used to predict which preterm newborns would later develop BPD; the model had an AUC of 0.751 . By restricting the population to preterm newborns, we eliminate the confounding effects of the strength of association between prematurity and BPD. A SHAP plot of the model is shown in Figure 7D and shows that lower levels of caffeine metabolites,especially xanthine and theophylline, are associated with higher predicted probability of BPD as revealed using machine learning modeling. Assuming all, or nearly all, of the detected caffeine and caffeine derived metabolites originate from maternal intake, this demonstrates the profound effect of maternal ingestion of a seemingly innocuous and common use substances with established human biologic effects on fetal development and newborn risk. Lower xanthines were associated with an increased disease risk. This example speaks to the tractability of our cord blood metabolomic approach and the potential clinical utility of effecting newborn outcomes through maternal diet and lifestyle habits.Associations Between Medication Use and Metabolic Gestational Age

[0096] Above we demonstrated the ability of various maternal use medications and substances to affect the newborn metabolic profile. We also showed the robust relationship between medications (magnesium sulfate, betamethasone) used when preterm delivery is clinically threatened, and a multitude of steroid hormones associated with gestational age. Accordingly, we next sought to determine if maternal medication use is associated with changes in a metabolic model of newborn developmental maturity using gestational age as the gold standard comparator. In brief, we explored how maternal medication use in the last two weeks of pregnancy is associated with changes in predicted gestational age using metabolic modeling. Interestingly, mothers who took hydralazine (blood pressure control medication typically given for preeclampsia), betamethasone, or insulin had metabolomic profiles that resulted in higher predicted gestational age compared to actual gestational age as calculated using gold standard measures (LMP and first trimester ultrasound), as shown in Figure 7E. Thus, newborns appeared more mature using a metabolic index of development than would be otherwise reflected in their calculated gestational age by traditional means. Among the top 25 features in the gestational age predictive model, women prescribed hydralazine, which is used to treat blood pressure in the last 14 days of their pregnancy had significantly (adjusting for multiple hypothesis testing) elevated levels of C6:0 Hexanoylcarnitine and lower levels of pregnendiol sulfate. Women given betamethasone in the last 14 days oftheir pregnancy had significantly lower levels of dehydrotestosterone glucuronide, pregnendiol sulfate, and an unidentified metabolite, C24H31 NO6S2. They had significantly higher levels of sorbitol / mannitol, carnosine, and an unidentified metabolite, C7H12N2O4S. Women given insulin in the last 14 days of their pregnancy had significantly higher levels of carnosine than those who were not. Taken together, these findings suggest that the modeling of newborn developmental maturity on a metabolic scale is a valid reflection of gestational biology given the tractable measurable effects of specific medications given late in pregnancy and in response to maternal co-morbidity.Elevated Levels of Medications in Cord Metabolites Are Associated with an Increased Odds of Disease

[0097] Four of the metabolites detected in our dataset: acetaminophen, ampicillin, xanthine, and lidocaine, are either drugs or drug metabolic products. Detecting these metabolites in cords are better measurements of fetal exposure than clinical records. We sought to determine if the births with higher levels of these metabolites were significantly associated with disease relative to births with lower levels of these metabolites. There were numerous diagnoses associated with these metabolites between the bottom tertile and top tertile of expression. CA is positively associated with both acetaminophen (OR = 3.60) and ampicillin (an antibiotic, OR = 3.71 ), while ampicillin was also positively correlated with Streptococcus carrier state (OR = 3.74). Lidocaine, a common short acting local anesthetic, is positively associated with birth by C-section (OR = 1 .60), negatively associated with birth without C-section (OR = -1.64) and was most positively associated with secondary uterine inertia (OR = 5.37). C-sections are performed in response to uterine inertia, making it notable that uterine inertia has an even stronger association with lidocaine than C-sections alone. Xanthine is negatively associated with many diagnoses, like retinopathy of prematurity, neonatal jaundice, anemia of prematurity, and low birth weight newborn (ORs = -2.38, -1.52, -2.57, and -3.99). (Figure 4F) The findings of acetaminophen, ampicillin and lidocaine confirm the robustness of our system, meanwhile the xanthine associations from umbilical cords are novel. Caffeine is given to preterminfants to prevent bradycardia and other complications, and low xanthine levels, a metabolite of caffeine, might be a contributing factor to this phenomenon.Bupivacaine Shows an A-V Gradient

[0098] We sought to determine if there was an A-V gradient for the drugs that were measured metabolomically, which would indicate metabolism of bupivacaine by the fetus. Metabolomically measured bupivacaine, a local anesthetic extensively used during childbirth, was differentially detected in the cord blood metabolomic profile reflecting the fetal A-V gradient and revealing that it is absorbed and metabolized by the fetus (LogFC = -0.68, -0.71 ; Figure 7G) - a perhaps un-recognized side-effect of the wide-spread clinical use of this medication with intended exclusive effects for maternal comfort.Differential Uptake of DH A is Associated with Cord Bupivacaine

