Advertisement

Cell-free DNA in maternal blood and artificial intelligence: accurate prenatal detection of fetal congenital heart defects

Open AccessPublished:August 07, 2022DOI:https://doi.org/10.1016/j.ajog.2022.07.062

      Background

      DNA cytosine nucleotide methylation (epigenomics and epigenetics) is an important mechanism for controlling gene expression in cardiac development. Combined artificial intelligence and whole-genome epigenomic analysis of circulating cell-free DNA in maternal blood has the potential for the detection of fetal congenital heart defects.

      Objective

      This study aimed to use genome-wide DNA cytosine methylation and artificial intelligence analyses of circulating cell-free DNA for the minimally invasive detection of fetal congenital heart defects.

      Study Design

      In this prospective study, whole-genome cytosine nucleotide methylation analysis was performed on circulating cell-free DNA using the Illumina Infinium MethylationEPIC BeadChip array. Multiple artificial intelligence approaches were evaluated for the detection of congenital hearts. The Ingenuity Pathway Analysis program was used to identify gene pathways that were epigenetically altered and important in congenital heart defect pathogenesis to further elucidate the pathogenesis of isolated congenital heart defects.

      Results

      There were 12 cases of isolated nonsyndromic congenital heart defects and 26 matched controls. A total of 5918 cytosine nucleotides involving 4976 genes had significantly altered methylation, that is, a P value of <.05 along with ≥5% whole-genome cytosine nucleotide methylation difference, in congenital heart defect cases vs controls. Artificial intelligence analysis of the methylation data achieved excellent congenital heart defect predictive accuracy (areas under the receiver operating characteristic curve, ≥0.92). For example, an artificial intelligence model using a combination of 5 whole-genome cytosine nucleotide markers achieved an area under the receiver operating characteristic curve of 0.97 (95% confidence interval, 0.87–1.0) with 98% sensitivity and 94% specificity. We found epigenetic changes in genes and gene pathways involved in the following important cardiac developmental processes: “cardiovascular system development and function,” “cardiac hypertrophy,” “congenital heart anomaly,” and “cardiovascular disease.” This lends biologic plausibility to our findings.

      Conclusion

      This study reported the feasibility of minimally invasive detection of fetal congenital heart defect using artificial intelligence and DNA methylation analysis of circulating cell-free DNA for the prediction of fetal congenital heart defect. Furthermore, the findings supported an important role of epigenetic changes in congenital heart defect development.

      Key words

      Introduction

      Precision medicine is a National Institutes of Health initiative that combines genetic and other advanced analytical techniques, such as artificial intelligence (AI), for elucidating the mechanisms, diagnosis, and development of targeted treatment for complex diseases.
      • Baek S.H.
      Challenges and future in precision cardiovascular medicine.
      The potential effect of these approaches in improving obstetrical outcomes has recently been addressed.
      • Sadovsky Y.
      • Mesiano S.
      • Burton G.J.
      • et al.
      Advancing human health in the decade ahead: pregnancy as a key window for discovery: a Burroughs Wellcome Fund Pregnancy Think Tank.
      Although there has been significant interest in “precision cardiovascular medicine,”
      • Antman E.M.
      • Loscalzo J.
      Precision medicine in cardiology.
      ,
      • Semsarian C.
      • Ingles J.
      • Ross S.B.
      • Dunwoodie S.L.
      • Bagnall R.D.
      • Kovacic J.C.
      Precision medicine in cardiovascular disease: genetics and impact on phenotypes: JACC Focus Seminar 1/5.
      little attention has been paid to fetal cardiology.

      Why was this study conducted?

      Although prenatal detection of fetal congenital heart defects (CHD) can significantly improve pediatric morbidity and mortality, ultrasound detection continues to have significant diagnostic limitations. We evaluated the feasibility of a minimally invasive approach for the detection of fetal CHD using DNA methylation (“epigenomic”) analysis of circulating cell-free DNA (cfDNA) in maternal blood.

      Key findings

      Using DNA methylation (“epigenomic) analysis of cytosine nucleotides (CpGs) in circulating cfDNA and artificial intelligence (AI), extensive epigenetic changes in genes involved in cardiac development and CHD were found. We achieved consistently accurate prenatal detection of fetal CHD. For example, a combination of 5 cytosine DNA markers achieved an area under the receiver operating characteristic curve of 0.97 (95% confidence interval, 0.87–1.0) with 98% sensitivity and 94% specificity for the detection of CHD.

      What does this add to what is known?

