Detecting congenital heart defects

Abstract

The present disclosure describes a method of detecting, diagnosing, or predicting congenital heart defects (CHD). The method is primarily a minimally invasive method as it uses a biological sample from a subject to detect methylation changes in nucleic acids of the subject. The method also involves the application of artificial intelligence (AI).

Description

[0001] Cross-references to related applications

[0002] This application claims the benefit of U.S. Provisional Patent Application 62 / 941,357, filed November 27, 2019, the entire contents of which are incorporated herein by reference. Technical Field

[0003] This disclosure relates to methods and systems for using molecular markers to predict and detect congenital heart defects in patients. Background Technology

[0004] Epigenetic changes (including DNA methylation) are known to be associated with cardiac embryogenesis (O′Meara & Lee 2015) and the development of congenital heart defects (Bahado-Singh et al. 2005; Bahado-Singh et al. 2016; Radhakrishna et al. 2016; Radhakrishna et al. 2019). With very few exceptions, cardiac tissue cannot be used for research analysis in living subjects. This is especially true for developing fetuses, and this reality hinders the assessment of epigenetic changes that are the cause or associated with the development of congenital heart defects (CHD). Therefore, there is significant interest in molecular markers in developing tissues (such as blood leukocytes) that also reflect epigenetic changes and are biologically linked to or associated with those in isolated organs such as the developing heart. Numerous studies (Bahado-Singh et al. 2005; Bahado-Singh et al. 2016; Radhakrishna et al. 2016; Radhakrishna et al. 2019) have demonstrated that the aforementioned goals may be achievable. These studies have used methylation of neonatal leukocyte DNA obtained through heel stick pricking to elucidate the mechanisms of various common nonsyndromic congenital heart defects (CHDs). Furthermore, using conventional statistical analysis and artificial intelligence (AI) methods, leukocyte DNA methylation accurately predicted different types of nonsyndromic congenital heart defects (CHDs) (Bahado-Singh et al. 2019b).

[0005] Extensive clinical and laboratory publications (Burton & Jauniaux 2018; Maslen 2018) demonstrate the crucial role of the placenta and its vascular distribution in cardiac embryogenesis and the development of CHD. Indeed, conditions that can affect placental morphology and vascular distribution (e.g., maternal hypertension) have shown in meta-analyses to significantly increase the risk of CHD in offspring of such pregnancies (Ramakrishnan et al. 2015; van Gelder et al. 2015). The placenta is an organ that can be extensively used for postnatal analysis.

[0006] In DNA methylation, a single carbon atom, or so-called "methyl group," is transferred to and covalently bound to position #5 of the cytosine nucleotide ring in a cytosine-guanine ('C-G' or 'CpG') dinucleotide. This process converts the cytosine base to 5-methylcytosine (5mC). Typically, DNA methylation, especially when it occurs in the promoter region of genes with numerous CpG repeats, leads to repression of gene transcription or gene silencing. 5mC can be further chemically modified to 5-hydroxymethylcytosine (5hmC). This hydroxymethylation has been found to occur through the action of a group of enzymes called 10-11 translocation (TET) proteins. These are a group of three dioxygenases that catalyze the conversion of 5mC to 5hmC (Tahiliani et al. 2009).

[0007] Since 5hmC is produced by the conversion of 5mC, 5mC is considered a mechanism for eliminating the former. The accumulation of 5hmC is associated with the regulation of gene transcription. Unlike 5mC, 5hmC has a much lower affinity for gene repressor proteins (e.g., methyl-CpG-binding proteins such as MBD1, MBD2, and MBD4) (Jin SG, Kadam S, Pfeifer GP. Examination of the specificity of DNA methylation profiling techniques towards 5-methylcytosine and 5-hydroxymethylcytosine. Nucleic Acid Res 2010;38:e125). Therefore, in addition to measuring 5mC concentration, contemporaneous information about 5hmC will provide more detailed epigenetic information and more precise gene transcription information. 5hmC levels have been found to be enhanced in the gene bodies of transcriptionally active genes (Nester et al. 2012), thus having an opposite effect on gene expression compared to 5mC.

[0008] Secondly, the concentration of 5hmC appears to be an important mechanism for tissue differentiation and a marker of tissue type (Nester et al. 2012). A major limitation of fetal cell-free (cfF) DNA research is that “fetal” DNA (more precisely, placental DNA as fetal organs) is heavily contaminated by maternal DNA leaking from hemolyzed leukocytes. Therefore, cfF DNA in maternal circulation accounts for approximately 10–20% of cell-free (cf) DNA in pregnancy (Rafi et al. 2017). Given the low 5hmC levels in maternal leukocytes, measuring 5hmC levels in fetal cfDNA should improve the specificity for identifying fetal and / or placental cfDNA in maternal circulation. Attached Figure Description

[0009] Figure 1A and 1B A heatmap is shown, generated using individual β values ​​of methylation markers that showed significant differences between cases and controls. Hierarchical cluster analysis revealed a clear separation between cases and controls based on differential CpG methylation. CpG marker IDs are provided in the right column.

[0010] Figure 2 The study revealed that differentially methylated gene networks play an important role in cardiac embryonic development and the formation of congenital heart defects. Summary of the Invention

[0011] The present invention is provided to introduce some concepts in a simplified form, which will be further described in the detailed description below. The summary is not intended to identify all key or essential features of the claimed subject matter, nor is it intended to be used alone to help determine the scope of the claimed subject matter.

[0012] Precision medicine is based on the integration of artificial intelligence (AI), big data, omics technologies, and blood tests to elucidate the pathogenesis and accurate detection of complex diseases (Mesko B, 2017). This study combines epigenomic analysis of cell-free fetal DNA (cfF DNA) with AI technology to inquire about and detect fetal non-syndromic CHD. In this prospective study, cfF DNA was extracted from maternal blood, and whole-genome DNA methylation analysis was performed using the Illumina Infinium MethylationEPIC BeadChip array. Six different AI platforms, including deep learning (DL) (a state-of-the-art AI approach), were used for CHD detection. Ingenuity pathway analysis based on significantly differentially methylated sites was used to investigate the molecular basis of CHD.

[0013] A total of 12 isolated cases of CHD and 26 controls were included. AI accurately detected CHD. Using RF, a combination of five CpG markers had an AUC (95% CI) of 0.98 (0.83–1), with 97% sensitivity and 93% specificity; followed by an SVM model, with an AUC (95% CI) of 0.97 (0.87–1), with 98% sensitivity and 94% specificity, for CHD detection. Logistic regression with cross-validation of CpG markers and the fetal history of CHD had an AUC (95% CI) of 0.98 (0.98–0.97), with 100% sensitivity and 85% specificity for CHD. Epigenetic dysregulations of genes and gene pathways involved in cardiac development and abnormal cardiac progression were found in non-syndromic CHD. This provides biological confirmation or plausibility that the findings were not random.

[0014] The method described in this paper accurately detects fetal CHD using AI analysis of cfF DNA in maternal blood. The applicant's data further demonstrate the role of epigenetic dysfunction in the development of non-syndrome CHD. The applicant's findings represent a further advancement in precision medicine for fetal cardiovascular health. Detailed Implementation

[0015] In the United States, birth defects (i.e., abnormalities that develop during fetal life and are present at birth) are the leading cause of infant mortality, defined as death within the first year of life. The frequency of CHD is 8–9 cases per 1,000 live births. CHD is the most common group of severe birth defects and is the most expensive in terms of hospitalization. Up to 25% of severe neonatal CHD cases go undiagnosed before discharge.

[0016] The development of the heart during embryonic and fetal life requires the coordination and arrangement of a large number of different genes. A relatively small percentage of CHD cases are known to be associated with gene mutations, which are changes in the normal sequence of basic structural units (“nucleotides”) arranged in the DNA of genes. Such mutations lead to dysfunction or nonfunction of genes that are important for normal heart development (i.e., changes in the amount or production of abnormal types of proteins).

[0017] Over the past sixty years, important mechanisms controlling gene function, known as "epigenetics," have been discovered and extensively studied. The term "epigenetics" describes the interactions between genes and the environment. These interactions do not lead to changes in the genome sequence itself (no changes in nucleotide sequence), but alter gene expression, which is still the cause of phenotypic changes. Epigenetics is defined as heritable (i.e., passed on to offspring) changes in cellular gene expression that are not primarily due to mutations or alterations in the nucleotide (adenine, thiamine, guanine, and cytosine) sequences of genes. Epigenetics involves the reversible regulation of gene expression through several potential mechanisms. One of the most widely studied mechanisms is DNA methylation. Other mechanisms include changes in the three-dimensional structure of DNA, histone modifications, and microRNA repressive activity. Epigenetic mechanisms are known to be broadly interconnected.

[0018] Cytosine is one of the four structural units, or nucleotides, that make up DNA. The chemical structure of cytosine is a pyrimidine ring. Other nucleotides or structural units found in DNA besides cytosine include thiamine, adenine, and guanosine.

[0019] The term methylation refers to the enzymatic addition of a "methyl" or a single carbon atom to the #5 position of the pyrimidine ring of cytosine, resulting in the conversion of cytosine to 5-methyl-cytosine. Cytosine methylation, as described above, is accomplished by the action of a family of enzymes called DNA methyltransferases (DNMTs). After formation, 5-methyl-cytosine is readily mutated or chemically transformed from the original cytosine to form thymine. 5-methyl-cytosine constitutes approximately 1% of all nucleotide bases in a normal genome.

[0020] The term hypermethylation refers to an increase in the frequency or percentage of methylation at a specific cytosine site when a sample from a target individual or group is compared with a normal or control group.

[0021] Cytosine typically pairs with guanosine (another nucleotide in a linear sequence along a single DNA strand) to form CpG pairs. “CpG” refers to a cytosine-phosphate-guanosine chemical bond, where the phosphate binds the two nucleotides together. In mammals, approximately 70% to 80% of the cytosine in these CpG pairs is methylated (Chatterjee R, Vinson C. Biochemica et Biophisica Acta 2012; 1819: 763-70). The term “CpG island” refers to a region in the genome with a high concentration of CG dinucleotide pairs, or CpG sites. CpG islands are frequently found near genes in mammalian DNA. The DNA length occupied by a CpG island is typically 300–3000 base pairs. CG clusters reside on the same DNA single strand. CpG islands are defined by various criteria, including: i) a length of repeating CG dinucleotide pairs occupying at least 200 bp of DNA; and ii) a CG content of at least 50%, and an observed / expected CpG ratio greater than 60%. In humans, approximately 70% of gene promoter regions have high CG content. CG dinucleotide pairs may exist elsewhere in genes or even outside of genes, and their association with specific genes is unknown.

[0022] Approximately 40% of the promoter regions (the gene regions that control their transcription or activation) of mammalian genes contain associated CpG islands, and three-quarters of these promoter regions have high CpG concentrations. Generally, cytosine nucleotides are methylated at most CpG sites scattered throughout the DNA. Conversely, cytosine is unmethylated at CpG sites located within CpG islands in gene promoter regions, indicating the role of cytosine methylation status within CpG islands in gene transcriptional activity.

[0023] Methylation of cytosine, which is associated with or located within genes, is generally associated with the repression of gene transcription. However, in some genes, increased methylation has the opposite effect and leads to gene activation or increased transcription. One potential mechanism to explain the latter phenomenon is that cytosine methylation may potentially inhibit the binding of gene repressive elements, thereby releasing the gene from repression. Epigenetic modifications, including DNA methylation, are mechanisms by which cells containing the same DNA and genes undergo the activation of different genes, leading to differentiation into unique tissues (e.g., the heart or intestine).

[0024] This disclosure describes the use of epigenomics and artificial intelligence analysis techniques to accurately diagnose or predict CHD (including CHD in prenatal and / or pediatric subjects) based on the detection of cytosine methylation in the nucleic acids of test subjects. Methylation analysis was performed using an Illumina Infinium array with methylation markers exceeding 850 kJ, according to the manufacturer's instructions. Methylation levels at CpG sites throughout the genome were examined in 12 CHD cases and compared with 26 unaffected healthy matched controls. Pathway analysis was performed using Ingenuity pathway analysis to elucidate the mechanisms of the disease. Furthermore, the diagnostic accuracy of the epigenomic markers used for CHD detection was determined. The area under the curve (AUC) of receiver operating characteristics, along with 95% CL and FDRp values, was calculated for CHD detection.

[0025] Several different artificial intelligence (AI) technologies, including the latest form of AI, deep learning (DL), are used to predict CHD using i) epigenetic markers (i.e., DNA methylation markers) and ii) clinical and demographic markers.

[0026] In its implementation, this disclosure describes a method for diagnosing CHD based on the methylation frequency or percentage of cytosine nucleotides at various identified sites in a nucleic acid sample. In its implementation, the nucleic acid sample can be obtained from a biological sample of a patient in need. The method includes: obtaining a biological sample from the patient; extracting nucleic acids from the sample; measuring the sample to determine the percentage of cytosine methylation at sites throughout the genome; comparing the patient's cytosine methylation level with a control; and calculating the individual's risk of being diagnosed with CHD based on the cytosine methylation levels at different sites throughout the genome. The control may be one or more characterized or known cases and / or one or more characterized or known groups.

[0027] The methods described herein include obtaining nucleic acids from a biological sample of a subject. Subjects are individuals or patients who need (or require) a diagnosis or are experiencing symptoms of CHD. Subjects may also be undergoing routine screening. Examples of subjects include, for example, adults, pediatric patients, embryos, or fetuses. “Embryo” refers to a patient from fertilization to the end of the eighth week of pregnancy. “Fetus” refers to a patient after the eighth week of pregnancy.

[0028] In the implementation scheme, patients may be adults, and controls may be a fully characterized normal (healthy) population and / or a fully characterized CHD patient population.

[0029] In the implementation plan, the patient can be a pediatric patient. Pediatric patients can be under approximately 19 years old, approximately 15 to 19 years old, under approximately 15 years old, approximately 10 to 15 years old, under 10 years old (childhood), approximately 5 to 10 years old, under approximately 4 years old, approximately 1 to 4 years old, under one year old (infant), or 28 days or less after birth (newborn or neonatal period). The patient can be an intrauterine patient, such as an embryo or fetus. When the patient is an embryo or fetus, DNA can be obtained from the biological sample of the mother, a pregnant woman carrying the embryo or fetus. The biological sample can be obtained from a pregnant woman in early, mid, or late pregnancy.

