Locus-specific DNA methylation prediction in cord blood and placenta

Abstract

DNA methylation is known to be responsive to prenatal exposures, which may be a part of the mechanism linking early developmental exposures to future chronic diseases. Many studies use blood to measure DNA methylation, yet we know that DNA methylation is tissue specific. Placenta is central to fetal growth and development, but it is rarely feasible to collect this tissue in large epidemiological studies; on the other hand, cord blood samples are more accessible. In this study, based on paired samples of both placenta and cord blood tissues from 169 individuals, we investigated the methylation concordance between placenta and cord blood. We then employed a machine-learning-based model to predict locus-specific DNA methylation levels in placenta using DNA methylation levels in cord blood. We found that methylation correlation between placenta and cord blood is lower than other tissue pairs, consistent with existing observations that placenta methylation has a distinct pattern. Nonetheless, there are still a number of CpG sites showing robust association between the two tissues. We built prediction models for placenta methylation based on cord blood data and documented a subset of 1,012 CpG sites with high correlation between measured and predicted placenta methylation levels. The resulting list of CpG sites and prediction models could help to reveal the loci where internal or external influences may affect DNA methylation in both placenta and cord blood, and provide a reference data to predict the effects on placenta in future study even when the tissue is not available in an epidemiological study.

Keywords: DNA methylation; cord blood; illumina humanmethylation 450; machine learning; placenta.

Figure 1.

R² of the 169 samples. The x-axis represents the sample index from 1 to 169, and the y-axis represents the CpG-wise R² of the 169 samples. The red line represents the CpG-wise R² of measured methylation beta values in the cord blood and measured methylation beta values in the placenta, and the purple line represents the CpG-wise R² of measured methylation beta values in the placenta and predicted methylation beta values in the placenta by single-CpG-based SVM and leave-one-out cross-validation.

Figure 2.

Methylation pattern across tissues and the between-tissue differences across individuals. (2a). Red circles indicate measured placenta beta values vs. measured cord blood beta values (x = measured placenta beta values, y = measured cord blood beta values), and purple circles indicate measured placenta beta values vs. SVM predicted placenta beta values (x = measured placenta beta values, y = predicted placenta beta values by single-CpG-based SVM and leave-one-out cross-validation). R²(raw) = CpG-wise R² between measured methylation beta values in the placenta and measured methylation beta values in the cord blood. R²(svm) = CpG-wise R² between measured methylation beta values in the placenta and predicted methylation beta values in the placenta by single-CpG-based SVM and leave-one-out cross-validation.

Figure 2b.

Red circles indicate measured placenta beta values–measured cord blood beta values of sample #1 vs. measured placenta beta values–measured cord blood beta values of sample #2(x = measured placenta beta values–measured cord blood beta values of sample #1, y = measured placenta beta values–measured cord blood beta values of sample #2). Purple circles indicate measured placenta beta values–SVM predicted placenta beta values in sample #1 vs. measured placenta beta values–SVM predicted placenta beta values in sample #2(x = measured placenta beta values–SVM predicted placenta beta values of sample #1, y = measured placenta beta values–SVM predicted placenta beta values of sample #2). Note: ‘–’ represents minus sign.

Figure 3.

Relationship between R² and mean absolute error. The y-axis is the sample-wise mean absolute error (MAE) of measured placenta methylation beta values and predicted placenta methylation beta values by single-CpG-based SVM and leave-one-out cross-validation, and the x-axis is the sample-wise R² of measured placenta methylation beta values and predicted placenta methylation beta values by single-CpG-based SVM and leave-one-out cross-validation (CpGs with SD > 0.1).

Figure 4.

Effect of sample size on prediction accuracy. The x-axis is the sample size of the training dataset, and the y-axis is the mean of CpG-wise R² for measured placenta methylation beta values and predicted placenta methylation beta values by single-CpG-based SVM for 100 testing samples. For the blue line, the extreme CpG sites with a minimum methylation beta value >0.8 or a maximum beta value <0.2 were removed.

Figure 5.

Venn diagram of well-predicted CpGs across the three datasets. The numbers in the circle represent the well-predicted CpGs (sample-wise R² > 0.8) in the three datasets, and the Venn diagram shows the intersection of the well-predicted CpGs across the following three datasets: cord blood–placenta, blood–artery and blood–atrium.