Since 2011, three generations of epigenetic clocks have been developed to estimate biological age[1-3]. The first-generation of models were generated by using DNA methylation data in various tissues to predict chronological age (outcome)[1, 4-6]. Second-generation models, such as PhenoAge and GrimAge used health outcomes, including all-cause mortality, for a more accurate determination of the latent biological age[7-9]. The latest, third-generation clocks like DunedinPoAm use longitudinal data to estimate the rate of aging[10]. This generation also includes universal clocks applicable to multiple species, such as the universal pan-mammalian epigenetic clock[11].
Biological age, as captured by these DNA methylation clocks, can be influenced by environmental factors, including smoking, obesity, sleep patterns, diet and exercise, stress, as well as diseases like cancer, diabetes, and Down syndrome[12-18]. The role of epigenetic programming in fetal development is crucial [19-21]. There is some evidence that in utero environmental factors are linked to variations in the epigenetic-predicted placenta age, yet research in this area is still limited in scope [22-27]. This may stem from the challenge involved in accurately estimating biological age of the placenta from epigenetics data so that accelerated ageing can be determined and linked to disease.
To date, two notable placental epigenetic clocks have been developed offering insight into the ability of DNA methylation profiles to predict gestational age. The first, developed by Mayne et al., utilized the Illumina 27k methylation array to demonstrate that gestational age could be reliably predicted through placental methylation data[22]. This clock, when evaluated on independent data, demonstrated a mean absolute error (MAE) of 2.6 weeks and a Pearson correlation coefficient of 0.94[28]. Subsequently, in 2019, Lee et al. introduced another epigenetic clock using the more advanced Illumina 450k methylation array and a larger training dataset, which showed improved performance on the same test dataset with an MAE of 0.96 weeks and a correlation of 0.96[28].
The objective of this DREAM challenge is to leverage the capabilities of the latest Illumina methylation arrays, specifically the EPIC arrays (850k CpG sites), which offer even greater coverage of the methylome, and herein we expand the training sample size. By developing a new epigenetic clock based on this technology, we aim to achieve greater accuracy in predicting gestational age, a surrogate for the biological age of the placenta.