[0030] Controls for pediatric patients could be a well-characterized group of normal (healthy) children under approximately 19 years of age and / or a well-characterized group of pediatric patients with CHD. Similarly, controls for intrauterine patients could be a well-characterized group of normal (healthy) intrauterine patients and / or a well-characterized group of intrauterine patients with CHD.

[0031] A well-characterized group of normal population (including adults, children, and intrauterine patients) or CHD patients may include one or more normal populations or CHD patients, or may include a group of normal populations or CHD patients.

[0032] Biological samples can be bodily fluids, such as blood, plasma, serum, urine, saliva, sputum, sweat, respiratory condensate, tears, genital secretions including cervical secretions, amniotic fluid, and umbilical cord blood obtained at birth. Biological samples can be cervical swabs, skin, hair, hair follicles / roots, and mucous membranes (cheek scrapings or tongue scrapings) used for cell-free nucleic acids or exfoliated trophoblast cells. Biological samples can also include any in vivo tissues of the patient, such as any tissue samples obtained from the patient, including placental tissue from the neonatal period or fetal life. Placental tissue from the embryo or fetus can be obtained through placental biopsy or chorionic villus sampling (CVS). In embodiments, biological samples may also include samples from CVS. In embodiments, biological samples from the mother may include maternal blood, placenta, amniotic fluid, other bodily fluids during pregnancy, or other maternal bodily fluids.

[0033] Cells and nucleic acids from any biological sample containing DNA can be used in the methods described herein for the diagnosis and prediction of CHD. Samples used for testing can be obtained from living or dead tissues, as well as archaeological samples containing cells or tissues.

[0034] The nucleic acids used in the methods described herein can be obtained from cells. In embodiments, the nucleic acids include fetal nucleic acids obtained directly from the fetus or embryo (e.g., obtained from the placenta or amniotic fluid via amniocentesis), or fetal nucleic acids obtained from maternal bodily fluids or placental tissue, circulating fetal cells harvested from maternal circulation, exfoliated placental cells, or cfF DNA in maternal circulation. In embodiments, the nucleic acids are obtained from amniotic fluid, fetal blood, or umbilical cord blood obtained during the fetus's life or at birth. In embodiments, the nucleic acids are DNA or RNA.

[0035] Since all cells (with a few exceptions, namely, mature red blood cells and mature platelets) contain a nucleus and therefore DNA, the methods described herein can be used to screen for CHD using DNA from any cell type other than the two mentioned above. Therefore, any biological or tissue sample, including cells containing nucleic acids (e.g., DNA), can be used in the methods described herein. Furthermore, cell-free DNA released from cells that have been damaged and can be recovered from bodily fluids can be used for this screening.

[0036] Cell-free DNA (cf DNA): Nucleic acids can be DNA that exists in the form of cf DNA. cf DNA refers to DNA released from cells due to natural cell death / renewal or due to disease processes. cf DNA is released into circulation and rapidly breaks down into DNA fragments, and can eventually end up in other bodily fluids, such as urine. Techniques for harvesting cf DNA from blood and other bodily fluids are well known in the art (Li Y et al. Size separation of circulatory DNA in maternal plasma permits ready detection of fetal DNA polymorphisms. ClinChem 2004; 50: 1002-1011; Zimmerman B et al. Noninvasive prenatal aneuploidytesting of chromosomes 13, 18, 21, X, and Y, using targeted sequencing of polymorphic loci. Prenat Diagn 2012; 32: 1233-41).

[0037] The cell-free DNA isolation technique used coats and stabilizes maternal leukocytes, preventing the breakdown, release, and contamination of leukocyte DNA from other maternal sources. To further enhance and target the analysis of placental / fetal cf DNA, rather than cf DNA derived from other maternal tissues, targeting techniques capable of identifying the tissue origin of cf DNA can be employed. One example is the study by Lehmann-Werman R et al. (Lehmann-Werman et al. 2016). Existing methylome databases were used to identify tissue-specific methylation patterns in cf DNA. Cell-free DNA was obtained from a blood donor and compared to methylome databases based on methylation, thus enabling the identification of circulating tissue origin for cf DNA fragments, such as pancreatic β cells and oligodendrocytes from the brain.

[0038] 5hmC levels are tissue-specific and extend beyond their reported correlation with gene expression levels. Beyond the gene expression information provided by 5hmC levels in cf DNA, other information regarding the tissue origin of cf DNA opens up the ability to obtain DNA and epigenetic data from different organs based on blood tests. Studies (Nester et al. 2012) have shown that 5hmC levels vary significantly between tissues. 5hmC levels are very low in blood cell DNA, relatively high in placental tissue, and even higher in the brain. Therefore, measuring 5hmC in CF fetal DNA (placental origin) would greatly address the problem of maternal leukocyte cfF DNA contamination, which can be problematic when only 5mC levels are measured in cfF DNA. This is because 5hmC levels in leukocytes are so low compared to cfF DNA derived from the placenta (the target in this CHD proposal). Therefore, overall, the use of 5mC and 5hmC measurements for placental tissue-targeted epigenomic analysis of cf DNA can be used to detect and provide further enhanced mechanistic information about CHD and other pregnancy conditions in which the placenta is affected.

[0039] Methylation assays: Several quantitative methylation assays are available. These methods include COBRA assays using methylation-sensitive restriction endonucleases. TM Detection is performed using gel electrophoresis and label-based hybridization probes. Another available technique is methylation-specific PCR (MSP) for amplifying target DNA segments. This is performed using methylation-sensitive probes after cytosine transformation with sodium bisulfite. (MethyLight) TM (Based on quantitative methylation assay) Fluorescence-based PCR is used. Another method used is quantitative methylation (QM). TM The assay combines PCR amplification with fluorescent probes designed to bind to putative methylation sites. Ms-SNuPETM This is a quantitative technique for determining differences in methylation levels at CpG sites. Like other techniques, it first involves bisulfite treatment, which converts unmethylated cytosine to uracil, while methylated cytosine remains unaffected. PCR primers specific to bisulfite-converted DNA are used to amplify the target sequence. The amplified PCR products are isolated and used to quantify the methylation status of the target CpG site. The preferred method for measuring cytosine methylation is the Illumina method.

[0040] Illumina Methods: For DNA methylation assays, Illumina... The Human Methylation 450 Beadchip or Illumina Infinium Methylation EPIC BeadChip assay can be used for quantitative methylation profiling. In short, nucleic acids, such as genomic DNA, are obtained. Nucleic acids are isolated using commercially available kits employing techniques widely known in the industry. Proteinase K is used to remove proteins and other contaminants from the nucleic acids. Nucleic acids are then removed from solution using available methods (e.g., organic extraction, salting out, or binding DNA to a solid support).

[0041] Methylation Analysis – Illumina’s Infinium Human Methylation 450 Bead Chip system or Illumina Infinium MethylationEpic BeadChip array can be used for genome-wide methylation analysis. Using the standard Infinium assay protocol, nucleic acids (e.g., DNA) (500 ng) are subjected to bisulfite conversion using the EZDNA Methylation Gold kit or the EZ-96 Methylation kit (Zymo Research) to deamination unmethylated cytosine for uracil. The DNA is enzymatically fragmented and hybridized to the Illumina BeadChip. The BeadChip contains site-specific oligomers and is paired, one specific to the methylated cytosine site and the other specific to the unmethylated site. Single-base extension is performed to incorporate biotin-labeled ddNTPs. After fluorescent staining and washing, the BeadChip is scanned, and the methylation status of each site is determined using BeadStudio software (Illumina). Experimental quality was assessed using a control dashboard with sample-dependent and sample-independent controls for target removal, staining, hybridization, extension, bisulfite conversion, specificity, negative controls, and non-polymorphic controls. Methylation status is the ratio of the methylated probe signal to the sum of methylated and unmethylated probes. The resulting ratio indicates whether the site is unmethylated (0) or fully methylated. Differentially methylated sites were identified using the Illumina Custom Model, and filtered based on p-values ​​using 0.05 as the cutoff value.

[0042] bisulfite conversion: such as in As described in the guidelines for methylation assays, nucleic acids are treated with sodium bisulfite, which converts unmethylated cytosine to uracil while leaving methylated cytosine unchanged. The bisulfite-converted nucleic acids are then denatured and neutralized. The denatured nucleic acids are then amplified. Bisulfite-based analysis (currently used to distinguish between methylated and unmethylated cytosine) cannot differentiate between 5mC and 5hmC. New technologies include, but are not limited to: thin-layer chromatography (Kriaucionis et al. 2009), chemical labeling of 5hmC (Song et al. 2011), immunoprecipitation (Nester et al., 2012), and recently commercially available 5hmC whole-exome and even whole-genome sequencing technologies that can provide detailed information about epigenetic changes in cfF DNA.

[0043] In the implementation scheme, a genome-wide (using genomic DNA) methylation study was performed using Illumina Infinium assays, which confirmed significant differences in the methylation frequency (level or percentage) of specific cytosine nucleotides associated with specific genes in the CHD group compared to the normal group. The differences in cytosine methylation levels were highly significant and sufficient to accurately distinguish between the CHD and normal groups. Therefore, the method described herein can be used as a test for screening CHD cases in a mixed population of patients with and without CHD.

[0044] The whole-genome application process increases the amount of DNA by thousands of times. The next step involves enzymatically fragmenting the DNA. The fragmented DNA is then precipitated with isopropanol and separated by centrifugation. The separated DNA is then resuspended in hybridization buffer. The fragmented DNA is then hybridized with magnetic beads covalently confined to 50-mer nucleotide segments at sites specific to the target cytosine nucleotide in the genome. There are over 500,000 types of magnetic beads specifically designed to anneal to the site where a specific cytosine nucleotide is located. The magnetic beads are bound to a silicon-based array. Two types of magnetic beads are designed for each site: one type represents a probe designed to match the methylation site where the cytosine nucleotide will remain unchanged; the other type corresponds to the initially unmethylated cytosine, which is converted to thiamine nucleotides after bisulfite treatment. Unhybridized (unannealed to the magnetic beads) DNA is washed away, leaving only the DNA segment bound to the appropriate magnetic bead and containing the target cytosine. After annealing with the corresponding patient DNA sequence, the oligomers bound to the magnetic beads then use the "protrusion" outside the target cytosine in the patient DNA sequence as an extension template to perform single-base extension with fluorescently labeled nucleotides.

[0045] If the target cytosine is unmethylated, it will perfectly match the unmethylated or "U" magnetic bead probe. This allows for single-base extension with a fluorescently labeled nucleotide probe and generates a fluorescent signal from the magnetic bead probe that can be read automatically. If the cytosine is methylated, a single-base mismatch will occur with the "U" magnetic bead probe oligomer. However, no further nucleotide extension occurs on the bead oligomer, preventing the incorporation of the fluorescently labeled nucleotide onto the bead. This will result in a low fluorescence signal from the "U" magnetic bead. The opposite will occur with "M" or methylated magnetic bead probes.

[0046] A laser is used to stimulate a fluorophore that binds to a single base for sequence extension. The methylation level at each cytosine site is determined by comparing the fluorescence intensity from the methylated magnetic beads to that from the unmethylated beads. The cytosine methylation level is denoted as "β," which is the ratio of the methylated bead probe signal to the total signal intensity at that cytosine site. These techniques for determining cytosine methylation have been previously described and are widely available for commercial use.

[0047] This disclosure describes the use of commercially available methylation technologies to cover up to 99% of a reference sequence gene across the entire genome, involving approximately 16,000 genes and 450,000 cytosine nucleotides, down to the single nucleotide level (Infinium Human Methylation 450 Beach Chip Kit or Infinium Methylation Epic BeadChip). The frequency of cytosine methylation at a single nucleotide in a cohort of CHD cases was used to estimate the risk or probability of being diagnosed with CHD, compared to a control group. The cytosine nucleotides analyzed using this technology include cytosine within CpG islands and cytosine at greater distances outside CpG islands, i.e., located on the “CpG island shore” and “CpG shelf,” and even further away from the islands in the so-called “CpG sea.”

[0048] The cytosines evaluated as described herein include, but are not limited to, cytosines located in CpG islands within the promoter regions of genes. Other regions targeted and measured include the so-called "shores" of CpG islands, located up to 2000 base pairs from the island, and the "shelves," which are the names of the DNA regions on either side of the shore. Differential cytosine methylation in regions further from the CpG islands (the so-called "sea") was analyzed. Extragenic cytosine sites located outside known genes (however, which may potentially maintain remote control of unspecified genes) were also detected with moderate, good, and excellent accuracy, as shown.

[0049] Specific cytosine nucleotide identification: Reliable identification of specific cytosine sites distributed throughout the genome has been described in detail in the following literature (Illumnia): "CpG Loci Identification. A guide to Illumina's method for unambiguous CpG loci identification and tracking for the..." and lnfinium TMAssays for Methylation. A brief overview is as follows. Illumina has developed a unique CpG site identifier that specifies the cytosine site based on the actual nucleotide position or the preceding and following sequences. It uses a similar strategy to that used in NCBI's Re SNP IPS (rs#) and is based on the sequences flanking the target cytosine. Therefore, a unique CpG site cluster ID number is assigned to each cytosine being evaluated. This system is reportedly consistent and unaffected by changes in public databases and genome assembly. The 60-base flanking sequences at 5′ and 3′ on either side of the CG site (i.e., a total of 122 bases) are used to identify the site. Therefore, a unique "CpG cluster number" or CG# is assigned to the 122 bp sequence containing the target CpG. The CG# is based on the human genome's Build... 37 (NCBI37). Therefore, there is only a risk that sites are assigned the same number and located in more than one position in the genome if the 122bp in the CpG cluster are identical. Based on this unique ID system, three separate criteria are used to track individual CpG sites: chromosome number, genomic coordinates, and genome construction. The smaller of the two coordinates "C" or "G" in the CpG is used for unique CG site identification. CG sites are also designated relative to the first 'definitive' nucleotide pair containing 'A' (adenine) through 'T' (thiamine). If one of these nucleotides is the 5' of the CG, the arrangement is designated TOP, and if such a nucleotide is the 3', it is designated BOT.

[0050] Furthermore, the forward or reverse DNA strand is indicated as the location of the cytosine being evaluated. It is assumed that the methylation state of cytosine bases within a specific chromosomal region is synchronous.

