The relationship between aging and epigenetic profiles has been highlighted in many recent studies. Models using somatic cell methylomes to predict age have been successfully constructed. However, gamete aging is quite distinct and as such age prediction using sperm methylomes is ineffective with current techniques.

We have produced a model that utilizes human sperm DNA methylation signatures to predict chronological age by utilizing methylation array data from a total of 329 samples. The dataset used for model construction includes infertile patients, sperm donors, and individuals from the general population. Our model is capable predicting age with an R2 of 0.89, a mean absolute error (MAE) of 2.04 years, and a mean absolute percent error (MAPE) of 6.28% in our data set. We additionally investigated the reproducibility of prediction with our model in an independent cohort where 6 technical replicates of 10 individual samples were tested on different arrays. We found very similar age prediction accuracy (MAE = 2.37 years; MAPE = 7.05%) with a high degree of precision between replicates (standard deviation of only 0.877 years). Additionally, we found that smokers trended toward increased age profiles when compared to ‘never smokers’ though this pattern was only striking in a portion of the samples screened.

The predictive model described herein was built to offer researchers the ability to assess “germ line age” by accessing sperm DNA methylation signatures at genomic regions affected by age. Our data suggest that this model can predict an individual’s chronological age with a high degree of accuracy regardless of fertility status and with a high degree of repeatability. Additionally, our data suggest that the aging process in sperm may be impacted by environmental factors, though this effect appears to be quite subtle and future work is needed to establish this relationship.