[0099] Not only did we determine the A-V gradients for every molecule, but we also investigated how medication as determined by clinical records might be altering the A-V gradient. This will demonstrate how medication given to the mothers alters the metabolic priorities of the fetus. The administration of bupivacaine is associated with a lower A-V gradient for both arachidonic acid and docosahexaenoic acid, showing that the baby is absorbing more of these important fatty acids when the birthing mother is given more bupivacaine (Figure 7G). These findings might suggest that bupivacaine administration is associated with healthier newborns and outcomes, perhaps due to the mother and infant being in lower pain distress.Effects of Preterm Birth on Cord Blood Metabolomics

[0100] Preterm birth between 28-36 weeks was significantly associated with 123 arterial and 190 venous metabolites which overlapped for both batches of cord blood data (Wilcoxon rank sum, all FDR < 0.05). One of the most significant metabolites shared between both venous batches was dihydrocortisol (logFC = -2.44; -2.08). There were many other glucocorticoids and sex hormones that were also lower in these newborns, including dehydrotestosterone glucuronide (logFC = -2.71 ; -2.45), cortisol (logFC = -2.36;-1.83), estrone (logFC = -2.34; -1.81 ), estriol sulfate glucuronide (logFC = -2.33; -1.92), and hydroxyandrosterone (logFC = -1 .64; -1 .28). These are known to be associated with gestational age and pregnancy progression in maternal serum, thus perhaps it was not surprising to find these metabolites also differentially present in cord blood at the time of delivery although their biological function in the fetus remains uncertain. Additional molecules that were found in lower levels in preterm births include bupivacaine (logFC = -1.56; -1.53) and undecanoylcarnitine (logFC = -0.82; -0.41 ). Molecules at significantly higher levels in these births include the bile acids taurocholic acid (logFC - 0.93; 0.89), chenodeoxycholic acid glycine conjugate (logFC = 0.84; 0.44), and carnosine (logFC = 1.02; 0.82). Carnosine is a dipeptide essential for muscle growth, and the bile acids are important molecules for lipid metabolism signaling. The metabolic profile of preterm birth looks like other diseases of prematurity with many shared corticosteroids and sex hormones, like dehydrotestosterone glucuronide, at significantly lower levels in the arterial and venous cord blood (Figure 8A).

[0101] Preterm birth between 28-36 weeks was significantly associated with 179 arterial and 404 venous lipids which overlapped for both batches of blood cord data. Once again, we observed an interesting pattern of a greater number of gradient specific features for the venous blood reflecting more fetal metabolism then placental. All molecular features reported are significant (FDR < 0.05). The two most significantly decreased lipids were triglycerides 48:4-14:0 (batch 1 logFC = -1.25; batch 2 logFC - 0.701 ) and 49:1-14:0. (logFC = -0.998; -0.691 ). The majority of triglycerides (337) were lower in preterm newborns including 52:4-22:4 (logFC =-1.23; -0.608) and 44:1-14:0 (logFC = -1.13; -0.942). Other complex lipids that were found in lower levels in these births include free fatty acids such as 16:1 (logFC = -0.293; -0.381 ) and diacylglycerol 18:2 / 20:4 (logFC = -0.814; -0.420) . Lipids at significantly higher levels in preterm births include cholesteryl esters such as 20:1 (logFC = 0.898; 0.668) and 18:0 (logFC = 0.749; 0.773) as well as phosphatidylcholines like 16:0 / 16:0 (logFC = 0.701 ; 0.590), ceramides such as 24:1 (logFC = 0.813; 0.545), and lysophosphatidylcholines like 20:4 (logFC = 0.390; 0.364). Hundreds of lipid species are significantly associated with preterm births, the vast majority of triacylglycerols, diacylglycerols, and free fatty acids are lower inpreterm births while most of the cholesteryl esters and phosphatidylcholines are higher in preterm births. Normal neonatal development results in a proportionally larger amount of serum triacylglycerides which, may explain part of these findings. Additionally, some of the other lipid abnormalities noted may be due to a dysregulation in the usual lipolysis and catabolism of lipid stores that occurs during the third trimester of normal pregnancy (Figure 8A).