[0051] Cytosine methylation is used to diagnose CHD using ROC curves: To determine the accuracy of methylation levels at specific cytosine sites for CHD prediction, different threshold levels of methylation at these sites (e.g., 10%, ≥20%, ≥30%, ≥40%, etc.) are used to calculate the sensitivity and specificity for CHD diagnosis. Thus, for example, using ≥10% methylation at a specific cg site, cases with methylation levels above this threshold will be considered to have a positive test, and cases below this threshold will be interpreted as a negative methylation test. In this embodiment, the percentage of CHD cases with a positive test (i.e., 10% methylation at the specific cytosine site) will equal the test sensitivity. The percentage of normal (non-CHD) cases with cytosine methylation levels <10% at this site will be considered the test specificity. The false positive rate is defined here as the number of normal cases with (incorrect) abnormal test results, and the sensitivity is defined as the number of CHD cases with (correct) abnormal test results, for example, a methylation level of 10% at the specific cg location. A series of threshold methylation values, such as ≥1 / 10, ≥1 / 20, ≥1 / 30, etc., were assessed and used to generate a series of paired sensitivity and false positive values ​​for each site. Receiver operating characteristic (ROC) curves were generated, plotting the data points with the Y-axis representing the sensitivity value and the X-axis representing the false positive rate. This method can be used to generate ROC curves for each individual cytosine site, showing significant methylation differences between the case and CHD groups. In this case, the computer program ROCR package - version 3.4 was used. https: / / CRAN.R-project.crg / The package = ROCR) is used to generate the ROC curve underside product.

[0052] The ROC curve is a plot of sensitivity (defined in this setting as the percentage of CHD cases with a positive test or abnormal cytosine methylation level at a specific cytosine site on the Y-axis) and false positive rate (1-specificity, or 100%-specificity when the latter is expressed as a percentage) (i.e., the number of normal (non-CHD) cases with abnormal cytosine methylation at the same site on the X-axis). Specificity is defined as the percentage of normal (non-CHD) cases with normal methylation levels or a negative test at the target site. The false positive rate is the percentage of normal individuals who are incorrectly found to have a positive test (i.e., abnormal methylation level); it can be calculated as 100-specificity (%), or expressed in decimal format [1-specificity (represented as a decimal point)].

[0053] The area under the ROC curve (AUC) represents the accuracy of a test in identifying normal cases from abnormal cases. AUC is the area under the ROC curve from the intersection of the X and Y axes to the diagonal (tilt angle 45°). The larger the area under the ROC curve, the higher the accuracy of the test in predicting the target condition. An area under the ROC curve of 1.0 represents a perfect test, which is positive (abnormal) in all cases with the disease and negative in all normal cases (without the disease). Methylation assays are used to determine the methylation level at specific cytosine sites in the genome, and many are commercially available. In this specific case, the method can be used to distinguish methylation levels in affected cases (CHD) compared to unaffected controls.

[0054] Logistic regression analysis can be used to calculate the sensitivity and specificity of CHD prediction based on cytosine site methylation.

[0055] Standard statistical tests can be performed, using a p-value to represent the probability that a difference can be observed between cytosine methylation at a given site in CHD and control DNA samples. More rigorous tests of statistical significance were also performed using the false discovery rate (FDR) for multiple comparisons. When performing multiple hypothesis testing using multiple comparisons, the FDR gives the probability that a positive result is accidental.

[0056] Statistical Analysis: This disclosure describes methods for predicting, diagnosing, detecting, and / or calculating the risk of a subject being diagnosed with CHD or even a specific type of CHD. This calculation can be based on logistic regression analysis, resulting in the identification of significant independent predictors among a number of possible predictors known to be associated with an increased risk of CHD or being diagnosed with CHD, such as methylation sites. Cytosine methylation levels at different sites can be used alone or in combination with other known CHD risk predictors, such as prenatal exposure to toxins – “yes” or “no” (e.g., alcohol or maternal smoking, maternal diabetes, family history combined with methylation levels at single or multiple sites), which are known to be associated with an increased risk of CHD as described in this application. For example, the probability of an individual being affected can be derived from a logistic regression-based probabilistic equation:

[0057] P CHD =1 / 1+e -(B1x1+B2x2+B3x3...Bnxn)

[0058] Here, 'x' refers to the magnitude or number of a specific predictor (e.g., methylation level at a specific site), and 'β' or β-coefficient refers to the magnitude of the change in the probability of an outcome (e.g., CHD) for each unit change in the level of a specific predictor (x). β values ​​are derived from the results of logistic regression analysis. These β values ​​will be obtained from multivariate logistic regression analysis of a large number of affected and unaffected individuals. In this case, the values ​​of x1, x2, x3, etc., representing the percentage of methylation at different pyrimidine sites, will be derived from the individuals being tested, while the β values ​​will be derived from logistic regression analysis of the aforementioned large reference population of affected (CHD) and unaffected cases. Based on these values, the probability of an individual having a certain type of CHD can be quantitatively estimated. Probability thresholds are used to define individuals at high risk (e.g., a probability ≥1 / 100 CHD can be used to define high-risk individuals triggering further evaluation, such as one or more of the following: echocardiography, pulse oximetry at birth, etc.), while individuals with a risk <1 / 100 will not require further follow-up. Among other factors, the thresholds used will be based on diagnostic sensitivity (the number of correctly identified CHD cases), specificity (the number of non-CHD cases correctly identified as normal), the risks and costs of echocardiography, and the interventions associated with designating an individual as "high-risk" for CHD and such factors. Logistic regression analysis is well-known as a method for estimating an individual's risk of having the condition in disease screening. (Royston P, Thompson SG. Model-based screening by risk with application in Down's syndrome. Stat Med 1992; 11: 257-68.)

[0059] Individual risk of CHD can also be calculated using different combinations of sites by distinguishing the percentage of methylation at individual cytosine sites (reported as the β coefficient) or by using methods based on overlapping Gaussian or multivariate Gaussian distributions (Wald NJ, Cuckle HS, Deusem JW et al (1988) Maternal serum screening for downsyndrome in early pregnancy. BMJ297, 883-887.), where the variable would be the so-called methylation level / methylation percentage at a specific (or multiple) site. Alternatively, if the methylation percentage or β coefficient is not normally distributed (i.e., non-Gaussian), a normal Gaussian distribution will be achieved by logarithmic transformation of these percentages if necessary.

[0060] As an example, two Gaussian distribution curves were derived for methylation at specific sites in the CHD group and the normal population. The mean, standard deviation, and overlap between the two curves were then calculated. The ratio of the heights of the distribution curves at a given methylation level gives the likelihood ratio, or factor, of the increased (or decreased) risk of having CHD at a specific methylation level at a given site. The likelihood ratio (LR) value can be multiplied by the background risk of CHD in the general population to give an individual's CHD risk based on the methylation level at the selected CG site. Information on the background population risk of CHD in the neonatal population is available from several sources (one such example is Hoffman JL et al Am Heart J 2004; 147: 425-439). Similar information can be used for prenatal and postnatal life.

[0061] Artificial Intelligence (AI): One or more AI algorithms can be used in combination with the methods described herein to improve the accuracy of predicting and / or diagnosing CHD. Representative examples of AI algorithms include Random Forest (RF), Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), Microarray Analysis Prediction (PAM), Generalized Linear Model (GLM), and Deep Learning (DL).

[0062] Random Forest (RF) is a supervised classification algorithm used for regression, classification, and other tasks. During the training phase, multiple decision tree prediction models are randomly generated, and the output consists of the pattern of the classes and the average prediction of each tree. There is a direct relationship between the number of trees in the forest and the results that can be obtained: the more trees, the more accurate the results. The difference between the Random Forest algorithm and the Decision Tree algorithm is that in a Random Forest, the process of finding the root node and splitting feature nodes is run randomly. A Decision Tree is a decision support tool that uses a dendrogram to show possible consequences. If a training dataset with a target and features is fed into a Decision Tree, it will formulate a set of rules. Overfitting is a key problem that can worsen the results in Decision Trees, but with the Random Forest algorithm, if there are enough trees in the forest, the classifier will not overfit the model. Another advantage is that the Random Forest classifier can handle missing values, and a final advantage is that the Random Forest classifier can model for class values.

[0063] Support Vector Machines (SVMs) are primarily a classifier method that performs classification tasks by constructing hyperplanes in a multidimensional space that separates different class labels. SVMs support both regression and classification tasks and can handle multiple continuous and categorical variables. Suppose we have some given data points that each belong to one of two classes, and the goal is to determine which class a new data point will belong to. In the case of an SVM, the data points are viewed as p-dimensional vectors (a list of p numbers), and we want to know if we can separate these points using a (p-1)-dimensional hyperplane. This is called a linear classifier. Many hyperplanes exist that could potentially classify the data. A reasonable choice for the optimal hyperplane is one that represents the maximum margin or margin between the two classes. We choose a hyperplane that maximizes the distance from it to the nearest data point on each side. If such a hyperplane exists, it is called a maximum margin hyperplane, and the linear classifier defined by it is called a maximum margin classifier; or, equivalently, a perceptron with optimal stability.

[0064] Linear discriminant analysis (LDA) is a classification method originally developed by Ra and A. Fisher in 1936. It is simple, mathematically robust, and typically produces models with accuracy as good as more complex methods. LDA is based on the concept of searching for the optimal linear combination of variables (predictors) that separate two classes (targets). Closely related to analysis of variance (ANOVA) and regression analysis, it also attempts to represent a dependent variable as a linear combination of other features or measures.

[0065] Microarray predictive analysis (PAM) is a statistical technique for class prediction from gene expression data using the nearest shrinking centroid. This method identifies the subset of genes that best characterize each class.

[0066] Generalized linear models (GLMs) are a large class of models, including linear regression, ANOVA, Poisson regression, and log-linear models, but GLMs have some limitations, such as linear functions, only linear predictors in system components, and independent responses.

[0067] Typically, classic machine learning techniques make predictions directly based on a pre-specified set of features. However, representation learning techniques transform features into intermediate representations before mapping them to the final prediction. Deep learning (DL) is a form of representation learning that uses multiple transformation steps to create highly complex features. DL is widely used in pattern recognition, image processing, computer vision, and more recently, bioinformatics. DL is classified as a feedforward artificial neural network (ANN), which uses more than one hidden layer (Y) connecting the input layer (X) and the output layer (Z) via a weight (W) matrix. The weight matrix W, which is expected to minimize the difference between the input layer (x) and the output layer (z), is considered optimal and is selected by the system to obtain the best result.

[0068] Types of CHD and Prevention and Treatment: The methods described in this article can be used to diagnose, detect, or predict one or more cardiac defects in a subject. There are many types of cardiac defects, which are generally grouped according to the part of the heart affected. Defects can be defects in heart valves, diaphragms, internal valves of the heart, or ventricles. Valve defects include, for example, defects in the aortic valve and pulmonary valve. Defects in the diaphragm include problems with the walls that separate the right and left ventricles of the heart. When the diaphragm fails to close completely during heart development, a small opening causes the walls to separate the right and left ventricles. Some defects are due to the valves that separate the top and bottom ventricles of the heart failing to form properly. Some types of cardiac defects involve combinations of different defects or must be related to the way the heart circulates during heart development.

[0069] Examples of CHD include aortic stenosis (AVS), hypoplastic left ventricular syndrome (HLHS), ventricular septal defect (VSD) (including VSD with atrial septal defect), tetralogy of Fallot (TOF), coarctation of the aorta (Coarct), atrial septal defect (ASD), pulmonary artery stenosis (PS) (including pulmonary valve stenosis and pulmonary valvular stenosis), truncus arteriosus, double aortic arch, and bicuspid AV valve (including those with dilated main pulmonary arteries). Patients diagnosed with one or more of these types of CHD require surgery to prevent serious complications or death.

[0070] The methods described in this article provide accurate and early detection and prediction of one or more cardiac defects. Early prenatal diagnosis of CHD allows for early assessment and optimal treatment at birth, and reduces mortality (Holland BJ et al. Ultrasound Ostet. Genecology 2015:45:631-638) and morbidity compared to cases diagnosed later after birth. Furthermore, early postnatal diagnosis of CHD is associated with improved survival compared to late diagnosis in critically ill CHD cases (Eckersley L, Sadler L et al. Arch Dis Child 2015:0:1-5). Therefore, the methods described in this article facilitate accurate prenatal diagnosis, which aids in early assessment, such as during the neonatal period, treatment, and improved survival in CHD cases.

[0071] The cytosine methylation described herein refers to the chemical addition of a methyl group or a single carbon atom to a cytosine nucleotide. A key dietary source of the carbon atom used for cytosine methylation is folic acid. Given the applicant's findings demonstrating the importance of cytosine methylation in the pathogenesis of CHD, it is reasonable to expect that dietary folic acid supplementation will reduce the risk of CHD. Currently, folic acid fortification in cereals and bread, as well as direct supplementation consumption by the entire population, including pregnant women, has been standard care for the prevention of neural tube defects. This provides a “natural” experiment to determine the value of folic acid supplementation in the prevention of CHD. Studies in China (Mao B. et al. 2017), the United States (Botto et al. 2003), Europe (Czeizel et al. 1998), and Canada (Ionescu-Ittu et al. 2009) have shown the beneficial effects of folic acid supplementation in reducing CHD. It should be noted that a significant proportion of pregnant women may not respond adequately to folic acid supplementation. This is because a significant proportion of women in some populations have mutations in the methylenetetrahydrofolate reductase (MTHFR) gene, which encodes a similarly named enzyme responsible for folate metabolism, the end result of which is the production of methyl carbon involved in cytosine methylation. Therefore, MTHFR results in weakened cytosine methylation. In some populations, the frequency of this mutation can be as high as 20%. Furthermore, when the mutation is present, enzyme activity is reduced by up to 30% (Botto et al., 2000). Therefore, MTHFR mutations are expected to reduce the effectiveness of folate supplementation in such populations.

[0072] Fortunately, there are alternative sources of methylation that are not affected by MTHFR gene mutations. These include choline and betaine, and are found in dietary sources such as broccoli, spinach, beets, liver, and other foods. Based on the applicant's findings on the importance of DNA methylation in CHD, population-based fortification and individual supplementation programs for choline and betaine can be evaluated. Furthermore, current evidence suggests that less than one-tenth of the US population, including pregnant women, has adequate choline consumption (Zeisel et al., 2009). The risk of pregnancy-related choline deficiency is amplified by the fact that this is a period of increased choline demand.