      This study reported using minimally invasive testing based on circulating cfDNA, epigenomic, and AI analyses in pregnancy for the accurate detection of isolated fetal CHD.
      Congenital heart defect (CHD), of which 80% of cases are nonsyndromic,
      • Blue G.M.
      • Kirk E.P.
      • Sholler G.F.
      • Harvey R.P.
      • Winlaw D.S.
      Congenital heart disease: current knowledge about causes and inheritance.
      remains a leading cause of pediatric morbidity and mortality.
      • van der Linde D.
      • Konings E.E.
      • Slager M.A.
      • et al.
      Birth prevalence of congenital heart disease worldwide: a systematic review and meta-analysis.
      Prenatal diagnosis is known to reduce the risk of newborn death in critical CHDs.
      • Holland B.J.
      • Myers J.A.
      • Woods Jr., C.R.
      Prenatal diagnosis of critical congenital heart disease reduces risk of death from cardiovascular compromise prior to planned neonatal cardiac surgery: a meta-analysis.
      However, population-based studies continue to show significant limitations of ultrasound screening for CHD
      • Olney R.S.
      • Ailes E.C.
      • Sontag M.K.
      Detection of critical congenital heart defects: review of contributions from prenatal and newborn screening.
      with close to half the cases being detected prenatally.
      • Combs C.A.
      • Hameed A.B.
      • Friedman A.M.
      • Hoskins I.A.
      Patient SafetyQuality Committee; Society for Maternal-Fetal Medicine
      Special statement: proposed quality metrics to assess accuracy of prenatal detection of congenital heart defects.
      The subject is of great importance to practicing maternal-fetal medicine and pediatric cardiology specialists as increased accuracy can result in improved CHD outcomes in infants and children.
      • Li Y.F.
      • Zhou K.Y.
      • Fang J.
      • Wang C.
      • Hua Y.M.
      • Mu D.Z.
      Efficacy of prenatal diagnosis of major congenital heart disease on perinatal management and perioperative mortality: a meta-analysis.
      Despite the above, a substantial percentage of CHDs remain undetected prenatally. An overall detection rate of 59.7% was achieved; however, for isolated CHD, the subject of this report, this fell to 44.2%.
      • van Velzen C.L.
      • Clur S.A.
      • Rijlaarsdam M.E.
      • et al.
      Prenatal detection of congenital heart disease--results of a national screening programme.
      Furthermore, newborn pulse oximetry, developed to counter the low prenatal detection rates, misses 1 of 6 newborn cases of critical CHD,
      PennState
      Using artificial intelligence to detect discrimination.
      leading to preventable deaths or severe morbidities. However, there are multiple barriers to improving the sonographic detection of CHD.
      • Pinto N.M.
      • Keenan H.T.
      • Minich L.L.
      • Puchalski M.D.
      • Heywood M.
      • Botto L.D.
      Barriers to prenatal detection of congenital heart disease: a population-based study.
      This realization has fueled efforts to improve the quality
      • Combs C.A.
      • Hameed A.B.
      • Friedman A.M.
      • Hoskins I.A.
      Patient SafetyQuality Committee; Society for Maternal-Fetal Medicine
      Special statement: proposed quality metrics to assess accuracy of prenatal detection of congenital heart defects.
      and accuracy of prenatal CHD diagnosis. The need for accurate prenatal detection takes on greater importance as data indicate a link between fetal CHD development and increasingly common prenatal risk factors, such as prepregnancy and gestational diabetes mellitus,
      • Tinker S.C.
      • Gilboa S.M.
      • Moore C.A.
      • et al.
      Specific birth defects in pregnancies of women with diabetes: National Birth Defects Prevention Study, 1997-2011.
      chronic hypertension,
      • Bateman B.T.
      • Huybrechts K.F.
      • Fischer M.A.
      • et al.
      Chronic hypertension in pregnancy and the risk of congenital malformations: a cohort study.
      obesity,
      • Cai G.J.
      • Sun X.X.
      • Zhang L.
      • Hong Q.
      Association between maternal body mass index and congenital heart defects in offspring: a systematic review.
      and even febrile illnesses.
      • Botto L.D.
      • Panichello J.D.
      • Browne M.L.
      • et al.
      Congenital heart defects after maternal fever.
      A priori high-risk status and ultrasound examination performed in high-risk centers are factors that are significantly associated with improved detection of CHD.
      • Pinto N.M.
      • Keenan H.T.
      • Minich L.L.
      • Puchalski M.D.
      • Heywood M.
      • Botto L.D.
      Barriers to prenatal detection of congenital heart disease: a population-based study.
      Strategies, such as repeated sonographic evaluation, may improve overall detection
      • Byrne J.J.
      • Morgan J.L.
      • Twickler D.M.
      • McIntire D.D.
      • Dashe J.S.
      Utility of follow-up standard sonography for fetal anomaly detection.
      ; however, this may not be cost-effective as most CHDs occur in low-risk pregnancies.
      In contrast, great progress has been achieved with the introduction of prenatal molecular testing for Down syndrome detection,
      • Porreco R.P.
      • Garite T.J.
      • Maurel K.
      • et al.
      Noninvasive prenatal screening for fetal trisomies 21, 18, 13 and the common sex chromosome aneuploidies from maternal blood using massively parallel genomic sequencing of DNA.
      a significantly less common disorder than CHD. Therefore, it seems reasonable to expect that with the use of accurate molecular markers, high-risk CHD pregnancies can be identified with greater accuracy, pooled, and referred to expert centers for fetal and subsequent newborn echocardiogram examination, thus improving early detection of CHD.
      Epigenetics is a common mechanism for altering gene expression in the absence of gene mutations and plays a fundamental role in heart and CHD
      • Martinez S.R.
      • Gay M.S.
      • Zhang L.
      Epigenetic mechanisms in heart development and disease.
      development. It is the mechanism by which the external environment, for example, cigarette smoking,
      • Zong D.
      • Liu X.
      • Li J.
      • Ouyang R.
      • Chen P.
      The role of cigarette smoke-induced epigenetic alterations in inflammation.
      a known risk factor for CHD, interacts with the genome to cause disease. DNA cytosine nucleotide (so-called CpG) methylation is the most extensively studied epigenetic mechanism. The placenta plays a crucial role in heart development, and measurement of placental DNA methylation was previously demonstrated to accurately predict common CHDs.
      • Radhakrishna U.
      • Albayrak S.
      • Zafra R.
      • et al.
      Placental epigenetics for evaluation of fetal congenital heart defects: ventricular septal defect (VSD).
      AI is a branch of computer sciences in which computer systems are able to perform high-order functions akin to “reasoning,” learning, and knowledge retention
      • Krittanawong C.
      • Zhang H.
      • Wang Z.
      • Aydar M.
      • Kitai T.
      Artificial intelligence in precision cardiovascular medicine.
      previously requiring human intelligence. One such application is the ability to identify complex patterns in data that permit accurate discrimination or predict group status. Thus, AI has been shown to be superior to conventional statistical approaches
      • Mazaki J.
      • Katsumata K.
      • Ohno Y.
      • et al.
      A novel prediction model for colon cancer recurrence using auto-artificial intelligence.
      for disease detection. There was an early interest in applying AI techniques to obstetrics and gynecology
      • Mango L.J.
      Reducing false negatives in clinical practice: the role of neural network technology.
      research.
      Using cytosine methylation and AI analyses of circulating cell-free DNA (cfDNA) in maternal blood, to which the placenta makes a significant contribution, we sought to investigate epigenetic pathogenesis of isolated, nonsyndromic CHD. Furthermore, we evaluated the feasibility of minimally invasive detection of isolated nonsyndromic CHD.

      Materials and Methods

      Study design

      The study cases were collected prospectively based on prenatal ultrasound suspicion for or the diagnosis of isolated CHD in the second or third trimester of pregnancy. The newborn or pediatric ECHO was the gold standard for final CHD diagnosis and study inclusion. Only isolated CHD cases were included with a negative microarray or karyotype. The exclusion criteria used were multiple pregnancies, postnatal extracardiac gross structural anomalies or suspicion or diagnosis of a chromosomal or genetic abnormality or a genetic syndrome, and finally suspected CHD cases without postnatal cardiac ECHO. Controls were those undergoing prenatal ultrasound with a negative cardiac examination or with a newborn clinical examination or ECHO read as negative. The maternal age of the control was within 2 years of the index case and the gestational age at sampling within 2 weeks of the index case. The protocol was approved by the Human Investigation Committee of William Beaumont Hospital, Royal Oak, Michigan (institutional review board protocol number 2017-145). All the study subjects provided previous approval in the form of written consent.

      Sample collection method for cell-free DNA extraction

      Circulating cfDNA was extracted from maternal plasma. A significant percentage of cfDNA in pregnancy is of placental (“fetal”) origin.
      • Taglauer E.S.
      • Wilkins-Haug L.
      • Bianchi D.W.
      Review: cell-free fetal DNA in the maternal circulation as an indication of placental health and disease.
      The blood was drawn directly into Streck Cell-Free DNA BCT tubes, to ensure optimal quality of the cfDNA.
      • Barták B.K.
      • Kalmár A.
      • Galamb O.
      • et al.
      Blood collection and cell-free DNA isolation methods influence the sensitivity of liquid biopsy analysis for colorectal cancer detection.
      The preservatives in the Streck Cell-Free DNA BCT tubes (Streck, La Vista, Nebraska, USA) coat blood cells and avoid hemolysis, and the release of additional leukocyte genomic DNA significantly reduces the contamination of the circulating cfDNA with maternal DNA. The cfDNA extraction, methylation array processing, and statistical methods are provided in the Supplemental Materials and Methods.