[0102] The single newborn diagnosis with the greatest number of significant metabolites detectable in cord blood was neonatal jaundice associated with preterm delivery. There were 148 metabolites associated with neonatal jaundice of prematurity in the venous dataset and 89 in the arterial dataset (Wilcoxon rank sum, all FDR < 0.05). The most significant metabolite in both venous batches was dihydrotestosterone glucuronide which is lower in the births associated with neonatal jaundice (logFC = -2.90; -2.33), and the next most significant was estrone (logFC = -2.53; -1.56). Numerous other glucocorticoids and hormones were also lower in these newborns including cortisol (logFC = -2.76; -1.52), hydroxyandrosterone glucuronide (logFC = -1.81 ; -1.15), estriol sulfate glucuronide (logFC = -2.52; -1.64), 16a-hydroxy DHEA 3-sulfate (logFC = -1.26; - 0.95), hydroxycorticosterone (logFC = -1.95; -2.25), estriol glucuronide (logFC = -1.26; - 1.19), pregnendiol disulfate (logFC = -1.10; -0.63), and hydroxyandrosterone glucuronide (logFC = -1 .82; -0.84). These follow a similar pattern to the premature birth; however, the levels and precise molecules vary depending on the disease associated with prematurity, or whether just prematurity itself was measured. There are other molecules lower in these births, like dihydroxy-cholestenoic acid (logFC = -0.66; -0.47). Molecules higher in these births include L-cystathionine (logFC = 1.08; 1.07), taurocholic acid (logFC = 1.00; 0.89), and carnosine (logFC = 1 .06; 0.76). These are molecules also associated with premature births. There were also many significant lipids associated with neonatal jaundice of prematurity, and 42 out of 48 significantly different triglycerides having fewer than 55 carbons. Other significant lipids include 14 CEs, 12 PCs, 4 CERs, 2 DAGs, 1 FFA, 2 HCERs, 1 LCER, and 1 LPC, with most of the nontriglycerides being higher in this cohort. Lipid dysregulation is known in premature newborns (Figure 8A).Metabolic Signatures Associated Post-Term Birth and Preterm Birth are Inverted

[0103] We found a metabolic signature associated with post term pregnancy (birth after 42 weeks), demonstrating the profound and consistent effect of gestational age on the metabolic maturity index. However, there were overall fewer metabolites associated with gestational age greater (23 venous and 8 arterial) than with preterm pregnancy (Wilcoxon rank sum, all FDR < 0.05). Those metabolites associated with post term birth tended to follow the inverse trend of the preterm metabolites. For example, dehydrotestosterone glucuronide (logFC - 1.02; 0.79) was one of the most significant metabolites, just as observed for premature birth, it was found to be generally higher in the post-term cords. Dehydrotestosterone glucuronide is a downstream metabolite of UDP-glucuronyl transferase which is an important component of removing toxins from blood. One possible explanation for our findings is that increased toxin exposures such as alcohol, tobacco, or certain medications result in both premature births as well as increased levels of dehydrotestosterone glucuronide. Other metabolites demonstrating an inverse relationship to preterm birth and therefore an increase relative to term newborns included several sex steroids and corticosteroids including: estriol glucuronide (logFC = 1.27; 0.37), pregnenetriol sulfate (logFC = 0.57; 0.21 ), tetrahydroaldosterone glucuronide (logFC = 0.96; 0.65), estrone (logFC = 0.79; 0.56), and estriol sulfate glucuronide (logFC = 0.77; 0.61 ), while carnosine (logFC = -0.43; -0.32) was decreased. These findings suggest that specific steroid hormone metabolites are key drivers of pregnancy progression and fetal development given the consistent relationship with gestational age observed herein (Figure 8B).Molecules Associated with Gestational Age and Related Diseases of Prematurity

[0104] A consistent, core group of many sex hormones and corticosteroid metabolites were associated with gestational age, preterm newborns, and post term newborns. We created a summary of these associations using the expression of selected metabolites from this category, comparing the odds ratio of disease between the bottom tertile of expression to top tertile of expression (all FDR <0.05, see methods). The number of metabolites associated with different diagnoses is extensive, but several examples areillustrative of the consistent involvement of the core hormone metabolites in both the metabolic maturity index and association with common diseases of prematurity Figure 5C. Several metabolites correlated with the following diagnoses, retinopathy of prematurity stage 0 bilateral, preterm newborn, unspecified weeks of gestation, respiratory distress syndrome of newborn, neonatal jaundice associated with preterm delivery, anemia of prematurity, and post term pregnancy, respectively. For example, tetrahydrodeoxycorticosterone (OR = -2.94, -2.17, -2.10, -2.04, -3.05, NS), dihydrocortisol (OR = -3.51 , -3.40, -2.75, -3.10, -3.08, 1 .72), carnosine (OR = 4.50, 2.94, 1 .87, 2.99, 2.45, -1.87), tetrahydroaldosterone glucuronide (OR = -2.86, -3.26, -1.82, -1.99, -2.94, 1.75), dehydrotestosterone glucuronide (OR = -Inf, -4.12, -4.80, -3.51 , -5.20, 2.45), estriol glucuronide (OR = -4.63, -4.08, -3.80, -2.09, -Inf, 2.10), and cortisol (OR = -3.58, -4.04, - 2.91 , -3.39, -3.04, 1.32). The first six are diseases of prematurity, while the last is post term pregnancy. Meanwhile, all listed metabolites except for tetrahydrodeoxycorticosterone showed an odds ratio in the opposite direction with poster term pregnancy relative to the diseases of prematurity. Overall, there is a strong signature of diseases associated with gestational age and umbilical cord blood metabolomics; however, each disease has its own 820 profile, weighting the metabolites differently (Figure 8C).