[0073] Laboratory evidence suggests the importance of betaine (an alternative source of single carbon for cytosine methylation) in the prevention of CHD (Karunamuni et al. 2017). Prenatal alcohol exposure is a significant risk factor for CHD. It is estimated that 10% of pregnant women in the United States drink alcohol, and 3% have high levels of alcohol consumption (Tan et al., 2015). Prenatal alcohol intake has been shown to interfere with single carbon metabolism and DNA methylation in mice (Liu et al., 2009). In the study by Karunamuni et al. (Karunamuni et al., 2017), betaine supplementation in pregnant mice exposed to alcohol reduced the frequency of CHD and had a beneficial effect on structural changes in the heart and great vessels induced by alcohol exposure. Importantly, alcohol led to a decrease in 5mC nucleotide levels in cardiac neural crest cells of pups with CHD. Cardiac neural crest cells are a specialized group of cells that are embryologically crucial for the development of important cardiac structures such as the interventricular septum, aorta, pulmonary artery, and other blood vessels. The 5mC levels of these cardiac neural crest cells were restored to normal by supplementing pregnant mice exposed to prenatal alcohol with betaine. The aforementioned information confirms the importance of methylation changes in cardiac development. However, it does not provide any information regarding the use of methylation levels in cf DNA for CHD detection. The method described in this article is a minimally invasive blood test to assess the mechanisms of CHD. This test can be performed during early cardiac development, close to early pregnancy. It is known that fetal cell-free DNA is expelled into the maternal circulation during early pregnancy. Understanding the methylation status of the developing fetus could potentially serve as a basis for supplementing individual patients with folic acid, choline, or betaine to help prevent or mitigate the development or severity of CHD.

[0074] From a population perspective, the insights generated by the applicant's findings (i.e., evidence of profound changes in cytosine methylation in key cardiac genes in CHD cases using cfF DNA from developing fetuses) can form the basis for population-based policies on choline and betaine supplementation. This is particularly important for populations with high MTHFR mutation rates, which render folic acid supplementation less effective. Therefore, the applicant's findings can serve as the basis for important preventative and even therapeutic interventions.

[0075] Once a patient is diagnosed with CHD, medication can be used to maintain a regular heartbeat. Immediate treatment for CHD involves providing adequate oxygen levels until cardiac repair is possible. Surgery, such as open-heart surgery, can be performed to repair any defects in the heart. Other treatments depending on the type of CHD include antibiotics, cardiac catheterization, open-heart surgery, and heart transplantation. Medications for CHD may include angiotensin-converting enzyme (ACE) inhibitors, angiotensin II receptor blockers (ARBs), beta-blockers, diuretics, and antihypertensive agents.

[0076] In the implementation plan, the methods described herein can prevent the development of CHD during future pregnancies. As an example, pregnant women can supplement with folic acid or folate.

[0077] Table of Genes and Genomic Loci (CHD Markers): This disclosure reports a strong correlation between cytosine methylation status at numerous cytosine sites throughout the genome using a strict false discovery rate (FDR) analysis, where q values ​​are <0.05 and many q values ​​are as low as <1×10⁻⁶. -30 This depends on the specific cytosine site considered. Epigenomic analysis was performed on 12 CHD patients and 26 normal controls. Significant differences in cytosine methylation patterns at multiple sites throughout the DNA were found in all tested CHD cases compared to normal controls. The genomic sites described herein are located in or associated with known genes. These findings are consistent with altered expression of multiple genes in CHD cases compared to controls.

[0078] Tables 1 through 6 provide genomic loci that can be used individually to predict, detect, or diagnose CHD in patients. Genomic loci are provided below each of Tables 2 through 6. One or more genomic loci from Tables 1 through 6 can be selected for predicting or diagnosing CHD in patients. In this embodiment, two or more of these loci may be selected from those in Tables 1, 2, 3, 4, 5, and / or 6.

[0079] A total of 5918 CpG genomic loci were identified, covering 4976 genes (FDRp value ≤ 0.01). Table 1 provides the top 1000 genomic loci obtained through whole-genome methylation profiling analysis using cfF DNA. One or more, two or more, or up to all 1000 genomic loci in Table 1 can be selected for predicting, detecting, or diagnosing CHD in patients. In an implementation, one or more, two or more, or up to 5, 10, 15, 20, 30, 40, 50, 60, 70, 80, 90, 100, 200, 300, 400, 500, 600, 700, 800, 900, or 1000 genomic loci disclosed in Table 1 can be selected for predicting CHD. In the implementation scheme, the genomic loci have an AUC (with 95% CI) of ≥0.70, 0.75, 0.80, 0.85, 0.86, 0.87, 0.88, 0.89, 0.90, 0.91, 0.92, 0.93, or 0.94. Of the 1000 genomic loci, 130 CpGs were hypomethylated, and 870 markers were hypermethylated and associated with CHD. 53 hypomethylated and 486 hypermethylated markers were found to have differential methylation differences ≥10%. Three genomic loci with methylation differences greater than 20% are: cg06301252 (PTPRN2) with 33.62% methylation difference; cg02807450 (MTMR2) with -21.15% methylation difference; and cg12900404 (DOCK10) with -20.13% methylation difference. 126 CpG loci individually showed an AUC ≥ 0.80, indicating very good to excellent predictive accuracy for disease prediction. In the implementation plan, the genomic loci used for the detection, prediction, and / or diagnosis of CHD include cg06301252, cg02807450, and cg12900404.

[0080] AUC combines sensitivity and specificity values ​​and provides a more precise indication of test accuracy. An AUC with a 95% CI indicates an AUC with a statistically significant 95% confidence interval. An AUC ≥ 0.70 indicates a clinically useful test.

[0081] Tables 2 through 6 provide genomic loci obtained through AI analysis using six different platforms: SVM, GLM, PAM, RF, LDA, and DL. Although loci are listed under each table, they are part of a numbered table summarizing the AUC, sensitivity, and specificity of each algorithm. In the implementation, the genomic loci used for prediction, detection, and diagnosis include one to five loci from each of the six algorithms listed in Tables 2 through 6. The top five predicted markers for each model are provided in descending order (under each table) within each table.

[0082] In this implementation, the genomic loci are selected from algorithms with an AUC (with 95% CI) of ≥0.8800, 0.8900, 0.9000, 0.9100, 0.9200, 0.9300, 0.9400, 0.9500, 0.9600, or 0.9700. In this implementation, the genomic loci are selected from algorithms with a sensitivity and / or specificity of ≥0.8700, 0.8800, 0.8900, 0.9000, or 0.9100. In this implementation, the genomic loci are selected from one to five loci from the RF, SVM, or DL ​​algorithms listed in Table 2.

[0083] Table 3 shows comparable predictive performance when demographic markers are considered together with CpG loci. Table 4 shows high performance when only markers meeting strict GWAS thresholds are considered. Table 5 shows high performance when considering both demographic markers and CpG loci meeting strict GWAS thresholds. Table 6 shows high predictive accuracy observed when only markers showing high levels of methylation change (e.g., 1.5-fold or greater) are used.

[0084] The range described throughout the application includes a specified range, subranges within the specified range, individual numbers within the range, and endpoints of the range. For example, a specification for a range such as one or more to up to 1000 includes subranges such as one or more to 500 or more, 10 or more to 20 or more, one or more to 5 or more, and individual numbers within that range, such as 1, 2, 3, 4, 5, 10, 20, 100, and 500. Furthermore, as another example, a specification for a range ≥0.70 would include 0.70 to 1.00 and include all individual numbers of 0.70 and 1.0.

[0085] The results presented in this article confirm that the predisposition or risk of having CHD can be determined based on the differences in methylation levels at cytosine sites between CHD and normal cases throughout the human genome.

[0086] The reported genomic loci enable targeted screening studies that predict and detect CHD based on cytosine methylation throughout the genome. It also allows for improved understanding of CHD development mechanisms, such as by evaluating cytosine methylation data using gene ontology analysis. In implementations, genomic loci are used in many different combinations to predict, detect, or diagnose CHD in subjects. In implementations, genomic loci are used to determine or calculate a subject's risk or predisposition to CHD at any time before birth or during any period after birth.

[0087] Table 1

[0088]

[0089]

[0090]

[0091]

[0092]

[0093]

[0094]

[0095]

[0096]

[0097]

[0098]

[0099]

[0100]

[0101]

[0102]

[0103]

[0104]

[0105]

[0106]

[0107]

[0108]

[0109]

[0110]

[0111]

[0112]

[0113]

[0114]

[0115]

[0116]

[0117]

[0118]

[0119]

[0120]

[0121]

[0122]

[0123]

[0124]

[0125]

[0126]

[0127]

[0128]

[0129]

[0130]

[0131]

[0132]

[0133]

[0134]

[0135]

[0136]

[0137]

[0138]

[0139]

[0140]

[0141]

[0142]

[0143]

[0144]

[0145]

[0146]

[0147]

[0148]

[0149]

[0150]

[0151]

[0152] Table 2: Artificial Intelligence and Circulating cfF DNA Prediction for CHD*

[0153]

[0154]

[0155] *Individual CpG markers were selected based on methylation difference ≥5% and CHD detection AUC ≥0.70.

[0156] Individual predictors in descending order of contribution:

[0157] SVM: cg04761177, cg21431091, cg01263077, cg09853933, cg27142059

[0158] GLM: cg24479965, cg01094213, cg22467129, cg24748945, cg01949461

[0159] PAM: cg27142059, cg09386284, cg16551159, cg04761177, cg01263077

[0160] RF: cg04761177, cg16551159, cg14957943, cg06978680, cg12592721

[0161] LDA: cg04761177, cg27142059, cg18073832, cg25731807, cg03790075

[0162] DL: cg04761177, cg21431091, cg01263077, cg09853933, cg27142059

[0163] Table 3: Artificial Intelligence and Circulating cfF DNA, Clinical and Demographic Predictions for CHD*

[0164]

[0165] *Individual CpG markers were selected based on methylation difference ≥5% and CHD detection AUC ≥0.70.

[0166] Individual predictors in descending order of contribution:

[0167] SVM: cg04761177, cg21431091, cg01263077, cg09853933, cg27142059

[0168] GLM: cg24479965, cg01094213, cg22467129, cg24748945, cg01949461

[0169] PAM: cg27142059, cg09386284, cg16551159, cg04761177, cg01263077

[0170] RF: cg04761177, cg16551159, cg14957943, cg06978680, cg12592721

[0171] LDA: cg04761177, cg27142059, cg18073832, cg25731807, cg03790075

[0172] DL: cg04761177, cg21431091, cg01263077, cg09853933, cg27142059

[0173] Table 4: Artificial Intelligence and Stringent Standards for Predicting Circulating cfF DNA in CHD*

[0174]

[0175] *Limited to individuals with CpG markers meeting the strict thresholds for GWAS studies (p-value < 5 x 10⁻⁶). -8 )

[0176] Individual predictors in descending order of contribution:

[0177] SVM: cg04761177, cg15277677, cg03790075, cg04626875, cg11782260

[0178] GLM: cg13598434, cg05349624, cg10259004, cg04761177, cg11196182

[0179] PAM: cg18198743, cg08316054, cg02394812, cg00280345, cg08052226

[0180] RF: cg04761177, cg244128480.62684, cg01637563, cg17720707, cg05349624

[0181] LDA: cg03790075, cg04761177, cg05349624, cg14809932, cg01637563

[0182] DL: cg13598434, cg05349624, cg10259004, cg04761177, cg11196182

[0183] Table 5: Artificial Intelligence, Circulating cfF DNA* Meeting Strict Standards, Clinical and Demographic Predictions for CHD

[0184]

[0185] *Limited to single CpG markers that meet the stringent thresholds for GWAS studies (p-value < 5 x 10⁻⁶). 8 )

[0186] Individual predictors in descending order of contribution:

[0187] SVM: cg04761177, cg15277677, cg03790075, cg04626875, cg11782260

[0188] GLM: cg13598434, cg05349624, cg10259004, cg04761177, cg11196182

[0189] PAM: cg18198743, cg08316054, cg02394812, cg00280345, cg08052226

[0190] RF: cg04761177, cg244128480.62684, cg01637563, cg17720707, cg05349624

[0191] LDA: cg03790075, cg04761177, cg05349624, cg14809932, cg01637563

[0192] DL: cg13598434, cg05349624, cg10259004, cg04761177, cg11196182

[0193] Table 6: Artificial Intelligence and Circulating cfF DNA* Prediction for CHD

[0194]

[0195] *Limited to individual CpG markers showing a ≥1.5-fold change in methylation compared to the control in CHD.

[0196] Individual predictors in descending order of contribution:

[0197] SVM: cg09493833, cg27563174, cg22752533, cg05279901, cg03741571

[0198] GLM: cg19803352, cg24479965, cg07287606, cg22467129, cg24509810

[0199] PAM: cg27142059, cg13598434, cg14200609, cg09493833, cg20360734

[0200] RF: cg27142059, cg26078733, cg14200609, cg23273875, cg24509810

[0201] LDA: cg27142059, cg09493833, cg19803352, cg17485454, cg07287606

[0202] DL: cg27142059, cg26078733, cg14200609, cg23273875, cg24509810

[0203] Microarrays: Differential methylation can be analyzed using microarray systems. Nucleic acids can be linked to chips, such as microchips. See, for example, U.S. Patent Nos. 5,143,854; 6,087,112; 5,215,882; 5,707,807; 5,807,522; 5,958,342; 5,994,076; 6,004,755; 6,048,695; 6,060,240; 6,090,556; and 6,040,138. Binding to nucleic acids on the microarray can be detected by scanning the microarray with various laser-based or charge-coupled device (CCD) scanners and extracting features using software packages such as IMAGENE (Biodiscovery, Hawthorne, CA), FeatureExtraction Software (Agilent), SCANALYZE (Eisen, M. 1999. SCANALYZE User Manual, Stanford Univ., Stanford, Calif. Ver 2.3.2.), or Genepix (Axon Instruments). The full set of loci will include one or more genomic loci listed in Tables 1 through 6, which have been individually shown as potentially clinically useful tests with an AUC ≥ 0.70.

[0204] Kit: Describes a kit for predicting and diagnosing CHD based on CpG site methylation on nucleic acids. The kit may include components for extracting nucleic acids (including DNA and RNA) from biological samples, components of a microarray system, and / or components for analyzing differentially methylated genomic sites.

[0205] As described in this article, biomarker testing for CHD can facilitate early and accurate diagnosis, thereby supporting CDC planning and management goals. Given the evidence that a significant percentage, or even the majority, of major CHD cases remain undiagnosed, accurate biomarkers are a crucial and necessary complement to any effective treatment strategy.

[0206] The methods disclosed herein include predicting, detecting, or diagnosing CHD and / or calculating the risk or predisposition to develop CHD. The methods described herein can be used for early prevention and / or treatment (including symptom relief or alleviation) of patients to prevent death or the development of severe CHD-related symptoms. Subjects or patients who require (or need) prediction, diagnosis, and / or treatment are those who may have CHD and require diagnosis and treatment.