      Artificial intelligence analysis

      We have extensively described the approaches used for the AI and machine learning analyses.
      • Bahado-Singh R.O.
      • Vishweswaraiah S.
      • Aydas B.
      • Mishra N.K.
      • Guda C.
      • Radhakrishna U.
      Deep learning/artificial intelligence and blood-based DNA epigenomic prediction of cerebral palsy.
      We briefly summarized those previously published descriptions. To generate the prediction models, 10-fold cross-validation
      • Bahado-Singh R.O.
      • Vishweswaraiah S.
      • Aydas B.
      • et al.
      Artificial intelligence and leukocyte epigenomics: evaluation and prediction of late-onset Alzheimer’s disease.
      with subjects serially divided into 2 groups, training, and test groups, each containing CHD cases and controls, was performed. The training group used to develop prediction models consisted of 80% of the study subjects with 20% in the test group. Cross-validation is commonly used when analyzing smaller datasets to minimize the so-called overfitting or inflated sensitivity and specificity values. Prediction models were generated from the training group, and the performance of these same models was determined for the independent test or validation group. A total of 6 different AI platforms or approaches were used in each analysis to determine the CHD detection performance. These AI techniques were random forest (RF), support vector machine (SVM), linear discriminant analysis, prediction analysis for microarrays, generalized linear model, and deep learning (DL). DL is an advanced form of AI and is increasingly being used in the analysis of biologic “big” data, such as genomics. These platforms are discussed in more detail in the Supplemental Materials and Methods.
      Regular logistic regression analyses were also performed for comparison. Finally, hierarchical cluster analysis
      • Bahado-Singh R.O.
      • Vishweswaraiah S.
      • Aydas B.
      • Radhakrishna U.
      Placental DNA methylation changes and the early prediction of autism in full-term newborns.
      and orthogonal partial least squares discriminant analysis (OPLS-DA) plots were generated to visually display the separation of CHD and control groups with the CpG methylation markers. Details are provided in the Supplemental Materials and Methods.

      Disease and functional enrichment analysis

      Ingenuity Pathway Analysis (IPA) systems (https://digitalinsights.qiagen.com/IPA)
      • Krämer A.
      • Green J.
      • Pollard Jr., J.
      • Tugendreich S.
      Causal analysis approaches in Ingenuity Pathway Analysis.
      were used to identify significant gene and gene pathways and biologic and disease pathways that were altered in CHD and thus provide information on the developmental mechanisms of CHD.

      Results

      Study subjects

      There were 12 major CHD cases and 26 controls. The mean gestational ages at blood draw were 25 3/7 (standard deviation [SD], 5 1/7) weeks of gestation vs 23 5/7 (SD, 5 4/7) weeks of gestation, respectively. There was no significant difference in these variables and other demographic and clinical characteristics between groups (Table 1). The CHD group included 4 cases of ventricular septal defects and 2 cases of tetralogy of Fallot. There was 1 case each of the following: pulmonary artery atresia with pulmonary valve stenosis, truncus arteriosus, pulmonary artery valve stenosis, ventricular septal defect with atrial septal defect, double aortic arch, and bicuspid atrioventricular valve with a dilated main pulmonary artery. There was no associated chromosomal and/or identified gene abnormality in these cases.
      Table 1Clinical and demographic characteristics of study subjects
      ParametersControlsCHDP value
      Number2612
      Age (y), mean (SD)31.36 (4.61)29.73 (5.59).32
      The P value is calculated using the t test
      Gestational age (wk) at sampling, mean (SD)23 5/7 (5 4/7)25 3/7 (5 1/7).15
      The P value is calculated using the Wilcoxon-Mann-Whitney test.
      BMI, mean (SD)30.93 (7.30)23.31 (10.01).03
      The P value is calculated using the Wilcoxon-Mann-Whitney test.
      Race, n (%)
      White22 (84.61)9 (75).81
      The P value is calculated using the Wilcoxon-Mann-Whitney test.
      Asian1 (3.84)1 (8.33)
      African American2 (7.69)2 (16.66)
      Other1 (3.84)0 (0)
      Gravidity, n (%)
      14 (15.38)3 (25.00).79
      The P value is calculated using the Wilcoxon-Mann-Whitney test.
      28 (30.79)2 (16.66)
      ≥314 (53.84)7 (58.33)
      Parity, n (%)
      05 (19.23)6 (50.00).03
      The P value is calculated using the Wilcoxon-Mann-Whitney test.
      19 (34.61)4 (33.33)
      25 (19.23)1 (8.33)
      ≥37 (26.92)1 (8.33)
      Previous child with CHD, n (%)
      No23 (88.46)12(100).08
      The P value is calculated using the t test
      Yes3 (11.53)0 (0)
      Family history of CHD, n (%)
      No21 (80.76)12 (100).43
      The P value is calculated using the Wilcoxon-Mann-Whitney test.
      Yes5 (19.23)0 (0)
      First-trimester alcohol use, n (%)
      No21 (80.76)9 (75)1.00
      The P value is calculated using the Wilcoxon-Mann-Whitney test.
      Yes5 (19.23)3 (25)
      First-trimester tobacco use, n (%)
      No25 (96.15)10 (83.33).29
      The P value is calculated using the Wilcoxon-Mann-Whitney test.
      Yes1 (3.84)2 (16.66)
      CHD, congenital heart defect; SD, standard deviation.
      Bahado-Singh. Artificial intelligence, circulating cell-free DNA, and detection of fetal congenital heart defect. Am J Obstet Gynecol 2022.
      a The P value is calculated using the t test
      b The P value is calculated using the Wilcoxon-Mann-Whitney test.

      Methylation changes in congenital heart defect

      A total of 5918 cytosine nucleotide or CpG loci (located in a total of 4976 genes), showed significant methylation change (P value of <.05 and a cytosine methylation difference of ≥5%) in the CHD group compared with controls. Each CpG locus predicted CHD with an area under the receiver operating characteristic curve (AUC) of ≥0.70 based on their methylation levels in the CHD group vs controls. A total of 527 separate CpG loci showed high individual predictive accuracy for CHD (AUC, ≥0.80). Each cytosine nucleotide, its associated gene based on the University of California Santa Cruz system, the level of CpG methylation in both CHD cases and controls, and individual AUC for the prediction of CHD for each locus are provided in the Supplemental Table 1. A total of 1278 cytosine loci throughout the genome were hypomethylated (reduced methylation in the CHD group vs controls), and 4640 cytosine loci were hypermethylated, that is, increased cytosine methylation in CHD. Hypermethylation of cytosines is classically associated with repressed gene expression or transcription, whereas hypomethylation is generally associated with increased gene expression. The greater the difference in cytosine methylation levels in a cytosine locus, the greater the likelihood that this will affect the expression of the pertinent gene
      • Kurdyukov S.
      • Bullock M.
      DNA methylation analysis: choosing the right method.
      in CHD.