[0105] Because there are many metabolites associated with diseases of developmental maturity there are several pathways also associated with these diseases (FDR < 0.05). Among the significant pathways associated with disease of prematurity, the most significant is steroid hormone biosynthesis, which is associated with the same diagnoses: retinopathy of prematurity, stage 0, bilateral, preterm newborn, unspecified weeks of gestation, respiratory distress syndrome of newborn, neonatal jaundice associated with preterm delivery, anemia of prematurity, and post-term pregnancy (ER = batch 1 4.84, batch 2 3.09; 4.33, 3.48; 5.09, 4.47; 5.30, 4.53; 4.70, 3.57; 3.13, NS, respectively). It is known that these molecules are associated with gestational age in mother’ s plasma. Other pathways associated with these diagnoses are beta-alanine metabolism (ER = 3.50, 2.66; 3.48, 3.03; 2.87, 3.76; 3.76, 3.43; 3.41 , 3.00; NS, 2.89, respectively) and histidine metabolism (ER = 2.59, 2.06; 2.55, 2.30; 2.02, 2.82; 3.05, 2.55;2.47, 2.22; NS, 2.02, respectively) showing the importance of amino acid and protein metabolism to these diseases of prematurity.Cord Blood Metabolomics Predict Gestational Age

[0106] Using venous metabolites as inputs and machine learning methods, we sought to compute a newborn metabolic maturity index and used the reported clinical gestational age of the newborns as the gold standard comparator. The resultant model had a high degree of concordance with clinical gestational age, Pearson correlation coefficient (r) of 0.83 (p = 6e-134). In general, the metabolic maturity index was in alignment with the clinically reported gestational age within 2 weeks for 86.9% of the cohort. Dehydrotestosterone glucuronide was the top feature in the model, just as it was the most significant metabolite for prematurity and neonatal jaundice associated with prematurity. Figure 8D shows the predicted gestational age versus actual gestational age.Other Diseases are Associated with Cord Blood Metabolic Indices

[0107] Respiratory distress is associated with many metabolic features. Respiratory distress syndrome of the newborn is associated with 95 arterial and 140 venous metabolites, respiratory failure of a newborn is associated with 20 metabolites in the venous dataset and the use of lung surfactants was associated with 78 metabolites in the venous cord blood (Wilcoxon rank sum, all FDR < 0.05). The profile of respiratory distress looks like diseases of prematurity, with many corticosteroids and sex hormones, lower in the diseased group.

[0108] Premature newborns sometimes require exogenous lung surfactants for survival, which are primarily lipids (40% PCs, 40% dipalmitoylphosphatidylcholines, 10% neutral lipids such as cholesterol). We, therefore, analyzed the correlations between lipids and the administration of lung surfactants (prevalence of use is 2.4% in our cohort), identifying lipids these infants are missing. Indeed, we found several lipids that were significantly associated with lung surfactant use in the newborn period and included six CEs, three CERs, one LPC, two PCs, and 14 TAGs (OR of top fertile metabolite expression compared to bottom fertile, all FDR <0.05. Figure 8E). These lipids alsocorrelate with other medications given to the newborn, such as xanthines, proparacaine, gentamicin, ferrous sulfate, caffeine, and ampicillin. Respiratory failure can be difficult to predict; however, we also show that using the metabolic profiles of the newborns that were given lung surfactants we can predict better than chance which newborns will require this treatment for respiratory failure. Using venous cord blood metabolites as inputs into an XGBoost model (additional detail in methods), we were able to predict future lung surfactant administration from metabolites. The predictive model has an AUC of 0.80. Additionally, within the preterm cohort, the AUC remains relatively high at 0.71. The performance in the preterm cohort provides confidence that we are identifying a true signal of newborns who will in the future require lung surfactant administration, and not just a signal of prematurity. The preterm cohort for newborns given surfactant are significantly different than those who did not receive it, showing a significant difference between preterm newborns who will not ultimately require surfactant, and those who will (p = 0.02).