[0207] As will be understood by those skilled in the art, each embodiment disclosed herein may include, consist of, or be substantially composed of the elements, steps, ingredients, or components specifically stated herein. Therefore, the terms “comprising” or “including” should be interpreted as stating: “comprising, consisting of, or substantially consisting of.” The transitional terms “comprising” or “including” mean, but are not limited to, and even in major quantities, allow for the inclusion of unstated elements, steps, ingredients, or components. The transitional phrase “consisting of” excludes any unstated elements, steps, ingredients, or components. The transitional phrase “substantially consisting of” limits the scope of the embodiment to the specified elements, steps, ingredients, or components and those that do not substantially affect the embodiment. For example, steps that do not affect the detection, prediction, diagnosis of CHD, or the prevention or treatment of CHD in patients.

[0208] Furthermore, unless otherwise stated, the figures used in the specification and claims to represent quantities of ingredients, components, reaction conditions, etc., should be understood to be modified by the term "about". Therefore, unless indicated to the contrary, the numerical parameters set forth in the specification and appended claims are approximate values ​​that may vary depending on the desired properties sought to be obtained from the subject matter presented herein. At least and without attempt to limit the application of the doctrine of equivalence to the scope of the claims, each numerical parameter should be interpreted at least according to the number of significant figures reported and by applying ordinary rounding techniques. While the wide range of numerical ranges and parameters set forth in the subject matter presented herein are approximate values, the values ​​set forth in specific embodiments are reported as precisely as possible. However, any numerical value inherently contains some error necessarily caused by the standard deviation found in its respective test measurements.

[0209] When further clarification is required, the term “about” has the meaning that a person skilled in the art would reasonably attribute to it when used in conjunction with the stated value or range, namely, slightly greater than or slightly less than the value or range, within ±20% of the value; within ±15% of the value; within ±10% of the value; within ±5% of the value; within ±4% of the value; within ±3% of the value; within ±2% of the value; within ±1% of the value; or any percentage between ±1% and 20% of the value.

[0210] Unless otherwise stated herein or clearly contradicted by the context, the terms “a,” “an,” “the,” and similar indicators used in the context of describing the claimed subject matter (particularly in the context of the following claims) shall be interpreted to cover both the singular and the plural.

[0211] The following exemplary embodiments and examples illustrate the exemplary methods provided herein. These exemplary embodiments and examples are not intended and should not be construed as limiting the scope of this disclosure. It will be clear that the methods may be implemented in ways other than those specifically described herein. In view of the teachings herein, many modifications and variations are possible and therefore within the scope of this disclosure.

[0212] Exemplary Implementation

[0213] Exemplary embodiments include, but are not limited to:

[0214] 1. The methods described herein include using nucleic acids obtained from biological samples of a subject to diagnose, detect, and / or predict CHD. Subjects may be fetuses, embryos, newborns, infants, children, adolescents, or adults. Subjects may be pregnant women. Biological samples include tissue samples, including placental tissue, or bodily fluids such as blood, plasma, serum, urine, saliva, sputum, sweat, tears, genital secretions (including cervical secretions), amniotic fluid, and umbilical cord blood obtained at birth. In embodiments, the nucleic acid is DNA. The DNA may be cellular DNA or cell-free (cf) DNA. cf DNA may be cf fetal (cfF) DNA.

[0215] 2. The methods described in this paper include using DNA and cfF DNA to determine the epigenetic mechanisms of CHD.

[0216] 3. The epigenetic mechanisms described herein include all forms of epigenetic testing based on the use of DNA and cfF DNA, such as changes in DNA methylation and histone modifications, including but not limited to: methylation, acetylation, ubiquitination, and phosphorylation. Histones are proteins that wrap around the DNA strand. Histone modifications contribute to direct alterations in DNA cytosine methylation and play a crucial role in genes.

[0217] 4. The nucleic acid methylation changes (including DNA methylation changes) described herein encompass various forms of methylation changes, including cytosine methylation, cytosine hydroxymethylation, and other forms of cytosine epigenetic modifications. Adenine nucleotides can also undergo DNA methylation changes. Therefore, the methods described herein include using cfF DNA to measure all forms of cytosine modifications that fall within the definition of adenine methylation for the purpose of predicting or monitoring CHD using cfF DNA.

[0218] 6. DNA epigenetic changes based on cff DNA as described herein (including all forms of methylation changes in cytosine or adenine, and including all forms of histone epigenetic modifications) for the detection of other fetal congenital abnormalities in which DNA methylation or histone modifications play a role in their development, such as neural tube defects, cleft lip, and cleft palate.

[0219] 7. The methods described herein include using cfF DNA for epigenetic monitoring of the impact of exposure on pregnancy in response to the risk of developing relevant cardiac abnormalities (e.g., exposure to alcohol or drugs, chemicals or other known risk factors) or conditions during an ongoing pregnancy (e.g., maternal diabetes or hypertension).

[0220] 8. A method for predicting or diagnosing congenital heart defects (CHD) in a subject in need, wherein the method comprises: measuring a biological sample containing cf nucleic acid obtained from the subject to determine the frequency or percentage of cytosine methylation at one or more sites throughout the genome; and comparing the cytosine methylation level of the sample with that of a control sample.

[0221] 9. The method according to any one of the implementation schemes 1-8, wherein the method further includes the use of artificial intelligence (AI) technology.

[0222] 10. The method according to any one of embodiments 1-9, wherein the method further comprises using artificial intelligence (AI) techniques including one or more of the following machine learning algorithms: random forest (RF), support vector machine (SVM), linear discriminant analysis (LDA), microarray analysis prediction (PAM), generalized linear model (GLM), or deep learning (DL).

[0223] 11. The method according to any one of the implementation schemes 1-10, wherein the method further includes: calculating the risk of the subject developing CHD.

[0224] 12. The method according to any one of embodiments 1-11, wherein the control sample comprises one or more biological samples from one or more normal (healthy) patients or from one or more patients diagnosed with CHD.

[0225] 13. The method according to any one of the implementation schemes 1-12, wherein the biological sample includes body fluids.

[0226] 14. The method according to any one of embodiments 1-13, wherein the biological sample includes blood, plasma, serum, urine, saliva, sputum, sweat, respiratory condensate, tears, genital secretions including cervical secretions, amniotic fluid, placental tissue, CVS sample and umbilical cord blood obtained at birth.

[0227] 15. According to any one of the implementation schemes 1-14, wherein cf nucleic acid includes cfF nucleic acid.

[0228] 16. The method according to any one of embodiments 1-15, wherein the biological sample contains cfF nucleic acid from early pregnancy, mid pregnancy and / or late pregnancy.

[0229] 17. According to any one of the implementation schemes 1-16, wherein the cf nucleic acid includes DNA.

[0230] 18. The method according to any one of embodiments 1-17, wherein the one or more sites include one or more sites from Tables 1 to 6.

[0231] 19. The method according to any one of embodiments 1-18, wherein the one or more sites include at least two, at least three, at least four, at least five, at least six, at least seven, at least eight, at least nine, or at least ten sites from Tables 1 to 6.

[0232] 20. The method according to any one of embodiments 1-19, wherein the one or more sites include an AUC (with 95% CI) greater than 0.70, 0.75, 0.80, 0.85, 0.90, 0.91, 0.92, 0.93, 0.94, 0.95, 0.96, or 0.97.

[0233] 21. The method according to any one of embodiments 8-20, wherein the determination is based on bisulfite methylation determination or whole-genome methylation determination.

[0234] 22. The method according to any one of the implementation schemes 1-21, wherein the one or more sites include cg06301252, cg02807450 or cg12900404.

[0235] 23. The method according to any one of embodiments 1-22, wherein the one or more sites include cg04761177, cg21431091, cg01263077, cg09853933, cg27142059, cg16551159, cg14957943, cg06978680 or cg12592721.

[0236] 24. The method according to any one of the implementation schemes 1-23, wherein the method further includes treating CHD.

[0237] 25. The method according to any one of embodiments 1-24, wherein the method further includes treating the subject by administering a drug and / or performing surgery.

[0238] 26. The method according to any one of the implementation schemes 1-23, wherein the method prevents CHD in future pregnancies.

[0239] 27. The method according to Example 26, wherein the method includes supplementing the pregnant mother with folic acid or folic acid salt.

[0240] Example

[0241] In this study, a total of 12 cases of CHD were analyzed. These 12 cases included: 1 case of pulmonary atresia with pulmonary valve stenosis; 4 cases of ventricular septal defect (VSD); 1 case of truncus arteriosus; 2 cases of tetralogy of Fallot; 1 case of pulmonary valve stenosis; 1 case of ventricular septal defect combined with atrial septal defect; 1 case of double aortic arch; and 1 case of mitral valve (AV) with dilated main pulmonary artery. A total of 12 cases of CHD were included. Demographic data are provided in Supplementary Table S1. Table 1 provides epigenomic data. Tables 2 through 6 provide epigenomic data obtained in combination with artificial intelligence techniques.

[0242] Introduction: In a series of studies (Radhakrishna et al. 2018; Bahado-Singh et al. 2019a; Radhakrishna et al. 2019), methylation changes were confirmed in genes and gene networks identified by placental DNA that were epigenetically dysregulated in isolated VSDs and nonsyndromic tetralogy of Fallot (two of the most important CHD categories). These studies contribute to providing important insights into the pathogenesis of these CHDs. Furthermore, CHDs can be accurately screened using DNA methylation markers derived from the placenta. While the placenta is ideal for such studies given its abundance and limited clinical value, applying the results to prenatal prediction of CHDs represents a significant challenge. Obtaining placental trophoblast tissue suitable for DNA analysis typically requires invasive procedures such as chorionic villus sampling or placental biopsy (Alfirevic et al., 2003). This procedure is painful, requires expertise, and may be associated with an increased risk of pregnancy complications. However, these previous studies did not address whether cf DNA released from the placenta into circulation could be used to detect CHD in a developing fetus or embryo.

[0243] Over the past two decades, genomic analysis of efF DNA present in maternal circulation during pregnancy has now developed into widespread clinical applications for detecting fetal aneuploidy (Goldwaser & Klugman 2018) and other chromosomes (Keet et al. 2015; Grace et al. 2016) and molecular pathology (Stewart et al. 2018). However, "cell-free fetal DNA" is an imprecise term, as the DNA actually originates from the placenta itself, i.e., embryonic fetal tissue. The placental trophoblast continuously proliferates, differentiates, and undergoes apoptosis (Taglauer et al., 2014). Apoptotic material from the placenta is continuously shed into the maternal circulation. This apoptotic material from trophoblast cells constitutes the cell-free "fetal" DNA found in maternal blood (Gupta et al. 2004; Tjoa et al. 2006) and in clinical studies (Wataganara et al. 2005; Alberry et al. 2007). Cell-free "fetal" DNA constitutes a significant percentage of cf DNA in maternal circulation. The contribution of fetal physiological fraction (fetal fraction) to overall maternal cfDNA blood levels accelerates with increasing gestational age (Rafaeli-Yehudai et al., 2018). In the following study, cfDNA was extracted from the plasma of pregnant women (Zolotukhina et al., 2005). DNA methylation analysis was performed on the cfDNA. cfDNA-based DNA methylation analysis has been used to help elucidate the epigenetic pathogenesis of CHD development and has also been used to predict non-syndromic CHD based on changes in DNA methylation observed in circulating cfF DNA in maternal circulation during mid-pregnancy.

[0244] Methods: Study of samples and cf DNA extraction. This study was approved by the Human Ethics Committee, and informed written consent was provided to each participant. A total of 12 cases and 26 controls were analyzed. Participants were mothers who gave birth to CHD cases and normal infants. The mean age of the cases was 29.7 years, and that of the controls was 31.3 years. cfF DNA was extracted from maternal blood. Blood samples were obtained from the participants during pregnancy. Blood samples were directly extracted into Streck cell-free DNA. To ensure good quality of cf-DNA from plasma, the sample tubes were then further processed by centrifuging at 3000×g for 15 minutes within 24 hours of blood collection. Plasma was aliquoted into 2.0 mL microcentrifuge collection tubes without disturbing the erythrocyte sedimentation rate (ESR) layer. The aliquots were stored at -80°C until further processing (Sheinerman et al. 2017). cf-DNA was extracted from 5 mL of plasma using the QIAamp Cyclic Nucleic Acid Kit (QiagenCat#55114). The method used was the manual vacuum method with a QIAvac 24 Plus vacuum manifold, following the manufacturer's protocol. The ability to incorporate silica membrane technology for fragmented DNA enabled efficient DNA recovery. This method allowed for further bisulfite conversion and methylation profiling analysis of the DNA by maintaining its methylation state.

[0245] Bisulfite conversion and methylation array processing: DNA bisulfite conversion was performed using the EZ DNA Methylation Gold Kit (Zymo, USA) with 10 μL of elution buffer (Hardy et al., 2017) according to the manufacturer's instructions. Methylation profiling was performed using an Illumina Infinium Methylation Epic BeadChip array with methylation labels exceeding 850K, according to the manufacturer's instructions. Samples were randomized on the chip. Both case and control sample processing were performed together to avoid any sample processing bias. The array was dried using a vacuum desiccator, and the processed magnetic bead chip was imaged using the Illumina iScan system according to the company's instructions.

[0246] Statistical and Bioinformatics Analysis: Following scanning, raw iDat files were downloaded from iScan software, and data were analyzed using GenomeStudio software. β values ​​were compared between case and control subjects for differential methylation analysis at individual CpG loci. Further downstream statistical analyses were described in previous publications (Bahado-Singh et al. 2019a; Bahado-Singh et al. 2019b). For further downstream analysis, CpG probes constituting the X and Y chromosomes were excluded, followed by dbSNP entries within 10 bp of the CpG locus, to avoid sex bias and the genetic effects on methylation sites, respectively (Wilhelm-Benartzi et al. 2013). β values ​​with a BH-adjusted FDR p-value < 0.01 were considered significant cutoffs. The area under the receiver operating characteristic curve (AUC-ROC) with 95% CI was calculated for each significant CpG locus using the dplyr, reshape2, and ROCR packages of R tools.

[0247] Artificial Intelligence (AI) Analysis: Methods for AI Analysis are described in Bahado-Singh et al., 2019b. The following is an overview of the previously published description. Epigenomic data were divided into two groups: a training group consisting of 80% of the study subjects and a test group comprising the remaining 20%. This is a commonly used approach when analyzing small datasets. We performed 10-fold cross-validation on the data from the training group to generate a predictive model. This model was then evaluated in either the independent validation or test group. Six appropriate AI methods or algorithms were used to determine the screening performance of CHD detection. These are Random Forest (RF), Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), Microarray Predictive Analysis (PAM), Generalized Linear Model (GLM), and Deep Learning (DL), which is the latest form of AI and is increasingly used to analyze complex and large volumes of biological data, such as epigenomics.