      Artificial intelligence prediction of congenital heart defect using cell-free DNA

      SVM, one of the AI platforms evaluated, achieved an AUC of 0.97 (95% confidence interval [CI], 0.87–1.00) with 98% sensitivity and 94% specificity (Table 2) for the detection of CHD. As noted previously, the SVM model is resistant to overfitting or overestimation of the predictive accuracy. The RF model, another AI approach, achieved an AUC of 0.98 (95% CI, 0.83–1.00) with 93% sensitivity and specificity (Table 2). The other AI platforms achieved consistently high diagnostic accuracies, with AUCs of ≥92% (Table 2). The inclusion of commonly employed demographic and clinical predictors of CHD (listed in Table 1) did not persist as significant predictors of CHD when considered along with epigenetic markers and did not improve overall model performance (Table 3) and so did not add to the predictive performance. Thus, epigenetic markers were found to be superior to widely used clinical factors employed in the assignment of fetal CHD risk. Furthermore, high accuracy was achieved when the analysis used a more stringent standard for defining statistically significant cytosine methylation change for the CpGs. We employed a “genome-wide association study” threshold,
      • Fadista J.
      • Manning A.K.
      • Florez J.C.
      • Groop L.
      The (in)famous GWAS P-value threshold revisited and updated for low-frequency variants.
      that is, with CpG methylation change significance defined as a P value of <5×10−8, (Supplemental Table 2), which is used to determine significance in genome-wide analyses that require multiple comparisons. Here again, using this new definition of significance, the inclusion of conventional demographic, history, and clinical CHD factors as predictors of CHD did not contribute to or further improve predictive performance (Supplemental Table 3). We used AI to identify and rank the top 5 predictive CpG predictors to be included in each algorithm as seen in the table legends (Tables 2 and 3).
      Table 2Artificial intelligence and circulating cell-free DNA prediction of CHD
      VariableSVMGLMPAMRFLDADL
      AUC (95% CI)0.9700 (0.8770–1.0000)0.9400 (0.8240–1.0000)0.9300 (0.8300–1.0000)0.9800 (0.8310–1.0000)0.9200 (0.7240–1.0000)0.9400 (0.8400–1.0000)
      Sensitivity0.98000.92000.87000.93800.97000.9300
      Specificity0.94000.87000.92000.93200.93000.9400
      Individual CpG markers were chosen on the basis of a methylation difference of ≥5% and individual AUC for CHD detection of ≥0.70.
      Individual predictors in decreasing order of contribution:
      SVM: cg04761177, cg21431091, cg01263077, cg09853933, cg27142059
      GLM: cg24479965, cg01094213, cg22467129, cg24748945, cg01949461
      PAM: cg27142059, cg09386284, cg16551159, cg04761177, cg01263077
      RF: cg04761177, cg16551159, cg14957943, cg06978680, cg12592721
      LDA: cg04761177, cg27142059, cg18073832, cg25731807, cg03790075
      DL: cg04761177, cg21431091, cg01263077, cg09853933, cg27142059
      AUC, area under the receiver operating characteristic curve; CHD, congenital heart defect; CI, confidence interval; DL, deep learning; GLM, generalized linear model; LDA, linear discriminant analysis; PAM, prediction analysis for microarrays; RF, random forest; SVM, support vector machine.
      Bahado-Singh. Artificial intelligence, circulating cell-free DNA, and detection of fetal congenital heart defect. Am J Obstet Gynecol 2022.
      Table 3Artificial intelligence and circulating cell-free DNA, clinical and demographic prediction of CHD
      VariableSVMGLMPAMRFLDADL
      AUC (95% CI)0.9680 (0.8770–1.0000)0.9450 (0.8240–1.0000)0.9370 (0.8300–1.0000)0.9670 (0.8310–1.0000)0.9120 (0.7240–1.0000)0.9340 (0.8340–1.0000)
      Sensitivity0.97000.91000.88000.92800.96000.9300
      Specificity0.92000.86000.91000.92200.91000.9200
      Individual CpG markers were chosen on the basis of a methylation difference of ≥5% and individual AUC for CHD detection of ≥0.70. Demographic, history, and clinical factors simultaneously considered in the analysis include maternal age, gravidity, parity, gestational age at sample collection, body mass index, race, previous child with CHD, other family history of CHD, first-trimester alcohol use, and first-trimester tobacco use.
      Individual predictors in decreasing order of contribution:
      SVM: cg04761177, cg21431091, cg01263077, cg09853933, cg27142059
      GLM: cg24479965, cg01094213, cg22467129, cg24748945, cg01949461
      PAM: cg27142059, cg09386284, cg16551159, cg04761177, cg01263077
      RF: cg04761177, cg16551159, cg14957943, cg06978680, cg12592721
      LDA: cg04761177, cg27142059, cg18073832, cg25731807, cg03790075
      DL: cg04761177, cg21431091, cg01263077, cg09853933, cg27142059
      AUC, area under the receiver operating characteristic curve; CHD, congenital heart defect; CI, confidence interval; DL, deep learning; GLM, generalized linear model; LDA, linear discriminant analysis; PAM, prediction analysis for microarrays; RF, random forest; SVM, support vector machine.
      Bahado-Singh. Artificial intelligence, circulating cell-free DNA, and detection of fetal congenital heart defect. Am J Obstet Gynecol 2022.
      Additional analyses using larger numbers of CpG predictors in the algorithms did not greatly improve predictive performance. For example, a biomarker model combining 50 CpGs achieved the following performance using DL: AUC of 0.98 (95% CI, 0.86–1.00) with 92.0% sensitivity and 92.9% specificity. For a 100-marker predictive model using DL, the AUC was 0.97 (95% CI, 0.86–1.0) with 92.0% sensitivity and 91.9 % specificity. Similar results were obtained for the other 5 AI platforms.

      Logistic regression–based prediction

      For comparison, we also performed a conventional logistic regression analysis for the prediction of CHD. For the training (or discovery group), the predictive algorithm achieved an AUC of 0.98 (95% CI, 0.98–0.97) with 100% sensitivity and 85% specificity. After 10-fold cross-validation, the algorithm achieved an AUC of 0.79 (95% CI, 0.61–0.98) with 83% sensitivity and 80% specificity in a test (or validation) group.
      Logistic regression analysis using only CpG markers with methylation differences meeting the stringent genome-wide association study threshold, that is, a P value of <5×10−8, achieved similar results. With 10-fold cross-validation, the AUC was 0.71 (95% CI, 0.50–0.92) with 75% sensitivity and 87% specificity in the test group. Overall, AI achieved greater predictive accuracy than conventional logistic regression approaches.

      Cluster analysis of differentially methylated targets in cell-free DNA congenital heart defect

      On hierarchical cluster analysis using individual CpG methylation markers, we observed well-separated clusters or segregation between the CHD group and controls (Figure 1). The OPLS-DA plot (Supplemental Figure) also showed excellent separation of the CHD and normal groups using CpG markers. Overall, these provided visual evidence that circulating cfDNA methylation markers are able to distinguish CHD cases from controls.
      Figure thumbnail gr1
      Figure 1Heat map clustering of methylated CpG markers of CHD in cfDNA
      Each column represents a single subject, and each row represents a CpG probe as categorized by Illumina. The color gradient code toward the positive number (+2) represents hypermethylation, and the color gradient toward the negative number (−2) represents hypomethylation. Controls are coded with class “0,” and CHD cases are coded with class “1.”
      CHD, congenital heart defect.
      Bahado-Singh. Artificial intelligence, circulating cell-free DNA, and detection of fetal congenital heart defect. Am J Obstet Gynecol 2022.