[0109] There was a metabolic signature in the cord blood metabolomics profile associated with newborns who were diagnosed with neonatal bradycardia (Wilcoxon rank sum, all FDR < 0.05). Epsilon-(gamma-glutamyl)-lysine (logFC = 0.27; 0.38), carnosine (logFC = 0.93; 0.62), and L-cystathionine (logFC = 0.94; 1.10) were higher in cords associated with bradycardia. Many glucocorticoids and sex hormones were lower with bradycardia like cortisol (logFC = -2.55; -0.99), dihydrocortisol (logFC = -2.64; -1.39), androstenediol (3beta,17beta) disulfate (logFC = -0.92; -0.44), dehydrotestosterone glucuronide (logFC = -2.60; -1.06), hydroxycorticosterone (logFC = -1.66; -1.46), and pregnendiol disulfate (logFC = -0.99; -0.46). There were numerous unknown molecules associated with bradycardia, many of which were lower in the births associated with bradycardia, for which we only have the structure. Examples include C21 H28O5 (logFC = -1.69; -1.00), C23H32O7 (logFC = -2.67; -1.07), C20H26O4 (logFC = -1.65; -1.12), C20H28O4 (logFC = -0.80; -1 .82), C21 H24N4O11 (logFC = -1.17; -0.74), C32H32N8O3S (logFC = -2.13; -0.97), and C20H3006S (logFC = -1.06; -0.48). These unknown metabolites, although not always listed, are associated with several medications and diagnoses. Neonatal bradycardia is associated with prematurity. The unknowns leave atantalizing glimpse of what is yet to be discovered in understanding the metabolomics associated with health outcomes.Healthy Neonates have a Distinct Metabolomic Signature

[0110] One question that interested us was whether healthy newborns, those without a disease, had a metabolic signature in the cord blood. We indeed discovered a newborn health signature, in this case denoted by the diagnosis: encounter for routine child health examination without abnormal findings (Wilcoxon rank sum, all FDR < 0.05). These newborns tended to have higher vitamin B5 (LogFC = 0.46; 0.53), as well as higher tryptophan betaine (LogFC = 0.46, 1.00) and piperine (LogFC = 0.78, 0.73), both diet derived metabolites. This group of healthy newborns had lower levels of LysoPE 22:5 (LogFC = -0.62, -0.55) and there are 49 lipid species higher in this group, dominated by 45 triglycerides (i.e. TAG60:12-FA22:6 LogFC = 0.69; 0.73). There are two diglycerides, 16:0 / 22:6 (LogFC = 0.49; 0.47) and 18:1 / 22:6 (LogFC = 0.39; 0.61 ), a phosphatidylcholine 18:0 / 22:6 (LogFC = 0.26; 0.30), and a lactosylceramide 16:0 that are differentially abundant (LogFC = 0.23; 0.14). The lipids follow a bimodal distribution of significance, with one group of lipids being more significant than the rest. These findings indicate that healthy newborns have a different metabolic profile than unhealthy newborns, particularly in their lipid profiles, and demonstrates the utility of this platform as a diagnostic of umbilical cord plasma for health and disease. A healthy metabolic and lipid profile is essential for a healthy newborn (Figure 9A).C-Section Procedures and Vaginal Birth have Different Cord Blood Metabolomic Profiles

[0111] C-section births were metabolically different from vaginal births as determined by Wilcoxon rank sum, all FDR < 0.05. The effect of C-section on neonatal physiology has been a subject of intense interest given the associations with later-life health. The most significant metabolite was lidocaine (LogFC = 3.10; 2.13), a local numbing agent, showing that this anesthetic transmits into the cord blood and is exposed to the fetus. The A-V gradient for lidocaine was also altered by C-sections (LogFC = -0.74; -0.48 for C- section births), indicating that this molecule is getting metabolized only in newbornswhose birth was from a C-section. We also find that C-section is associated with lower cord cortisol (LogFC = -0.81 ; -0.61 ), consistent with the literature. (63) In addition, newborns delivered by C-section have lower levels of corticosterone (LogFC = -0.31 ; - 0.60), dihydrocortisol (LogFC = -1.23; -0.86), lactic acid (LogFC = -0.21 ; -0.22), pyroglutamic acid (LogFC = -0.24; -0.18), carboxyethylleucine (LogFC = -0.51 ; -0.28), and carboxyethylphenylalanine (LogFC = -0.30; -0.44) demonstrating the board effects of C-section on the metabolic profile. There were 163 TAGs (i.e. TAG48:3-FA12:0 LogFC = -0.65, -0.46) with a wide range of carbon structures that were lower in cords of C-sections births, along with one diglyceride, 18:1 / 18:1 (LogFC = -0.22; -0.14) (Figure 9B). These findings could reflect the fact that C-sections are given to a set of mothers who have different risks, but C-sections themselves may be influencing the metabolomics of cord blood. Vaginal births follow the inverse pattern.MethodsSamples

[0112] Plasma was collected from umbilical cords at the Lucile Packard Children’s Hospital Stanford with permission from the Institutional Review Board (IRB #46411 ). All umbilical cords were collected and sent for clinically-indicated testing to Pathology where blood gas analysis was performed following Stanford protocol. Leftover and discarded samples were obtained for research. Arterial and venous blood were extracted from umbilical cords by heparinized syringe and transferred to low binding tubes. Samples were centrifuged at 1300g for 10 minutes. Plasma was aliquoted in cryovials and frozen at -80°C.