[0248] Randomized Decision Trees (RF) is a supervised classification algorithm used for classification, regression, and other functions. A forest of decision trees is randomly created, and the average prediction of each tree is determined. There is a direct correlation between the number of trees in the forest and the accuracy of the generated results. Increasing the number of trees will increase the accuracy of the obtained results. RF has the advantage of being able to handle missing values ​​in the dataset and can utilize categorical values ​​(Huang et al., 2013). First, labeled data identifying different groups is fed to an SVM (supervised learning), thereby building a model to distinguish the groups. Subsequently, when provided with unlabeled fresh data, it is possible to develop a model or hyperplane to cluster one group with another. SVMs are capable of performing regression and classification tasks and can handle multiple continuous and categorical variables (Mahadevan et al., 2008). Least Significant Analytical Trees (LDA) are used to reduce the number of features or predictors needed for accurate classification and grouping. This is particularly useful for epigenomic analysis, as studies begin with nearly 900,000 latent features for CHD detection. LDA is a simple method, yet it can still achieve excellent accuracy. The accuracy is as good as more complex methods. LDA is based on the identification of linear combinations of variables (predictors) that optimally separate two categories (targets) (Liland 2011). It is closely related to Analysis of Variance (ANOVA) and regression analysis, which attempts to define outcome variables based on combinations of explanatory variables. PAM is a statistical technique for class prediction from gene expression data using the nearest contracting centroid (Alakwaa et al. 2018; Candel et al. 2018). This method identifies the subset of genes that best represent each category. GLM is a broad class of models that includes linear regression, ANOVA, Poisson regression, log-linear models, etc. (Alakwaa et al., 2018; Candel et al. 2018). DL is a form of representation learning that uses multiple transformation steps to create very complex features. DL is classified as a feedforward artificial neural network (ANN) that uses more than one hidden layer (y) connecting the input layer (x) and the output layer (z) via a weight (W) matrix. The expected weight matrix minimizes the difference between the input and output layers and is considered the best AI method (Alakwaa et al., 2018; Candel et al., 2018).

[0249] Multivariate regression analysis provides the model: quantile normalization and automatic scaling are performed on the epigenomic data prior to the multivariate regression analysis. As a quality control step, principal component analysis (PCA) was performed on all categories using MetaboAnalyst (v4.0) (Chong et al., 2018) to investigate the presence of any systematic variations and detect potential outliers. Subsequently, these preprocessed data were used to perform partial least squares discriminant analysis (PLSDA). The ideal number of CpG β-value variables was carefully selected based on predictive accuracy and cross-validation using leave-one-out cross-validation methods available in MetaboAnalyst. The goodness-of-fit (R²) and predictability (Q²) values ​​for each PLSDA model have been reported. During cross-validation, a permutation test with 2000 iterations was performed, which minimizes the likelihood of segregation observed on PLSDA, considering the possibility of p-values ​​< 0.05.

[0250] Cluster analysis of differentially methylated targets in cfF DNA CHD: Heatmaps were generated using individual β values ​​of significantly differentially methylated markers between cases and controls (Figure 1). If less than 20% of the values ​​were missing, the median for the specific methylation site was used to estimate the missing values. Any methylation markers showing more than 20% missing values ​​were omitted. Furthermore, data were filtered using the interquantile range (IQR), and quantiles were normalized to avoid dilution effects. In hierarchical cluster analysis, Euclidean distance and the Ward clustering algorithm were used, and the results were viewed using PLADA-VIP plot scores. Control and case groups were assigned codes 0 and 1, respectively.

[0251] Gene ontology analysis and functional enrichment: Disease and functional enrichment analyses were performed using genes identified with significantly differentially methylated FDR p-values ​​< 0.01, employing the Ingenuity Pathway Analysis (IPA) (Qiagen IPA) system. The IPA platform enables the systematic analysis of array data related to biological functions (Haddad et al., 2016). Gene networks were considered based on their interactions and roles in cardiac development and disease.

[0252] Results: Genome-wide methylation profiling was performed using cfF DNA extracted from mothers who gave birth to 12 CHD cases and 26 controls. There were no statistically significant differences in gestational age at sampling (p = 0.15). The mean (SD) gestational age at blood draw was 23 weeks and 5 days (case group) and 24 weeks and 6 days (control group) (p < 0.15). There were no significant differences in age range and gestational age at sampling between mothers giving birth to CHD infants and mothers giving birth to normal infants, with p values ​​of 0.32 and 0.15, respectively. Demographic details are provided in Table S1.

[0253] Table S1: Clinical and demographic characteristics of study participants*

[0254]

[0255]

[0256] W: Wilcoxon rank-sum test

[0257] *There was no difference in racial composition between the two groups.

[0258] Differential methylation analysis of cfF DNA identified 5918 CpG markers covering 4976 genes (FDR p-value ≤ 0.01). Table 1 provides the top 1000 important CpG sites and related genes with University of California, Santa Cruz (UCSC) gene symbols, along with statistical data. Of these markers, 130 CpGs were hypomethylated, while 870 were hypermethylated and associated with CHD. 53 hypomethylated and 486 hypermethylated markers showed differential methylation differences ≥ 10%. The three markers with methylation differences greater than 20% were cg06301252 (PTPRN2) with a 33.62% methylation change, cg02807450 (MTMR2) with a -21.15% methylation change, and cg12900404 (DOCK10) with a -20.13% methylation change. The 126 CpG sites showed an AUC ≥ 0.80, indicating excellent predictive accuracy for disease prediction.

[0259] AI prediction of methylation markers for cfF DNA CHDs. Significantly differentially methylated CpGs with a 5% difference and AUC > 0.70 were used for AI analysis using six prediction algorithms. Among them, the RF and SVM models were found to show the highest performance, with AUC (95% CI) = 0.98 (0.831–1) and 93.8% sensitivity and 93.2% specificity, and AUC (95% CI) = 0.97 (0.877–1) and 98% sensitivity and 94% specificity, followed by the DL model with AUC (95% CI) = 0.94 (0.840–1.0) and 93% sensitivity and 94% specificity (Table 2). High performance was also achieved for each of the other four platforms. Comparable predictive performance was achieved when demographic markers were considered together with CpG sites (Table 3), and high performance was also achieved when only markers meeting strict GWAS thresholds were considered (Table 4). High performance was achieved when demographic markers and CpG sites that met strict GWAS thresholds were considered (Table 5). Table 6 shows that high prediction accuracy was observed when only markers showing high levels of methylation change (e.g., 1.5-fold or greater) were used. The top 5 predicted markers for each model are provided in descending order (below each table) in each table.

[0260] Logistic Regression Analysis: A standard logistic regression analysis with 10-fold cross-validation was performed, and the labels were compared with AI predictions with a CpG of 5% difference and AUC > 0.70. On the training set, an AUC of 0.98 (0.98–0.97) (95% CI) with 100% sensitivity and 85% specificity was obtained. Subsequently, the trained logistic regression model with 10-fold cross-validation provided an AUC of 0.79 (0.61–0.98) (95% CI) with 83% sensitivity and 80% specificity on the test set. The logistic regression equation is as follows:

[0261] logit(P)=log(P / (1-P))=-5.469-2.698cg08230215-3.924cg04761177-2.478cg10259004-2.477cg06009031.

[0262] For a strict p-value of 5 × 10 -8 The CpG marker showed an AUC (95% CI) of 0.99 (0.92–0.99), with 90% sensitivity and 95% specificity in the training / discovery test. 10-fold cross-validation showed an AUC (95% CI) of 0.71 (0.50–0.92), with 75% sensitivity and 87% specificity, according to the following equation:

[0263] logit(P)=log(P / (1-P))=-2.583-5.115cg04761177-6.351cg23306063+1.126cg10259004-0.369cg23136742-2.917cg11196182.

[0264] Cluster analysis of differentially methylated targets in cfF DNA CHD: In hierarchical cluster analysis using individual β values ​​of significantly differentially methylated markers between cases and controls, well-separated clusters of hypomethylation and hypermethylation were observed between cases and controls. Figure 1A and Figure 1B The CpG tag ID is provided in the right column.

[0265] Disease and Functional Enrichment Analysis: Pathway analysis was performed to determine the biological validity of the statistical findings. The pathway analysis system revealed significant functional enrichment of disease and CHD-related genes. The top three cardiac development and disease functions were enriched in congenital heart disease (P-6.69E-03), cardiac hypertrophy (P-4.34E-03), and cardiac development (P-3.73E-05). Figure 2 The gene networks involved in the function of these diseases were depicted. Therefore, key genes involved in differential methylation appear to play a role in cardiac embryology and CHD development.

[0266] Discussion: A comprehensive methylation profiling analysis was performed using cell-free circulating cfF DNA from maternal blood in combination with AI predictions. This study showed significant differential methylation of 5918 CpG markers containing 4976 genes between cases and controls. Cf DNA is a promising source of biomarkers for prenatal diagnosis, being minimally invasive and slowly replacing other invasive tests such as amniocentesis or chorionic villus sampling (Nagy 2019). Epigenetic factors regulate gene expression and influence cardiac development during embryogenesis (Vallaster et al. 2012). 1278 hypomethylated genes tend to be overexpressed in cells (Klasic et al. 2016), while the remaining 4640 hypermethylated genes may show downregulated expression in cells (Razin & Kantor 2005). 539 CpGs showed methylation differences ≥10%, indicating biological relevance to gene expression. The study samples were separated based on PLS-DA analysis, and the differences between high-methylation and low-methylation markers were shown in the heatmap (Figure 1).

[0267] AI analysis was performed using six different algorithms, including SVM, GLM, PAM, RF, LDA, and DL. The SVM model provided the best predictions, with an AUC of 0.97 (98% sensitivity and 94% specificity). The top five predicted labels included cg04761177 (ATP2A1), cg21431091 (TMEM9), cg01263077 (MYO9B), cg09853933 (ATG2B; GSKIP), and cg27142059 (TRIM15). To compare with the AI ​​predictions, the labels were tested using conventional logistic regression analysis with 10-fold cross-validation, and cg08230215 (MAST3), cg04761177 (ATP2A1), cg10259004 (MY19), and cg06009031 (C7orf50) were found. A common marker was found in the prediction algorithms SVM, PAM, RF, LDA, DL, and logistic regression: cg04761177 (ATP2A1). cg04761177 was demethylated, exhibiting 9% methylation differential, with an AUC of 0.94 (CI 0.86–1.00) for CHD prediction. The gene ATP2A1 is also known as SERCA1. This gene regulates the electrical and systolic properties of the heart, and dysregulation of this gene is associated with various heart diseases, including congestive heart failure (Peters et al. 1997; Enniset et al. 2002).

[0268] From having a strict p-value of -5 × 10 -8 The second most common CpG marker predicted as an excellent marker by both AI algorithms and conventional logistic regression is cg10259004 (MY19). The MYL9 gene encodes the myosin light chain and is expressed in cardiac smooth muscle and involved in cardiac morphogenesis (England & Loughna 2013).

[0269] Functional enrichment of CpG-containing genes in disease- and CHD-associated cf DNA: Disease-based pathway enrichment analysis identified significantly differentially methylated genes currently known or predicted to be associated with cardiac hypertrophy, cardiac development, and congenital heart disease. This demonstrates the biological plausibility of the identified CHD-associated genes. All identified CpG methylation sites were located within promoters or gene regions and were associated with mechanisms of cardiac hypertrophy, cardiac development, and congenital heart disease, with significant p-values ​​<0.001. For example, some hypomethylated genes included HSPB11, POFUT1, NFATC4, DTNBP1, CFLAR, KCNH2, B3GAT3, and BMP4, while hypermethylated genes included BMP7, NOG, MAP2K2, FGF9, ADAM17, and MAPK3.

[0270] HSPB11 has been found to be associated with heart development and is highly expressed during late ventricular tissue development in zebrafish (Singh et al. 2016). POFUT1 is an essential component of Notch signaling and plays a crucial role in heart valve development, cardiac outflow tract formation, and ventricular septum formation (Penton et al. 2012). Mice lacking the protein POFUT1 die in mid-pregnancy and exhibit severe defects in angiogenesis and heart development (Shi & Stanley 2003). NFATC4 and DTNBP1 are involved in inducing human cardiac hypertrophy (Poirier et al. 2003; Rangrez et al. 2013). Reduced expression of the CFLAR gene has been shown to result in severe defects in cardiac trabeculae formation, myocardial thinning, cardiac lethality, and ventricular structure (Imanishi et al. 2000; Yeh et al. 2000; Lakhani et al. 2006; Ye et al. 2013). KCNH2 encodes a protein known as Kv11.1 (a potassium channel) that conducts potassium ions out of cardiomyocytes (Newton-Cheh et al. 2007; Park et al. 2013). A retrospective study in a large cohort investigated mutations in KCNH2 resulting in atrioventricular block, tetralogy of Fallot, coarctation, atrial septal defect, atrioventricular canal, bicuspid aortic valve, patent ductus arteriosus, tricuspid atresia, and ventricular septal defect (Ebrahim et al. 2017). Another gene, B3GAT3, plays an important role in proteoglycan biosynthesis (von Oettingen et al. 2014), and mutations in these proteoglycan chains result in serious congenital cardiac abnormalities, including bicuspid aortic valve, ventricular septal defect, and mitral valve prolapse (Baasanjav et al. 2011; Bloor et al. 2017). BMP4 is an important source for endocardial cushion development and normal outflow tract septation. In mice, loss of BMP4 function leads to ventricular septal defects, abnormal semilunar valve formation, abnormal cushion remodeling, persistent truncus arteriosus, and insufficient cardiac differentiation in developing epicardial cushions (McCulley et al., 2008).