      Disease and functional enrichment analysis

      Further elucidating the pathogenesis of CHD was an important objective of this study. This approach was used to understand the epigenetic basis of isolated CHD development. IPA systems showed significant disease and functional enrichment of the genes known to be associated with heart development and heart disease. Given the relationship between epigenetic alterations and change in gene expression, the results suggested that there is statistically significant differential gene expression between the 2 phenotypes, CHD and controls. The top 4 enriched or altered disease functions were “cardiovascular system development and function (p-3.8E-19),” “cardiac hypertrophy signaling (p-3.7E-6),” “congenital heart anomaly (p-8.55E-6),” and “cardiovascular disease (p-7.44E-15).” In other words, many of the genes that were epigenetically altered in CHD are known to be involved in the aforementioned roles. Figure 2 depicts the network of genes involved in these molecular pathways. Overall, these results have added biologic plausibility to our use of cytosine (gene) methylation markers given that the significant markers have been linked to biologic functions involving the heart.
      Figure thumbnail gr2
      Figure 2Ingenuity Pathway Analysis of significantly differentially methylated genes
      The top 4 significantly disease-enriched networks linked to congenital heart defect are depicted with their significant values (P values). The hyper- and hypomethylated genes are provided using different color codes.
      Bahado-Singh. Artificial intelligence, circulating cell-free DNA, and detection of fetal congenital heart defect. Am J Obstet Gynecol 2022.

      Comment

      Principal findings

      Given the previously stated limitations of prenatal ultrasound and newborn pulse oximetry screening, one of our objectives was to evaluate a new approach for the detection of fetal CHD. Using AI combined with cytosine methylation analysis of circulating cfDNA in maternal blood, we consistently and accurately predicted fetal CHDs using a minimally invasive approach. The SVM AI platform achieved a high detection rate for CHD: AUC of 0.97 (95% CI, 0.87–1.00) with a 98% sensitivity and 94% specificity. Comparably high predictive accuracies were also achieved using 5 other AI platforms. Of clinical interest is the finding that standard clinical and demographic predictors that are routinely used to determine high-risk CHD status and for the selection for fetal echocardiographic examination proved inferior to circulating cfDNA markers.
      Another important objective of precision medicine and this study was to further elucidate the pathogenesis of complex disorders. We found evidence that DNA methylation–based epigenetic changes are significantly associated with isolated, nonsyndromic CHD. Using pathway enrichment analysis, we found that the genes that were epigenetically altered had pertinent functional roles related to “cardiovascular system development and function,” “cardiac hypertrophy,” “congenital heart anomaly,” and “cardiovascular disease” (Figure 2), conferring biologic plausibility to our CpG results. In Supplemental Table 4, we summarized the known or suspected cardiac roles of some of the individual genes with altered cytosine methylation in our study.

      Results in the context of what is known

      We found evidence of a significant role of DNA methylation changes in the development of nonsyndromic or idiopathic CHD. Epigenetic mechanisms are known to regulate gene expression and affect cardiac embryogenesis
      • Vallaster M.
      • Vallaster C.D.
      • Wu S.M.
      Epigenetic mechanisms in cardiac development and disease.
      ,
      • O’Meara C.C.
      • Lee R.T.
      Peering Into the cardiomyocyte nuclear epigenetic State.
      and the development of CHD.
      • Radhakrishna U.
      • Albayrak S.
      • Zafra R.
      • et al.
      Placental epigenetics for evaluation of fetal congenital heart defects: ventricular septal defect (VSD).
      ,
      • Bahado-Singh R.O.
      • Wapner R.
      • Thom E.
      • et al.
      Elevated first-trimester nuchal translucency increases the risk of congenital heart defects.
      • Radhakrishna U.
      • Albayrak S.
      • Alpay-Savasan Z.
      • et al.
      Genome-wide DNA methylation analysis and epigenetic variations associated with congenital aortic valve stenosis (AVS).
      • Bahado-Singh R.O.
      • Zaffra R.
      • Albayarak S.
      • et al.
      Epigenetic markers for newborn congenital heart defect (CHD).
      Cardiac tissue is not accessible for research in living embryos. Given this fact, there is interest in developing molecular markers in surrogate, accessible tissues, such as blood leukocytes for the assessment and prediction of CHD in newborns.
      • Radhakrishna U.
      • Albayrak S.
      • Zafra R.
      • et al.
      Placental epigenetics for evaluation of fetal congenital heart defects: ventricular septal defect (VSD).
      ,
      • Bahado-Singh R.O.
      • Wapner R.
      • Thom E.
      • et al.
      Elevated first-trimester nuchal translucency increases the risk of congenital heart defects.
      • Radhakrishna U.
      • Albayrak S.
      • Alpay-Savasan Z.
      • et al.
      Genome-wide DNA methylation analysis and epigenetic variations associated with congenital aortic valve stenosis (AVS).
      • Bahado-Singh R.O.
      • Zaffra R.
      • Albayarak S.
      • et al.
      Epigenetic markers for newborn congenital heart defect (CHD).
      However, accessing fetal blood is too invasive and risky to be clinically useful. Extensive publications, both clinical and laboratory,
      • Burton G.J.
      • Jauniaux E.
      Development of the human placenta and fetal heart: synergic or independent?.
      ,
      • Maslen C.L.
      Recent advances in placenta-heart interactions.
      indicate that the placenta and its vasculature play a crucial role in cardiac embryogenesis and CHD development. For example, disorders, such as hypertension, which affect fetal-placental vasculature, have been shown in meta-analysis to significantly increase the risk of CHD.
      • Ramakrishnan A.
      • Lee L.J.
      • Mitchell L.E.
      • Agopian A.J.
      Maternal hypertension during pregnancy and the risk of congenital heart defects in offspring: a systematic review and meta-analysis.
      ,
      • van Gelder M.M.
      • Van Bennekom C.M.
      • Louik C.
      • Werler M.M.
      • Roeleveld N.
      • Mitchell A.A.
      Maternal hypertensive disorders, antihypertensive medication use, and the risk of birth defects: a case-control study.
      Furthermore, mutations in >750 genes in the mouse models resulted in alteration of both placental morphology and resulted in CHD.
      • Camm E.J.
      • Botting K.J.
      • Sferruzzi-Perri A.N.
      Near to One’s heart: the intimate relationship Between the placenta and fetal heart.
      In a series of studies, we demonstrated that methylation changes in the placental DNA significantly correlated with the development of CHD and that these placental epigenetic changes detected CHD with high accuracy.
      • Radhakrishna U.
      • Albayrak S.
      • Zafra R.
      • et al.
      Placental epigenetics for evaluation of fetal congenital heart defects: ventricular septal defect (VSD).
      ,
      • Bahado-Singh R.
      • Vishweswaraiah S.
      • Mishra N.K.
      • Guda C.
      • Radhakrishna U.
      Placental DNA methylation changes in detection of tetralogy of Fallot.