[0113] Metabolites and lipids were extracted in 96-well high throughput fashion using a liquid-liquid biphasic separation with cold methyl tert-butyl ether (MTBE), methanol, and water. To begin, 1 mL MTBE was added to 40 pl of plasma and spiked with 40 pl of deuterated lipid internal standards (Sciex, cat# 5040156, lot# LPISTDKIT-103). The samples were agitated at 4°C for 30 minutes. After the addition of 250 pl cold water, samples were vortexed for 1 minute then centrifuged at 3,800 g for 5 minutes at 4°C. The upper organic phase contained the lipids while the lower aqueous phase containedmetabolites with precipitated proteins at the bottom of the tube. For quality control, reference plasma samples (40 pl plasma), as well as controls lacking samples (blanks), were processed in parallel.

[0114] Metabolites: To further precipitate proteins, 500 pl 1 :1 :1 acetone: acetonitrile: methanol spiked with 16 labeled metabolite internal standards was added to 300 pl of the aqueous phase and 200 pl of the organic phase and incubated overnight at -20°C. After centrifugation at 3,800 g for 10 min at 4°C, the metabolic extracts were dried down under a stream of nitrogen gas and resuspended in 100 pl 50 / 50 methanol / water for LC-MS.

[0115] Complex lipids: 700 p of the organic phase was 1139 dried down under a stream of nitrogen and resolubilized in 200 pl of methanol for storage at -20°C until analysis. The day of the analysis, samples were dried down, resuspended in 300 pl of 10 mM ammonium acetate in 90 / 10 methanol / toluene, and centrifuged at 3,800 g for 5 min at 4°C.Data Acquisition

[0116] Metabolite extracts were analyzed using a broad-spectrum untargeted LC-MS platform while complex lipids were quantified using a targeted MS-based approach.

[0117] Untargeted Metabolomics by Liquid Chromatography (LC)-MS'. Metabolic extracts were analyzed four times using HILIC and RPLC separation in both positive and negative ionization modes. Data were acquired on a mass spectrometer for HILIC and a mass spectrometer for RPLC. Both instruments were equipped with a HESI-II probe and operated in full MS scan mode. MS / MS data were acquired on quality control samples (QC) consisting of an equimolar mixture of all samples in the study. HILIC experiments were performed using a ZIC-HILIC column 2.1 x 100 mm, 3.5 p m, 200A (Merck Millipore, Darmstadt, Germany) and mobile phase solvents consisting of 10 mM ammonium acetate in 50 / 50 aceton itrile / water (A) and 10 mM ammonium acetate in 95 / 5 acetonitrile / water (B). RPLC experiments were performed using a Zorbax SBaq column 2.1 x 50 mm, 1.7 pm, 100A and mobile phase solvents consisting of 0.06% acetic acid in water (A) and 0.06% acetic acid in methanol (B). Data quality were ensured by (i) injecting 6 and 12 pool samples to equilibrate the LC-MS system prior to running the sequence for RPLCand HILIC, respectively, (ii) injecting a pooled sample every 10 injections to control for signal deviation with time, and (iii) checking mass accuracy, retention time and peak shape of internal standards in each sample.

[0118] Targeted Lipidomics using the Lipidyzer Platform-. Lipid extracts were analyzed using the Lipidyzer platform that comprises a 5500 QTRAP system equipped with a SelexION differential mobility spectrometry (DMS) interface (Sciex) and a high flow LC- 30AD solvent delivery unit (Shimadzu, Columbia, MD). Briefly, lipid molecular species were identified and quantified using multiple reaction monitoring (MRM) and positive / negative ionization switching. Two acquisition methods were employed covering 13 lipid classes; method 1 had SelexION voltages turned on while method 2 had SelexION voltages turned off. Data quality were ensured by i) tuning the DMS compensation voltages using a set of lipid standards (cat# 5040141 , Sciex) after each cleaning, more than 24 hours of idling or 3 days of consecutive use; by ii) performing a quick system suitability test (QSST) (cat# 5040407, Sciex) before each batch to ensure acceptable limit of detection for each lipid class; and by iii) triplicate injection of lipids extracted from a reference plasma sample throughout the batch.Data Processing