[0271] Hypermethylated genes, such as BMP7, encode a secretory ligand for TGF-β, which plays a crucial regulatory role in the coronary vascular system, ventricular myocardial development, and compaction (Azhar et al., 2003). In mouse embryos, low BMP7 levels are associated with increased cardiovascular disease morbidity and mortality (Silverman et al., 2004) (Freedman et al., 2009). Another hypermethylated gene, NOG (noggin), interacts strongly with BMP2 and BMP4, but also with BMP7 (Choi et al., 2007). Notably, both BMP4 and BMP7, along with NOG, were differentially methylated in this study. Noggin knockout mice exhibited several abnormalities in cardiovascular development, which are a result of BMP signaling and are predicted to play a similar role in humans (Choi et al., 2007). MAP2K2 (mitogen-activated protein kinase kinase 2) is another hypermethylated gene, and early reports indicated its association with increased prevalence of cardiac hypertrophy through activation of the ERK pathway (Gillespie-Brown et al. 1995; Gallo et al. 2019). Another MAP kinase family gene, MAPK3, is involved in the regulation of meiosis, mitosis, and postmitotic function, and its dysregulation promotes cardiac hypertrophy (Bueno et al. 2000; Mutlak & Kehat 2015). The FGF9 gene is primarily expressed in the epicardium, where it maintains myocardial proliferation during mid-pregnancy cardiac development. Altered FGF9 expression is associated with significantly reduced cardioblast proliferation and ventricular hypoplasia (Lavine et al., 2005). FGF9 also functions as a paracrine signal during embryonic cardiac development, and its loss of function is associated with reduced cardioblast proliferation (Itoh et al.). The ADAM17 gene, containing deintegration and metalloproteinase domains, plays a crucial role in structural cardiac remodeling by altering cell surface matrix receptors, and loss of ADAM17 function leads to cardiac hypertrophy (Wang et al. 2009; Takayanagi et al. 2016). Altered ADAM17 expression is also a marker of ventricular remodeling, indicating its important role in the late stages of cardiac remodeling (Zheng et al. 2016).

[0272] This proof-of-concept study confirms the impact of DNA methylation on CHD and uses cfF DNA identification as a possible approach. The diagnostic accuracy predicted using various statistical methods and AI platforms helps prioritize novel genes for CHD prediction.

[0273] In summary, significant differentially methylated markers associated with CHD were identified in cfF DNA. AI algorithms predicted significant CHD-related markers, and disease enrichment analysis revealed important genes related to cardiac development and function. Developing minimally invasive methods for prenatal diagnosis is of clinical importance. Using cfF DNA in this study to understand the roles and regulation of the identified genes will further form the basis for understanding the molecular mechanisms of embryological development of normal and abnormal hearts in early pregnancy.

[0274] The foregoing subject matter is provided by way of illustration only and should not be construed as limiting. Various modifications and changes may be made to the subject matter described herein without following the exemplary embodiments and applications shown and described and without departing from the true spirit and scope of this disclosure as set forth in the appended claims.

[0275] All publications, patents, and patent applications cited in this specification are incorporated herein by reference in their entirety, as if each individual publication, patent, or patent application were specifically and individually indicated to be incorporated by reference. While the foregoing has been described according to various embodiments, those skilled in the art will understand that various modifications, substitutions, omissions, and alterations may be made without departing from its spirit.

[0276] References

[0277] Alakwaa FM, Chaudhary K. & Garmire LX(2018)Deep Learning AccuratelyPredicts Estrogen Receptor Status in Breast Cancer MetabolomicsData.JProteome Res 17,337-47.

[0278] Alberry M., Maddocks D., Jones M., Abdel Hadi M., Abdel-Fattah S., Avent N. & Soothill PW (2007) Free fetal DNA in maternal plasma in anembryonicpregnancies: confirmation that the origin is the trophoblast. Prenat Diagn 27,415-8.

[0279] Alfirevic Z.,Sundberg K.&Brigham S.(2003)Amniocentesis and chorionicvillus sampling for prenatal diagnosis.Cochrane Database Syst Rev,CD003252.

[0280] Azhar M.,Schultz Jel J.,Group I.,Dorn GW,2nd,Meneton P.,Molin DG,Gittenberger-de Groot AC&Doetschman T.(2003)Transforming growth factor betaine cardiovascular development and function.Cytokine Growth Factor Rev 14,391-407.

[0281] Baasanjav S.,A1-Gazali L.,Hashiguchi T.,Mizumoto S.,Fischer B.,HornD.,Seelow D.,Ali BR,Aziz SA,Langer R.,Saleh AA,Becker C.,Nurnberg G.,Cantagrel V.,Gleeson JG,Gomez D.,Michel JB,Stricker S.,Lindner TH,Nurnberg P.,Sugahara K.,Mundlos S.&Hoffmann K.(2011)Faulty initiation ofproteoglycan synthesis causes cardiac and joint defects.Am JHum Genet 89,15-27.

[0282] Bahado-Singh R.,Vishweswaraiah S.,Mishra NK,Guda C.&Radhakrishna U.(2019a)Placental DNA methylation changes for the detection of tetralogy ofFallot.Ultrasound Obstetrics Gynecol.

[0283] Bahado-Singh R.O.,Vishweswaraiah S.,Aydas B.,Mishra N.K.,Guda C.&Radhakrishna U.(2019b)Deep Learning / Artificial Intelligence and Blood-BasedDNA Epigenomic Prediction of Cerebral Palsy.Int JMol Sci 20.

[0284] Bahado-Singh R.O.,Wapner R.,Thom E.,Zachary J.,Platt L.,Mahoney M.J.,Johnson A.,Silver R.K.,Pergament E.,Filkins K.,Hogge W.A.,Wilson R.D.,JacksonL.G.,First Trimester Maternal Serum B.&Fetal Nuchal Translucency ScreeningStudy G.(2005)Elevated first-trimester nuchal translucency increases the riskof congenital heart defects.Am JObstet Gynecol 192,1357-61.

[0285] Bahado-Singh R.O.,Zaffra R.,Albayarak S.,Chelliah A.,Bolinjkar R.,Turkoglu O.&Radhakrishna U.(2016)Epigenetic markers for newborn congenitalheart defect(CHD).J Matern Fetal Neonatal Med 29,1881-7.

[0286] Bartak B.K.,Kalmar A.,Galamb O.,Wichmann B.,Nagy Z.B.,Tulassay Z.,Dank M.,Igaz P.&Molnar B.(2019)Blood Collection and Cell-Free DNA IsolationMethods Influence the Sensitivity of Liquid Biopsy Analysis for ColorectalCancer Detection.Pathol Oncol Res 25,915-23.

[0287] Bloor S.,Giri D.,Didi M.&Senniappan S.(2017)Novel Splicing Mutationin B3GAT3 Associated with Short Stature,GH Deficiency,Hypoglycaemia,Developmental Delay,and Multiple Congenital Anomalies.Case Rep Genet 2017,3941483.

[0288] Botto LD,Mulinare J,Erickson JD.Do multivitamin or folic acidsupplements reduce the risk for congenital heart defects?Evidence and gaps.AmJ Med GenetA 2003;121A:95-101)

[0289] Botto LD,Yang Q.5,10-Methylenetetrahydrofolate reductase genevariants and congenital anomalies:a HuGE review.Am JEpidemiol2000;151:862-77

[0290] Bueno O.F.,De Windt L.J.,Tymitz K.M.,Witt S.A.,Kimball T.R.,KlevitskyR.,Hewett T.E.,Jones S.P.,Lefer D.J.,Peng C.F.,Kitsis R.N.&Molkentin J.D.(2000)The MEK1-ERK1 / 2 signaling pathway promotes compensated cardiachypertrophy in transgenic mice.EMBO J 19,6341-50.

[0291] Burton G.J.&Jauniaux E.(2018)Development of the Human Placenta andFetal Heart:Synergic or Independent?Front Physiol 9,373.

[0292] Candel A.,Parmar V.,LeDell E.&Arora A.(2018)Deep Learning with H2O.

[0293] Choi M.,Stottmann R.W.,Yang Y.P.,Meyers E.N.&Klingensmith J.(2007)Thebone morphogenetic protein antagonist noggin regulates mammalian cardiacmorphogenesis.Circ Res 100,220-8.

[0294] Chong J.,Soufan O.,Li C.,Caraus I.,Li S.,Bourque G.,Wishart D.S.&XiaJ.(2018)MetaboAnalyst 4.0:towards more transparent and integrativemetabolomics analysis.Nucleic Acids Res46,W486-W94.

[0295] Czeizel AE,Merhala Z.Bread fortification with folic acid,vitamin B12,and vitamin B6 in Hungary.Lancet 1998;352:1225

[0296] Ebrahim M.A.,Williams M.R.,Shepard S.&Perry J.C.(2017)GenotypePositive Long QT Syndrome in Patients With Coexisting Congenital HeartDisease.Am J Cardiol 120,256-61.

[0297] England J.&Loughna S.(2013)Heavy and light roles:myosin in themorphogenesis of the heart.Cell Mol Life Sci 70,1221-39.

[0298] Ennis I.L.,LiR.A.,Murphy A.M.,Marban E.&Nuss H.B.(2002)Dual genetherapy with SERCA1and Kir2.1 abbreviates excitation without suppressingcontractility.J Clin Invest 109,393-400.

[0299] Freedman B.I.,Bowden D.W.,Ziegler J.T.,Langefeld C.D.,Lehtinen A.B.,Rudock M.E.,Lenchik L.,Hruska K.A.,Register T.C.&Carr J.J.(2009)Bonemorphogenetic protein7(BMP7)gene polymorphisms are associated with inverserelationships between vascular calcification and BMD:the Diabetes HeartStudy.JBone Miner Res 24,1719-27.

[0300] Gallo S.,Vitacolonna A.,Bonzano A.,Comoglio P.&Crepaldi T.(2019)ERK:AKey Player in the Pathophysiology of Cardiac Hypertrophy.Int J Mol Sci 20.

[0301] Gillespie-Brown J.,Fuller S.J.,Bogoyevitch M.A.,Cowley S.&Sugden P.H.(1995)The mitogen-activated protein kinase kinase MEK1 stimulates a patternof gene expression typical of the hypertrophic phenotype in rat ventricularcardiomyocytes.JBiol Chem270,28092-6.

[0302] Goldwaser T.&Klugman S.(2018)Cell-free DNA for the detection of fetalaneuploidy.Fertil Steril 109,195-200.

[0303] Grace M.R.,Hardisty E.,Dotters-Katz S.K.,Vora N.L.&Kuller J.A.(2016)Cell-Free DNA Screening:Complexities and Challenges of ClinicalImplementation.Obstet Gynecol Surv 71,477-87.

[0304] Gupta A.K.,Holzgreve W.,Huppertz B.,Malek A.,Schneider H.&Hahn S.(2004)Detection of fetal DNA and RNA in placenta-derived syncytiotrophoblastmicroparticles generated in vitro.Clin Chem 50,2187-90.

[0305] Haddad D.,Socci N.,Chen C.H.,Chen N.G.,Zhang Q.,Carpenter S.G.,MittraA.,Szalay A.A.&Fong Y.(2016)Molecular network,pathway,and functional analysisof time-dependent gene changes associated with pancreatic cancersusceptibility to oncolytic vaccinia virotherapy.Mol Ther Oncolytics3,16008.

[0306] Huang J.H.,Xie H.L.,Yan J.,Lu H.M.,Xu Q.S.&Liang Y.Z.(2013)Usingrandom forest to classify T-cell epitopes based on amino acid properties andmolecular features.Anal Chim Acta 804,70-5.

[0307] Imanishi T.,Murry C.E.,Reinecke H.,Hano T.,Nishio I.,Liles W.C.,Hofsta L.,Kim K.,O′Brien K.D.,Schwartz S.M.&Han D.K.(2000)Cellular FLIP isexpressed in cardiomyocytes and down-regulated in TUNEL-positive graftedcardiac tissues.Cardiovasc Res 48,101-10.

[0308] Ionescu-Ittu R,Marelli AJ,Mackie AS et al.Prevalence of severecongenital heart disease after folic acid fortification of grain products:time trend analysis in Quebec,Canada.BMJ 2009;338:b1673doi:10.1136 / bmj.b1673

[0309] Itoh N.,Ohta H.,Nakayama Y.&Konishi M.(2016)Roles of FGF Signals inHeart Development,Health,and Disease.Front Cell Dev Biol 4,110.

[0310] Karunamuni G,Sheehan MM,Doughman YQ et al.Supplementation with themethyl donor betaine prevents congenital defects induced by prenatal alcoholexposure.Alcohol Clin Exp Res 2017;41:1917-27

[0311] Ke W.L.,Zhao W.H.&Wang X.Y.(2015)Detection of fetal cell-free DNA inmaternal plasma for Down syndrome,Edward syndrome and Patau syndrome of highrisk fetus.Int J Clin Exp Med 8,9525-30.

[0312] Klasic M.,Kristic J.,Korac P.,Horvat T.,Markulin D.,Vojta A.,ReidingK.R.,Wuhrer M.,Lauc G.&Zoldos V.(2016)DNA hypomethylation upregulatesexpression of the MGAT3 gene in HepG2cells and leads to changes in N-glycosylation of secreted glycoproteins.Sci Rep 6,24363.

[0313] Kriaucionis S,Heintz N.The nuclear DNA base 5-hydroxymethylcytosineis present in Purkinje neurons and the brain.Science 2009;324:929-30

[0314] Lakhani S.A.,Masud A.,Kuida K.,Porter G.A.,Jr.,Booth C.J.,Mehal W.Z.,Inayat I.&Flavell R.A.(2006)Caspases 3 and 7:key mediators of mitochondrialevents of apoptosis.Science 311,847-51.

[0315] Lavine K.J.,YuK.,White A.C.,Zhang X.,Smith C.,Partanen J.&Omitz D.M.(2005)Endocardial and epicardial derived FGF signals regulate myocardialproliferation and differentiation in vivo.Dev Cell 8,85-95.

[0316] Lehmann-Werman R,Neiman D,Zemmour H et al.Identification of tissue-specific cell death using methylation patterns of circulating DNA.ProcNatlAcadSci U.S.A.2016:113(13):E1826-34

[0317] Liland K.H.(2011)Multivariate methods in metabolomics-from pre-processing to dimension reduction and statistical analysis.TrAC Trends inAnalytical Chemistry 30,827-41.

[0318] Liu Y,Balaraman Y,Wang G et al.Alcohol exposure alters DNAmethylation profiles in mouse embryos at early neurulation.Epigenetics 2009;4:500-11

[0319] Mahadevan S.,Shah S.L.,Marrie T.J.&Slupsky C.M.(2008)Analysisofmetabolomic data using support vector machines.Anal Chem 80,7562-70.

[0320] Mao B,Qui J,Zhao N et al.Maternal folic acid supplementation anddietary folate intake and congenital heart defects.PLoS One 2017 November 16;12:e0187996.doi:10.1371 / journal.pone.0187996

[0321] Maslen C.L.(2018)Recent Advances in Placenta-Heart Interactions.FrontPhysiol9,735.