      Clinical implications

      There is ongoing proliferation, differentiation, and apoptosis of the placental trophoblast,
      • Taglauer E.S.
      • Wilkins-Haug L.
      • Bianchi D.W.
      Review: cell-free fetal DNA in the maternal circulation as an indication of placental health and disease.
      and apoptotic DNA material is continuously shed into the maternal circulation generating cell-free “fetal” DNA.
      • Tjoa M.L.
      • Cindrova-Davies T.
      • Spasic-Boskovic O.
      • Bianchi D.W.
      • Burton G.J.
      Trophoblastic oxidative stress and the release of cell-free feto-placental DNA.
      • Gupta A.K.
      • Holzgreve W.
      • Huppertz B.
      • Malek A.
      • Schneider H.
      • Hahn S.
      Detection of fetal DNA and RNA in placenta-derived syncytiotrophoblast microparticles generated in vitro.
      • Alberry M.
      • Maddocks D.
      • Jones M.
      • et al.
      Free fetal DNA in maternal plasma in anembryonic pregnancies: confirmation that the origin is the trophoblast.
      • Wataganara T.
      • Gratacos E.
      • Jani J.
      • et al.
      Persistent elevation of cell-free fetal DNA levels in maternal plasma after selective laser coagulation of chorionic plate anastomoses in severe midgestational twin-twin transfusion syndrome.
      This cfDNA is now routinely used clinically for noninvasive aneuploidy
      • Tjoa M.L.
      • Cindrova-Davies T.
      • Spasic-Boskovic O.
      • Bianchi D.W.
      • Burton G.J.
      Trophoblastic oxidative stress and the release of cell-free feto-placental DNA.
      • Gupta A.K.
      • Holzgreve W.
      • Huppertz B.
      • Malek A.
      • Schneider H.
      • Hahn S.
      Detection of fetal DNA and RNA in placenta-derived syncytiotrophoblast microparticles generated in vitro.
      • Alberry M.
      • Maddocks D.
      • Jones M.
      • et al.
      Free fetal DNA in maternal plasma in anembryonic pregnancies: confirmation that the origin is the trophoblast.
      • Wataganara T.
      • Gratacos E.
      • Jani J.
      • et al.
      Persistent elevation of cell-free fetal DNA levels in maternal plasma after selective laser coagulation of chorionic plate anastomoses in severe midgestational twin-twin transfusion syndrome.
      detection. We reported a potential approach that could have utility for CHD screening, a significantly more common disorder that the chromosomal disorders for which routine prenatal screening has long been a standard of care. However, it bears emphasizing that molecular testing is unlikely to replace imaging studies. Determining the location and 3-dimensional relationships of heart lesions is crucial to prognostication, immediate newborn and ultimate surgical management. These features can only be determined by imaging. Therefore, epigenetic testing has the potential to better identify pregnancies at elevated risk of CHD and thus permit the funneling of such cases for prenatal and newborn ECHO in centers with the appropriate expertise. Should our findings be confirmed, the clinical benefits of improved prenatal detection seem clear. For example, it would allow for the arrangement to deliver at tertiary care centers, which would result in optimizing neonatal and perioperative care and favorably affect pediatric morbidity and mortality.
      • van Velzen C.L.
      • Clur S.A.
      • Rijlaarsdam M.E.
      • et al.
      Prenatal detection of congenital heart disease--results of a national screening programme.
      Of note, we focused on isolated CHD with no other structural or chromosomal anomalies, the most difficult category for ultrasound detection. Improving the quality of ultrasound examinations by itself will not solve the problem of low prenatal detection rates as data show that significant segments of the population, rural and also impoverished communities, currently have limited access to expert prenatal ultrasound services.
      • Hill G.D.
      • Block J.R.
      • Tanem J.B.
      • Frommelt M.A.
      Disparities in the prenatal detection of critical congenital heart disease.
      The performance of a blood test to identify high-risk cases to be subsequently referred to expert centers for ultrasound could significantly improve access and prenatal detection. In this regard, we reported, in this study, that the molecular markers were significantly superior to standard clinical “markers,” for example, family history for identifying high-risk patients who would need to be referred for expert evaluation. As outlined in the Introduction, the major potential clinical implication of our findings could be a reduction in infant and pediatric morbidity and mortality.

      Research implications

      Oncology has been one of the principal focus areas in precision medicine. This has contributed to a surge in interest in research in circulating cell-free tumors for cancer diagnostics, prognostics, and therapeutics and has led to recent Food and Drug Administration approval of a few diagnostic tests based on this approach.
      • Ignatiadis M.
      • Sledge G.W.
      • Jeffrey S.S.
      Liquid biopsy enters the clinic - implementation issues and future challenges.
      Although the focus has been on mutation analysis of cell-free tumor circulating tumor DNA (ctDNA), more recent publications have demonstrated the value
      • Bahado-Singh R.
      • Vlachos K.T.
      • Aydas B.
      • Gordevicius J.
      • Radhakrishna U.
      • Vishweswaraiah S.
      Precision oncology: artificial intelligence and DNA methylation analysis of circulating cell-free DNA for lung cancer detection.
      ,
      • Bahado-Singh R.O.
      • Radhakrishna U.
      • Gordevičius J.
      • et al.
      Artificial intelligence and circulating cell-free DNA methylation profiling: mechanism and detection of Alzheimer’s disease.
      and the advantages of DNA methylation analysis of ctDNA in precision oncology. Furthermore, recent publications have used genome-wide methylation analysis of circulating cfDNA for the accurate detection of and mechanisms of Alzheimer disease, a brain disorder.
      • Bahado-Singh R.O.
      • Vishweswaraiah S.
      • Aydas B.
      • et al.
      Artificial intelligence and leukocyte epigenomics: evaluation and prediction of late-onset Alzheimer’s disease.
      Precision cardiology aims to understand the specific pathogenesis of cardiovascular diseases, thus informing the development of effective therapies. Circulating cfDNA screening offers a new potential “on-ramp” in the development of precision fetal cardiology and investigation of the fetal heart. Dietary folate is the primary source of the “methyl group” (carbon atom) used in DNA and other methylation reactions.
      • Mahmoud A.M.
      • Ali M.M.
      Methyl donor micronutrients that modify DNA methylation and cancer outcome.
      Epidemiologic studies have linked population dietary folate supplementation with a reduced risk of CHD.
      • Li Y.
      • Huang T.
      • Zheng Y.
      • Muka T.
      • Troup J.
      • Hu F.B.
      Folic acid supplementation and the risk of cardiovascular diseases: a meta-analysis of randomized controlled trials.
      A national population-based study found that this is particularly the case when folate supplementation is targeted to the embryonic period.
      • Czeizel A.E.
      • Vereczkey A.
      • Szabó I.
      Folic acid in pregnant women associated with reduced prevalence of severe congenital heart defects in their children: a national population-based case-control study.
      This supports the role of DNA methylation change in CHD development and should be investigated as a research tool for assessing folate deficiency and the therapeutic benefits of targeted maternal supplementation to prevent and even potentially “rescue” the CHD phenotype.
      • Obeid R.
      • Holzgreve W.
      • Pietrzik K.
      Folate supplementation for prevention of congenital heart defects and low birth weight: an update.
      To address those objectives, future studies should concentrate on cfDNA testing in the embryonic period of <8 weeks, when the heart is developing.
      Although our study was preliminary and far removed from clinical introduction, a screening approach using a panel of cfDNA epigenomic markers should be considered in future research to identify high-risk pregnancies, given its apparent superiority to conventional clinical risk prediction. Perhaps more significant is the opportunity this approach could provide for the investigation and ongoing minimally invasive monitoring of the molecular mechanisms of heart development in vivo.

      Strengths and limitations

      A strength of our study was the finding of high predictive accuracy for fetal CHD using a minimally invasive approach. This study reported using circulating cfDNA, genome-wide epigenomics, and AI analysis in pregnancy. Another strength is that we focused on isolated, nonsyndromic CHD, which accounts for most CHDs. Our study was not without limitations; one such limitation was the small sample size. There was no previous study that permitted meaningful assessment of the sample size needed. Larger prospective studies will be needed to validate our findings. However, despite this limitation, highly statistically significant predictive accuracies were observed. A related limitation was that a narrow variety of CHDs were examined, although it included some of the most common and clinically significant varieties of CHD encountered in practice. The evaluation of other CHD types is necessary.