[0119] Metabolomics: Data from each mode were independently analyzed using Progenesis QI software (v2.3). Metabolic features from blanks and those that did not show sufficient linearity upon dilution in QC samples (r<0.6) were discarded. Only metabolic features present in >2 / 3 of the samples were kept for further analysis. Missing values were imputed by drawing from a random distribution of low values in the corresponding sample. Intensity drift was corrected using SERRF. Data quality postnormalization was verified by ensuring clustering of pooled sample replicates on a principal component analysis (PCA) plot. Data from each mode were merged and metabolic features were annotated as follows. Peak annotations were first performed by matching experimental m / z, retention time, and MS / MS spectra to an in-house library of analytical-grade standards. Remaining peaks were identified by matching experimental m / z and fragmentation spectra to publicly available databases including HMDB, MoNA,and MassBank using the R package ‘metID’ (v0.2.0). We used the Metabolomics Standards Initiative (MSI) level of confidence to grade metabolite annotation confidence (level 1 - level 3). Level 1 represents formal identifications where the biological signal matches accurate mass, retention time, and fragmentation spectra of an authentic standard run on the same platform. For level 2 identification, the biological signal matches accurate mass and fragmentation spectra available in one of the public databases listed above. Level 3 represents putative identifications that are the most likely name based on previous knowledge.Data Analytics

[0120] Data from the processing pipeline comes out in two data matrices, one for metabolomics and one for lipidomics. There was a single matrix for each batch. In batch two, 20 samples were reran from batch one for correction. Batch correction was performed in R (4.1.1 ) using the sva::ComBat() (v3.40.0) function. Only metabolites confidently identified (MSI level 3) were used for batch correction. The matrices were then split into arterial or venous umbilical cord samples for analysis. Because veins from the umbilical cords were easier to sample, there were approximately 50% more samples in the venous dataset. A third “subtracted” dataset was5 created using the difference in metabolite values between venous and arterial samples. Data were Iog2 transformed prior to statistical analysis to facilitate statistics and visualizations.

[0121] Correlations were performed in R using superman correlations of Iog2 transformed data and the clinical values. For the sparse correlations where there were between 25 and 200 samples, samples without clinical values were discarded. For the en masse correlations, for clinical values for which we had over 200 samples per a batch k nearest neighbor imputation was performed on the missing samples.

[0122] Associations were determined for all ICD10 diagnoses and medications using the same analysis approach for each. Data were aggregated by the Stanford Research Repository (STARR) from Epic Electronic Health Records. Only diagnoses or medications associated with 25 births combined in the entire cohort were analyzed. Associations were determined using diagnoses and medications from both mother and child. Statisticalanalyses were performed using a Wilcoxon rank sum test on all the births that had a particular diagnosis or medication against those births without that diagnosis or medication. The rank sum test was used to ensure robust statistics even for metabolites that followed a nonparametric distribution. Benjamini Hochberg false discovery rate correction (FDR) was applied to all tests. The results were then compared between batch one and batch two and only metabolites significant (FDR < 0.05) in both batches after FDR are reported. This was done to ensure reproducibility. Figures were made using the ggplot2(v3.4.2) package from R.

[0123] In order to find if high levels of plasma metabolites were associated with health outcomes, every birth was split into tertiles of metabolite abundance. Then an odds ratio (OR) was created between the first and third tertile for the diagnoses discussed. The data were then adjusted for FDR, and only the metabolites significant after correction are reported. For this analysis both batches were grouped together. To account for batch differences, batches were corrected using sva::ComBat with 20 samples from batch 1 being rerun in batch 2 to facilitate correction.Machine Learning

[0124] Machine learning models to predict maternal medication usage were built using XGBoost (v1.6.2) in Python (v3.10.6). For each newborn, the maternal medical record was pulled from Stanford STARR OMOP data. The list of drug concept IDs that we used to identify each drug are included in the appendix. We then used intensity values for each of 874 metabolites that could be consistently identified across batches with high confidence in venous cord blood samples as features to predict whether or not the mother had taken the drug during pregnancy. We consider drugs taken from the time of birth up to 280 days prior to be considered as taken during pregnancy. We used nested cross validation with a 5-fold outer CV and 3-fold inner CV. The optimal model hyperparameters were chosen using the 3-fold inner CV on the 4 folds of training data. These optimal hyperparameters are used to train a model on the 4 folds of training data from the outer CV loop, which are then tested on the 5th test fold. This process is repeated 5 times. The grid search parameters are: learning_rate: 0.01 , 0.1 , n_estimators: 25,50,75,100,max_depth: 2, 3, 4, 5, scale_pos_weight: 1 ,3,5,10. The reported results are the mean and standard deviation of the model performance within each of the 5 test sets from the outer cross validation loop. A final model is trained on all the data using the mean of the 5 sets of the optimal hyperparameters, which is used to compute feature importance. The top features were determined using the gain importance type in XGBoost.

[0125] The metabolite set enrichment analysis was done using MetaboAnalyst (v3.2.0) in R (v4.2.0). Metabolites and drugs were processed in the same way as described above. All p-values were adjusted using the Bonferroni correction. We analyzed 59 metabolite pathways across 6 drugs, so the adjusted significance threshold was 1 .41*10-4 - all findings that are not significant at this threshold were made to be 0.

