DNA Methylation of Window of Implantation Genes in Cervical Secretions Predicts Ongoing Pregnancy in Infertility Treatment

Abstract

Window of implantation (WOI) genes have been comprehensively identified at the single cell level. DNA methylation changes in cervical secretions are associated with in vitro fertilization embryo transfer (IVF-ET) outcomes. Using a machine learning (ML) approach, we aimed to determine which methylation changes in WOI genes from cervical secretions best predict ongoing pregnancy during embryo transfer. A total of 2708 promoter probes were extracted from mid-secretory phase cervical secretion methylomic profiles for 158 WOI genes, and 152 differentially methylated probes (DMPs) were selected. Fifteen DMPs in 14 genes (BMP2, CTSA, DEFB1, GRN, MTF1, SERPINE1, SERPINE2, SFRP1, STAT3, TAGLN2, TCF4, THBS1, ZBTB20, ZNF292) were identified as the most relevant to ongoing pregnancy status. These 15 DMPs yielded accuracy rates of 83.53%, 85.26%, 85.78%, and 76.44%, and areas under the receiver operating characteristic curves (AUCs) of 0.90, 0.91, 0.89, and 0.86 for prediction by random forest (RF), naïve Bayes (NB), support vector machine (SVM), and k-nearest neighbors (KNN), respectively. SERPINE1, SERPINE2, and TAGLN2 maintained their methylation difference trends in an independent set of cervical secretion samples, resulting in accuracy rates of 71.46%, 80.06%, 80.72%, and 80.68%, and AUCs of 0.79, 0.84, 0.83, and 0.82 for prediction by RF, NB, SVM, and KNN, respectively. Our findings demonstrate that methylation changes in WOI genes detected noninvasively from cervical secretions are potential markers for predicting IVF-ET outcomes. Further studies of cervical secretion of DNA methylation markers may provide a novel approach for precision embryo transfer.

Keywords: DNA methylation; boruta algorithm; cervical secretion; endometrial receptivity; in vitro fertilization; machine learning; noninvasive; window of implantation.

Figure 1

Study logistics. A total of 158 WOI genes were selected from the single cell endometrial transcriptomic profiles. In total, 2708 promoter probes of the 158 WOI genes were extracted from methylomic profiles generated by an EPIC 850K BeadChip array, from 31 pregnancy and 37 nonpregnancy mid-secretory phase cervical secretion samples. Probes were ranked by p-value to select 152 DMPs. These DMPs were used in the following refinement step using the Boruta algorithm to identify the 15 DMPs most relevant to ongoing pregnancy status. Three selected genes were verified by qMSP in an independent set of 30 pregnancy and 35 nonpregnancy mid-secretory phase cervical secretion samples. Different sets of selected DMPs or genes were used in four ML algorithms for predicting ongoing pregnancy in the two datasets (array and qMSP datasets). RF: random forest, NB: naïve Bayes, SVM: support vector machine, KNN: k-nearest neighbors, NP: nonpregnancy, P: pregnancy, DMPs: differentially methylated probes, WOI: window of implantation, qMSP: quantitative methylation-specific PCR, CV: cross-validation.

Figure 2

Methylation differences of WOI genes between pregnancy and nonpregnancy cervical secretion samples. A heatmap showing methylation differences of 152 DMPs (87 genes) between the pregnancy and nonpregnancy cervical secretion sample groups. Samples are presented vertically, and values of DNA methylation of the DMPs are presented horizontally. Green and pink columns represent nonpregnancy and pregnancy samples, respectively. Methylation levels are presented as low (dark blue) to high (dark red).

Figure 3

Selection of 15 DMPs and their predictive performance in four ML algorithms. (a) A box plot demonstrates the importance ranking of 152 attributes (DMPs) compared with shadow attributes by BA. The blue, red, and green box plots represent the importance levels of the shadow, rejected, and confirmed attributes, respectively. (b) Fifteen selected DMPs from the feature selection by BA. (c) ROC curves (AUCs) of the best models from four ML algorithms using the 15 selected DMPs for predicting ongoing pregnancy in the array dataset. AUC values were generated from a 100−time repeated fivefold cross-validation. BA: Boruta algorithm, ML: machine learning, DMP: differentially methylated probe, ROC: receiver operating characteristic, AUC: area under the ROC curve, RF: random forest, NB: naïve Bayes, SVM: support vector machine, KNN: k-nearest neighbor.

Figure 4

Methylation levels and predictive performance of three selected candidate genes in four ML algorithms. (a) A box plot shows differences in methylation levels (β-values) measured by an EPIC 850 K BeadChip array of three selected DMPs from three candidate genes between pregnancy (n = 31) and nonpregnancy (n = 37) cervical secretion samples. (b) A box plot shows differences in methylation levels (ΔCp values, transformed into 2^−(ΔCp) values) measured by qMSP of three selected candidate genes between pregnancy (n = 30) and nonpregnancy (n = 35) cervical secretion samples. (c) ROC curves (AUCs) of the best models of four ML algorithms using three selected candidate genes for predicting ongoing pregnancy in the qMSP dataset. AUC values were generated from a 100−time repeated fivefold cross-validation. ML: machine learning, ROC: receiver operating characteristic, AUC: area under the ROC curve, RF: random forest, NB: naïve Bayes, SVM: support vector machine, KNN: k-nearest neighbor.