      Conclusions

      This study reported using circulating cfDNA, whole-genome epigenetic, and AI analyses to predict and investigate the pathogenesis of fetal nonsyndromic CHD. We identified several significantly differentially methylated CpG markers. Many of the genes that were found to have altered methylation in our study subjects are currently known or believed to have important roles in cardiac development, CHD, and adult cardiovascular disease. The identified functional roles of these genes will add biologic plausibility to our findings and provide additional mechanistic information, which could be a crucial launching pad for the future development of targeted therapy and prevention strategies consistent with precision cardiology objectives. Finally, the development of minimally invasive methods for the detection of prenatal CHD is an important clinical priority and, as shown in this study, seems to be feasible.

      References

        • Baek S.H.
        Challenges and future in precision cardiovascular medicine.
        Cardiovasc Prev Pharmacother. 2019; 1: 10-18
        • Sadovsky Y.
        • Mesiano S.
        • Burton G.J.
        • et al.
        Advancing human health in the decade ahead: pregnancy as a key window for discovery: a Burroughs Wellcome Fund Pregnancy Think Tank.
        Am J Obstet Gynecol. 2020; 223: 312-321
        • Antman E.M.
        • Loscalzo J.
        Precision medicine in cardiology.
        Nat Rev Cardiol. 2016; 13: 591-602
        • Semsarian C.
        • Ingles J.
        • Ross S.B.
        • Dunwoodie S.L.
        • Bagnall R.D.
        • Kovacic J.C.
        Precision medicine in cardiovascular disease: genetics and impact on phenotypes: JACC Focus Seminar 1/5.
        J Am Coll Cardiol. 2021; 77: 2517-2530
        • Blue G.M.
        • Kirk E.P.
        • Sholler G.F.
        • Harvey R.P.
        • Winlaw D.S.
        Congenital heart disease: current knowledge about causes and inheritance.
        Med J Aust. 2012; 197: 155-159
        • van der Linde D.
        • Konings E.E.
        • Slager M.A.
        • et al.
        Birth prevalence of congenital heart disease worldwide: a systematic review and meta-analysis.
        J Am Coll Cardiol. 2011; 58: 2241-2247
        • Holland B.J.
        • Myers J.A.
        • Woods Jr., C.R.
        Prenatal diagnosis of critical congenital heart disease reduces risk of death from cardiovascular compromise prior to planned neonatal cardiac surgery: a meta-analysis.
        Ultrasound Obstet Gynecol. 2015; 45: 631-638
        • Olney R.S.
        • Ailes E.C.
        • Sontag M.K.
        Detection of critical congenital heart defects: review of contributions from prenatal and newborn screening.
        Semin Perinatol. 2015; 39: 230-237
        • Combs C.A.
        • Hameed A.B.
        • Friedman A.M.
        • Hoskins I.A.
        • Patient Safety
        • Quality Committee; Society for Maternal-Fetal Medicine
        Special statement: proposed quality metrics to assess accuracy of prenatal detection of congenital heart defects.
        Am J Obstet and Gynecol. 2020; 222: B2-B9
        • Li Y.F.
        • Zhou K.Y.
        • Fang J.
        • Wang C.
        • Hua Y.M.
        • Mu D.Z.
        Efficacy of prenatal diagnosis of major congenital heart disease on perinatal management and perioperative mortality: a meta-analysis.
        World J Pediatr. 2016; 12: 298-307
        • van Velzen C.L.
        • Clur S.A.
        • Rijlaarsdam M.E.
        • et al.
        Prenatal detection of congenital heart disease--results of a national screening programme.
        BJOG. 2016; 123: 400-407
        • PennState
        Using artificial intelligence to detect discrimination.
        (Available at:) (Accessed July 10, 2019)
        • Pinto N.M.
        • Keenan H.T.
        • Minich L.L.
        • Puchalski M.D.
        • Heywood M.
        • Botto L.D.
        Barriers to prenatal detection of congenital heart disease: a population-based study.
        Ultrasound Obstet Gynecol. 2012; 40: 418-425
        • Tinker S.C.
        • Gilboa S.M.
        • Moore C.A.
        • et al.
        Specific birth defects in pregnancies of women with diabetes: National Birth Defects Prevention Study, 1997-2011.
        Am J Obstet Gynecol. 2020; 222: 176.e1-176.e11
        • Bateman B.T.
        • Huybrechts K.F.
        • Fischer M.A.
        • et al.
        Chronic hypertension in pregnancy and the risk of congenital malformations: a cohort study.
        Am J Obstet Gynecol. 2015; 212 (e1–14): 337
        • Cai G.J.
        • Sun X.X.
        • Zhang L.
        • Hong Q.
        Association between maternal body mass index and congenital heart defects in offspring: a systematic review.
        Am J Obstet Gynecol. 2014; 211: 91-117
        • Botto L.D.
        • Panichello J.D.
        • Browne M.L.
        • et al.
        Congenital heart defects after maternal fever.
        Am J Obstet Gynecol. 2014; 210 (e1–11): 359
        • Byrne J.J.
        • Morgan J.L.
        • Twickler D.M.
        • McIntire D.D.
        • Dashe J.S.
        Utility of follow-up standard sonography for fetal anomaly detection.
        Am J Obstet Gynecol. 2020; 222 (e1–9): 615
        • Porreco R.P.
        • Garite T.J.
        • Maurel K.
        • et al.
        Noninvasive prenatal screening for fetal trisomies 21, 18, 13 and the common sex chromosome aneuploidies from maternal blood using massively parallel genomic sequencing of DNA.
        Am J Obstet Gynecol. 2014; 211 (e1–12): 365
        • Martinez S.R.
        • Gay M.S.
        • Zhang L.
        Epigenetic mechanisms in heart development and disease.
        Drug Discov Today. 2015; 20: 799-811
        • Zong D.
        • Liu X.
        • Li J.
        • Ouyang R.
        • Chen P.
        The role of cigarette smoke-induced epigenetic alterations in inflammation.
        Epigenetics Chromatin. 2019; 12: 65
        • Radhakrishna U.
        • Albayrak S.
        • Zafra R.
        • et al.
        Placental epigenetics for evaluation of fetal congenital heart defects: ventricular septal defect (VSD).
        PLoS One. 2019; 14e0200229
        • Krittanawong C.
        • Zhang H.
        • Wang Z.
        • Aydar M.
        • Kitai T.
        Artificial intelligence in precision cardiovascular medicine.
        J Am Coll Cardiol. 2017; 69: 2657-2664
        • Mazaki J.
        • Katsumata K.
        • Ohno Y.
        • et al.
        A novel prediction model for colon cancer recurrence using auto-artificial intelligence.
        Anticancer Res. 2021; 41: 4629-4636
        • Mango L.J.
        Reducing false negatives in clinical practice: the role of neural network technology.
        Am J Obstet Gynecol. 1996; 175: 1114-1119
        • Taglauer E.S.
        • Wilkins-Haug L.
        • Bianchi D.W.
        Review: cell-free fetal DNA in the maternal circulation as an indication of placental health and disease.
        Placenta. 2014; 35: S64-S68