[0126] For the caffeine metabolite analysis, caffeine metabolites were identified from DrugBank and mapped to our metabolites. All caffeine-related metabolites were extracted from venous cord blood measurements. We then built machine learning models to predict newborn outcomes. The models were built with nested CV with a 3-fold outer loop and 3- fold inner loop, similar to the process described above. The grid search for the hyperparameters was max_depth: 2, 3, 4, 5, 6, 7; n_estimator': 50, 75, 100, 125, 150, 175, 200; learning_rate: 0.1 , 0.05, 0.01 ; scale_pos_weight: 1 , 5, 10. A final model is trained on all the data using the mean of the 3 sets of the optimal hyperparameters, which is used to compute feature importances. The SHAP plots were created using the SHAP package (v0.41.0).

[0127] For our metabolic model of gestational age, we used intensity measurements from the 874 metabolites described above to predict gestational age. The models were built with a nested CV with a 3-fold outer loop and 3-fold inner loop, similar to the process described above. The grid search for the hyperparameters was max_depth: 2, 3, 4, 5, 6, 7, 8; n_estimators: 50, 75, 100, 125, 150, 175, 200 'learning_rate: 0.1 , 0.05, 0.01 ; colsample_bytree: 0.5, 0.75, 1. The reported results are the mean and standard deviation of the model performance within each of the 3 test sets. A final model is trained on all the data using the mean of the 3 sets of the optimal hyperparameters, which is used to compute feature importance. Using the same drug concept ID listed in the appendix, we identified maternal drug use and split the newborns into two populations based onwhether or not their mother had been given a specific drug within 14 days prior to their birth. We analyzed the error of the predictions within these two populations.

[0128] We built a machine learning model to predict lung surfactant use any time after birth using venous metabolites with a nested CV with a 3-fold outer loop and 3-fold inner loop, similar to the process described above. The grid search for the hyperparameters was max_depth: 2, 3, 4; n_estimators: 10, 25, 50, 75, 100, 150, 200; learning_rate: 0.1 , 0.01 ; scale_pos_weight: 1 , 3, 5, 10, 15, 20. For this analysis, we saved the predictions made for each patient when they were in the test set in the outer CV loop. We then computed the AUC from these predictions. This analysis makes it fairly straightforward to compute an AUC within the preterm cohort, which was done by only considering predictions and outcomes for preterm newborns. A final model is trained on all the data using the mean of the 3 sets of the optimal hyperparameters, which is used to compute feature importance.

Claims

WHAT IS CLAIMED IS:1 . A method to assess neonate heath status, comprising: measuring a panel of analytes derived from umbilical cord or placenta blood to yield a set of analyte measurements; and predicting, utilizing a trained computational model, a health statues of a neonate by entering the set of analyte measurements as features into the trained computational model to yield the health status.

2. The method according to claim 1 , wherein the panel of analytes comprises one or more metabolites or one or more lipids.

3. The method according to claim 1 , wherein the health status is good health and the panel of analytes comprises one or more of: vitamin B5, tryptophan betaine, piperine, LysoPE 22:5, TAG60: 12-FA22:6, 16:0 / 22:6, 18:1 / 22:6, 18:0 / 22:6, and 16:0.

4. The method according to claim 1 , wherein the health status is jaundice and the panel of analytes comprises one or more of: hydroxyhippurate, sulfolithocholylglycine, gamma-glutamyl-L-putrescine, and taurine.

5. The method according to claim 1 , wherein the health status is premature birth, atelectasis, or abdominal distension, and the panel of analytes comprises: 3- indolepropionic acid.

6. The method according to claim 1 , wherein the health status is respiratory distress syndrome and the panel of analytes comprises: hydroxyhippurate.

7. The method according to claim 1 , wherein the health status is retinopathy and the panel of analytes comprises: vanillin 4-sulfate.

8. The method according to claim 1 , wherein the health status is hypoxic ischemia and the panel of analytes comprises one or more acylcarnitines.

9. The method according to claim 1 , wherein the health status is of bronchopulmonary dysplasia and the panel of analytes comprises one or more of: xanthine, theophylline, 5-acetylamino-6-amino-3-methyluracil, theobromine, methylxanthine, hypoxanthine, and caffeine.

10. The method according to claim 1 , wherein the health status is an assessment of drug or medication intake during pregany and the panel of analytes comprises metabolites related to steroid hormone biosynthesis and histidine metabolism.11 . The method according to claim 1 , wherein the computational model decision trees, ridge regression, kemalized ridge regression, K-nearest neighbors, logistic regression, LASSO, elastic net, least angle regression (LAR), random forest, neural network, and principal components analysis.

12. The method of claim 1 , further comprising administering a treatment to the individual for the health condition.

Images