[0322] McCulley D.J.,Kang J.O.,Martin J.F.&Black B.L.(2008)BMP4 is requiredin the anterior heart field and its derivatives for endocardial cushionremodeling,outflow tract septation,and semilunar valve development.Dev Dyn237,3200-9.

[0323] Mutlak M.&Kehat I.(2015)Extracellular signal-regulated kinases 1 / 2 asregulators of cardiac hypertrophy.FrontPharmacol 6,149.

[0324] Nagy B.(2019)Cell-free nucleic acids in prenatal diagnosis andpregnancy-associated diseases.EJIFCC 30,215-23.

[0325] Nester CE,Ottaviano R,Reddington J et al.Tissue type is a majormodifier of the 5-hydroxymethylcytosine content of human genes.Genome Res2012;22:467-77

[0326] Newton-Cheh C.,Guo C.Y.,Larson M.G.,Musone S.L.,Surti A.,CamargoA.L.,Drake J.A.,Benjamin E.J.,Levy D.,D′Agostino R.B.,Sr.,Hirschhorn J.N.&O′Donnell C J.(2007)Common genetic variation in KCNH2 is associated with QTinterval duration:the Framingham Heart Study.Circulation 116,1128-36.

[0327] O′Meara C.C.&Lee R.T.(2015)Peering Into the Cardiomyocyte NuclearEpigenetic State.Circ Res 117,392-4.

[0328] Park J.K.,Oh Y.S.,Choi J.H.&Yoon S.K.(2013)Single nucleotide deletionmutation of KCNH2gene is responsible for LQT syndrome in a 3-generationKorean family.JKorean Med Sci 28,1388-93.

[0329] Penton A.L.,Leonard L.D.&Spinner N.B.(2012)Notch signaling in humandevelopment and disease.Semin Cell Dev Biol23,450-7.

[0330] Peters D.G.,Mitchell H.L.,McCune S.A.,Park S.,Williams J.H.&KandarianS.C.(1997)Skeletal muscle sarcoplasmic reticulum Ca(2+)-ATPase geneexpression in congestive heart failure.Circ Res 81,703-10.

[0331] Poirier O.,Nicaud V.,McDonagh T.,Dargie HJ,Desnos M.,Dorent R.,Roizes G.,Schwartz K.,Tiret L.,Komajda M.&Cambien F.(2003)Polymorphisms ofgenes of the cardiac calcineurin pathway and cardiac hypertrophy.Eur J HumGenet 11,659–64.

[0332] Radhakrishna U.,Albayrak S.,Alpay-Savasan Z.,Zeb A.,Turkoglu O.,Sobolewski P.&Bahado-Singh RO(2016)Genome-Wide DNA Methylation Analysis andEpigenetic Variations Associated with Congenital Aortic Valve Stenosis(AVS).PLoS One 11,e0

[0333] Radhakrishna U.,Albayrak S.,Zafra R.,Baraa A.,Vishweswaraiah S.,Veerappa AM,Mahishi D.,Saiyed N.,Mishra NK,Guda C.,Ali-Fehmi R.&Bahado-Singh RO(2019)Placental epigenetics for evaluation of fetal congenitalheart defects:Ventricular Septal Defect(VSD).PLoS One 14,e0200229.

[0334] Radhakrishna U.,Vishweswaraiah S.,Veerappa A.M.,Zafra R.,Albayrak S.,Sitharam P.H.,Saiyed N.M.,Mishra N.K.,Guda C.&Bahado-Singh R.(2018)Newbornblood DNA epigenetic variations and signaling pathway genes associated withTetralogy of Fallot(TOF).PLoS One 13,e0203893.

[0335] Rafaeli-Yehudai T.,Imterat M.,Douvdevani A.,Tirosh D.,Benshalom-Tirosh N.,Mastrolia S.A.,Beer-Weisel R.,Klaitman V.,Riff R.,Greenbaum S.,Alioshin A.,Rodavsky Hanegbi G.,Loverro G.,Catalano M.R.&Erez O.(2018)Maternal total cell-free DNA in preeclampsia and fetal growth restriction:Evidence of differences in maternal response to abnormal implantation.PLoSOne 13,e0200360.

[0336] Rafi I,Hill M,Hayward Jet al.Non-invasive prenatal testing:use ofcell-free fetal DNA in Down syndrome screening.Br J Gen Pract 2017;67:298-299

[0337] Ramakrishnan A.,Lee L.J.,Mitchell L.E.&Agopian A.J.(2015)MaternalHypertension During Pregnancy and the Risk of Congenital Heart Defects inOffspring:A Systematic Review and Meta-analysis.Pediatr Cardiol36,1442-51.

[0338] Rangrez A.Y.,Bernt A.,Poyanmehr R.,Harazin V.,Boomgaarden I.,Kuhn C.,Rohrbeck A.,Frank D.&Frey N.(2013)Dysbindin is a potent inducer of RhoA-SRF-mediated cardiomyocyte hypertrophy.J Cell Biol203,643-56.

[0339] Razin A.&Kantor B.(2005)DNA methylation in epigenetic control of geneexpression.Prog Mol Subcell Biol38,151-67.

[0340] Sheinerman K.S.,Toledo J.B.,Tsivinsky V.G.,Irwin D.,Grossman M.,Weintraub D.,Hurtig H.I.,Chen-Plotkin A.,Wolk D.A.,McCluskey L.F.,Elman L.B.,Trojanowski J.Q.&Umansky S.R.(2017)Circulating brain-enriched microRNAs asnovel biomarkers for detection and differentiation of neurodegenerativediseases.Alzheimers Res Ther 9,89.

[0341] Shi S.&Stanley P.(2003)Protein O-fucosyltransferase 1 is an essentialcomponent of Notch signaling pathways.Proc Natl Acad Sci U S A 100,5234-9.

[0342] Silverman S.L.,Delmas P.D.,Kulkarni P.M.,Stock J.L.,Wong M.&PlouffeL.,Jr.(2004)Comparison of fracture,cardiovascular event,and breast cancerrates at 3 years in postmenopausal women with osteoporosis.JAm Geriatr Soc52,1543-8.

[0343] Singh A.R.,Sivadas A.,Sabharwal A.,Vellarikal S.K.,Jayarajan R.,VermaA.,Kapoor S.,Joshi A.,Scaria V.&Sivasubbu S.(2016)Chamber Specific GeneExpression Landscape of the Zebrafish Heart.PLoS One 11,e0147823.

[0344] Song CX,Szulwach KE,Fu Y et al.Selective chemical labeling revealsthe genome-wide distribution of 5-hydroxymethylcytosine.Nat Biotechol2011;29:68-72

[0345] Stewart C.M.,Kothari P.D.,Mouliere F.,Mair R.,Somnay S.,Benayed R.,Zehir A.,Weigelt B.,Dawson S.J.,Arcila M.E.,Berger M.F.&Tsui D.W.(2018)Thevalue of cell-free DNA for molecular pathology.JPathol244,616-27.

[0346] Taglauer E.S.,Wilkins-Haug L.&Bianchi D.W.(2014)Review:cell-freefetal DNA in the maternal circulation as an indication of placental healthand disease.Placenta 35 Suppl,S64-8.

[0347] Tahiliani M,Koh KP,Shen Y et al.Conversion of 5-methylcytosine to 5-hydroxymethylcytosine in mammalian DNA by MLL partner TET1.Science 2009;324:930-5

[0348] Takayanagi T.,Forrester S.J.,Kawai T.,Obama T.,Tsuji T.,Elliott K.J.,Nuti E.,Rossello A.,Kwok H.F.,Scalia R.,Rizzo V.&Eguchi S.(2016)VascularADAM17 as a Novel Therapeutic Target in Mediating Cardiovascular Hypertrophyand Perivascular Fibrosis Induced by Angiotensin II.Hypertension 68,949-55.

[0349] Tan CH,Denny CH,Cheal NE et al.Alcohol use and binge drinking amongwomen of childbearing age-United States,2011-2013.MMWR Morb Mortal Wkly Rep2015;64:1042-6

[0350] Tjoa M.L.,Cindrova-Davies T.,Spasic-Boskovic O.,Bianchi D.W.&BurtonG.J.(2006)Trophoblastic oxidative stress and the release of cell-free feto-placental DNA.Am JPathol 169,400-4.

[0351] Vallaster M.,Vallaster C.D.&Wu S.M.(2012)Epigenetic mechanisms incardiac development and disease.Acta Biochim Biophys Sin(Shanghai)44,92-102.

[0352] van Gelder M.M.,Van Bennekom C.M.,Louik C.,Werler M.M.,Roeleveld N.&Mitchell A.A.(2015)Maternal hypertensive disorders,antihypertensivemedication use,and the risk of birth defects:a case-control study.BJOG 122,1002-9.

[0353] von Oettingen J.E.,Tan W.H.&Dauber A.(2014)Skeletal dysplasia,globaldevelopmental delay,and multiple congenital anomalies in a 5-year-old boy-report of the second family with B3GAT3mutation and expansion of thephenotype.Am J Med Genet A 164A,1580-6.

[0354] Wang X.,Oka T.,Chow F.L.,Cooper S.B.,Odenbach J.,Lopaschuk G.D.,Kassiri Z.&Fernandez-Patron C.(2009)Tumor necrosis factor-alpha-convertingenzyme is a key regulator of agonist-induced cardiac hypertrophy andfibrosis.Hypertension 54,575-82.

[0355] Wataganara T.,Gratacos E.,Jani J.,Becker J.,Lewi L.,Sullivan L.M.,Bianchi D.W.&Deprest J.A.(2005)Persistent elevation of cell-free fetal DNAlevels in maternal plasma after selective laser coagulation of chorionicplate anastomoses in severe midgestational twin-twin transfusion syndrome.AmJ Obstet Gynecol192,604-9.

[0356] Ye J.,Llorian M.,Cardona M.,Rongvaux A.,Moubarak R.S.,Comella J.X.,Bassel-Duby R.,Flavell R.A.,Olson E.N.,Smith C.W.&Sanchis D.(2013)A pathwayinvolving HDAC5,cFLIP and caspases regulates expression of the splicingregulator polypyrimidine tract binding protein in the heart.J Cell Sci 126,1682-91.

[0357] Yeh W.C.,Itie A.,Elia A.J.,Ng M.,Shu H.B.,Wakeham A.,Mirtsos C.,Suzuki N.,Bonnard M.,Goeddel D.V.&Mak T.W.(2000)Requirement for Casper(c-FLIP)in regulation of death receptor-induced apoptosis and embryonicdevelopment.Immunity 12,633-42.

[0358] Zeisel SH,da Costa KA.Choline:an essential nutrient for publichealth.Nut Rev 2009;67:615-23

[0359] Zheng D.Y.,Zhao J.,Yang J.M.,Wang M.&Zhang X.T.(2016)Enhanced ADAM17expression is associated with cardiac remodeling in rats with acutemyocardial infarction.Life Sci 151,61-9.

[0360] Zolotukhina T.V.,Shilova N.V.&Voskoboeva E.Y.(2005)Analysis of cell-free fetal DNA in plasma and serum ofpreguant women.JHistochem Cytochem 53,297-9.

Claims

1. An electronic device, comprising: A memory, on which instructions are stored; processor, The instructions, when executed by the processor, perform the following steps: The frequency or percentage of cytosine methylation at one or more sites throughout the genome is determined by measuring a biological sample containing cell-free nucleic acid obtained from a pregnant woman carrying a subject, wherein the subject is an embryo or fetus, and wherein, in determining the frequency or percentage of cytosine methylation at one site throughout the genome, the one site is cg04761177, or wherein, in determining the frequency or percentage of cytosine methylation at multiple sites throughout the genome, the multiple sites include cg04761177 and one or more sites from Tables 1 to 6 other than cg04761177; and By comparing the cytosine methylation level of the sample with that of the control sample, congenital heart defects in the subject can be predicted or diagnosed.

2. The electronic device according to claim 1, wherein, The prediction or diagnosis also includes the use of artificial intelligence technology.

3. The electronic device according to claim 1 or 2, wherein, The prediction or diagnosis also includes the use of artificial intelligence techniques, including one or more of the following machine learning algorithms: random forest, support vector machine, linear discriminant analysis, microarray analysis prediction, generalized linear model, or deep learning.

4. The electronic device according to claim 1 or 2, wherein, The prediction or diagnosis also includes: calculating the risk that the subject will develop a congenital heart defect.

5. The electronic device according to claim 1 or 2, wherein, The control samples were obtained from one or more normal patients or from one or more patients diagnosed with congenital heart defects.

6. The electronic device according to claim 1 or 2, wherein, The biological sample contains bodily fluids.

7. The electronic device according to claim 1 or 2, wherein, The biological samples include blood, plasma, serum, urine, saliva, sputum, sweat, tears, genital secretions including cervical secretions, amniotic fluid, placental tissue, and umbilical cord blood obtained at birth.

8. The electronic device according to claim 1 or 2, wherein, The cell-free nucleic acid includes fetal cell-free nucleic acid.

9. The electronic device according to claim 1 or 2, wherein, The biological sample contains cell-free fetal nucleic acids from early, mid, or late pregnancy.

10. The electronic device according to claim 1 or 2, wherein, The free nucleic acid contains DNA.

11. The electronic device according to claim 1 or 2, wherein, In determining the frequency or percentage of cytosine methylation at multiple sites throughout the genome, the multiple sites also include at least one, at least two, at least three, at least four, at least five, at least six, at least seven, at least eight, or at least nine sites from Tables 1 to 6, excluding cg04761177.

12. The electronic device according to claim 1 or 2, wherein, The one or more sites contain an AUC greater than 0.70, 0.75, 0.80, 0.85, 0.90, 0.91, 0.92, 0.93, 0.94, 0.95, 0.96 or 0.97, with a 95% CI.

13. The electronic device according to claim 1 or 2, wherein, The assays are based on bisulfite methylation assays or whole-genome methylation assays.

14. The electronic device according to claim 1 or 2, wherein, In determining the frequency or percentage of cytosine methylation at multiple sites throughout the genome, the sites from Tables 1 to 6 other than cg04761177 are cg06301252, cg02807450, or cg12900404.

15. The electronic device according to claim 1 or 2, wherein, In determining the frequency or percentage of cytosine methylation at multiple sites throughout the genome, the sites from Tables 1 to 6 other than cg04761177 are cg21431091, cg01263077, cg09853933, cg27142059, cg16551159, cg14957943, cg06978680, or cg12592721.

Images