        • Barták B.K.
        • Kalmár A.
        • Galamb O.
        • et al.
        Blood collection and cell-free DNA isolation methods influence the sensitivity of liquid biopsy analysis for colorectal cancer detection.
        Pathol Oncol Res. 2019; 25: 915-923
        • Bahado-Singh R.O.
        • Vishweswaraiah S.
        • Aydas B.
        • Mishra N.K.
        • Guda C.
        • Radhakrishna U.
        Deep learning/artificial intelligence and blood-based DNA epigenomic prediction of cerebral palsy.
        Int J Mol Sci. 2019; 20: 2075
        • Bahado-Singh R.O.
        • Vishweswaraiah S.
        • Aydas B.
        • et al.
        Artificial intelligence and leukocyte epigenomics: evaluation and prediction of late-onset Alzheimer’s disease.
        PLoS One. 2021; 16e0248375
        • Bahado-Singh R.O.
        • Vishweswaraiah S.
        • Aydas B.
        • Radhakrishna U.
        Placental DNA methylation changes and the early prediction of autism in full-term newborns.
        PLoS One. 2021; 16e0253340
        • Krämer A.
        • Green J.
        • Pollard Jr., J.
        • Tugendreich S.
        Causal analysis approaches in Ingenuity Pathway Analysis.
        Bioinformatics. 2014; 30: 523-530
        • Kurdyukov S.
        • Bullock M.
        DNA methylation analysis: choosing the right method.
        Biology (Basel). 2016; 5: 3
        • Fadista J.
        • Manning A.K.
        • Florez J.C.
        • Groop L.
        The (in)famous GWAS P-value threshold revisited and updated for low-frequency variants.
        Eur J Hum Genet. 2016; 24: 1202-1205
        • Vallaster M.
        • Vallaster C.D.
        • Wu S.M.
        Epigenetic mechanisms in cardiac development and disease.
        Acta Biochim Biophys Sin (Shanghai). 2012; 44: 92-102
        • O’Meara C.C.
        • Lee R.T.
        Peering Into the cardiomyocyte nuclear epigenetic State.
        Circ Res. 2015; 117: 392-394
        • Bahado-Singh R.O.
        • Wapner R.
        • Thom E.
        • et al.
        Elevated first-trimester nuchal translucency increases the risk of congenital heart defects.
        Am J Obstet Gynecol. 2005; 192: 1357-1361
        • Radhakrishna U.
        • Albayrak S.
        • Alpay-Savasan Z.
        • et al.
        Genome-wide DNA methylation analysis and epigenetic variations associated with congenital aortic valve stenosis (AVS).
        PLoS One. 2016; 11e0154010
        • Bahado-Singh R.O.
        • Zaffra R.
        • Albayarak S.
        • et al.
        Epigenetic markers for newborn congenital heart defect (CHD).
        J Matern Fetal Neonatal Med. 2016; 29: 1881-1887
        • Burton G.J.
        • Jauniaux E.
        Development of the human placenta and fetal heart: synergic or independent?.
        Front Physiol. 2018; 9: 373
        • Maslen C.L.
        Recent advances in placenta-heart interactions.
        Front Physiol. 2018; 9: 735
        • Ramakrishnan A.
        • Lee L.J.
        • Mitchell L.E.
        • Agopian A.J.
        Maternal hypertension during pregnancy and the risk of congenital heart defects in offspring: a systematic review and meta-analysis.
        Pediatr Cardiol. 2015; 36: 1442-1451
        • van Gelder M.M.
        • Van Bennekom C.M.
        • Louik C.
        • Werler M.M.
        • Roeleveld N.
        • Mitchell A.A.
        Maternal hypertensive disorders, antihypertensive medication use, and the risk of birth defects: a case-control study.
        BJOG. 2015; 122: 1002-1009
        • Camm E.J.
        • Botting K.J.
        • Sferruzzi-Perri A.N.
        Near to One’s heart: the intimate relationship Between the placenta and fetal heart.
        Front Physiol. 2018; 9: 629
        • Bahado-Singh R.
        • Vishweswaraiah S.
        • Mishra N.K.
        • Guda C.
        • Radhakrishna U.
        Placental DNA methylation changes in detection of tetralogy of Fallot.
        Ultrasound Obstet Gynecol. 2020; 55: 768-775
        • Tjoa M.L.
        • Cindrova-Davies T.
        • Spasic-Boskovic O.
        • Bianchi D.W.
        • Burton G.J.
        Trophoblastic oxidative stress and the release of cell-free feto-placental DNA.
        Am J Pathol. 2006; 169: 400-404
        • Gupta A.K.
        • Holzgreve W.
        • Huppertz B.
        • Malek A.
        • Schneider H.
        • Hahn S.
        Detection of fetal DNA and RNA in placenta-derived syncytiotrophoblast microparticles generated in vitro.
        Clin Chem. 2004; 50: 2187-2190
        • Alberry M.
        • Maddocks D.
        • Jones M.
        • et al.
        Free fetal DNA in maternal plasma in anembryonic pregnancies: confirmation that the origin is the trophoblast.
        Prenat Diagn. 2007; 27: 415-418
        • Wataganara T.
        • Gratacos E.
        • Jani J.
        • et al.
        Persistent elevation of cell-free fetal DNA levels in maternal plasma after selective laser coagulation of chorionic plate anastomoses in severe midgestational twin-twin transfusion syndrome.
        Am J Obstet Gynecol. 2005; 192: 604-609
        • Hill G.D.
        • Block J.R.
        • Tanem J.B.
        • Frommelt M.A.
        Disparities in the prenatal detection of critical congenital heart disease.
        Prenat Diagn. 2015; 35: 859-863
        • Ignatiadis M.
        • Sledge G.W.
        • Jeffrey S.S.
        Liquid biopsy enters the clinic - implementation issues and future challenges.
        Nat Rev Clin Oncol. 2021; 18: 297-312
        • Bahado-Singh R.
        • Vlachos K.T.
        • Aydas B.
        • Gordevicius J.
        • Radhakrishna U.
        • Vishweswaraiah S.
        Precision oncology: artificial intelligence and DNA methylation analysis of circulating cell-free DNA for lung cancer detection.
        Front Oncol. 2022; 12790645
        • Bahado-Singh R.O.
        • Radhakrishna U.
        • Gordevičius J.
        • et al.
        Artificial intelligence and circulating cell-free DNA methylation profiling: mechanism and detection of Alzheimer’s disease.
        Cells. 2022; 11: 1744
        • Mahmoud A.M.
        • Ali M.M.
        Methyl donor micronutrients that modify DNA methylation and cancer outcome.
        Nutrients. 2019; 11: 608
        • Li Y.
        • Huang T.
        • Zheng Y.
        • Muka T.
        • Troup J.
        • Hu F.B.
        Folic acid supplementation and the risk of cardiovascular diseases: a meta-analysis of randomized controlled trials.
        J Am Heart Assoc. 2016; 5e003768
        • Czeizel A.E.
        • Vereczkey A.
        • Szabó I.
        Folic acid in pregnant women associated with reduced prevalence of severe congenital heart defects in their children: a national population-based case-control study.
        Eur J Obstet Gynecol Reprod Biol. 2015; 193: 34-39
        • Obeid R.
        • Holzgreve W.
        • Pietrzik K.
        Folate supplementation for prevention of congenital heart defects and low birth weight: an update.
        Cardiovasc Diagn Ther. 2019; 9: S424-S433