Prediction of gestational age based on genome-wide differentially methylated regions

Jon Bohlin, Siri Eldevik Haberg, Per Magnus, Sarah Reese, Hakon Gjessing, Maria Magnus, Christine Parr, Christian Page, Stephanie London, Wenche Nystad

Research output: Contribution to journalArticle (Academic Journal)peer-review

113 Citations (Scopus)
323 Downloads (Pure)



We explored the association between gestational age and cord blood DNA methylation at birth and whether DNA methylation could be effective in predicting gestational age due to limitations with the presently used methods. We used data from the Norwegian Mother and Child Birth Cohort study (MoBa) with Illumina HumanMethylation450 data measured for 1753 newborns in two batches: MoBa 1, n = 1068; and MoBa 2, n = 685. Gestational age was computed using both ultrasound and the last menstrual period. We evaluated associations between DNA methylation and gestational age and developed a statistical model for predicting gestational age using MoBa 1 for training and MoBa 2 for predictions. The prediction model was additionally used to compare ultrasound and last menstrual period-based gestational age predictions. Furthermore, both CpGs and associated genes detected in the training models were compared to those detected in a published prediction model for chronological age.


There were 5474 CpGs associated with ultrasound gestational age after adjustment for a set of covariates, including estimated cell type proportions, and Bonferroni-correction for multiple testing. Our model predicted ultrasound gestational age more accurately than it predicted last menstrual period gestational age.


DNA methylation at birth appears to be a good predictor of gestational age. Ultrasound gestational age is more strongly associated with methylation than last menstrual period gestational age. The CpGs linked with our gestational age prediction model, and their associated genes, differed substantially from the corresponding CpGs and genes associated with a chronological age prediction model.
Original languageEnglish
Article number207
Number of pages9
JournalGenome Biology
Publication statusPublished - 7 Oct 2016


Dive into the research topics of 'Prediction of gestational age based on genome-wide differentially methylated regions'. Together they form a unique fingerprint.

Cite this