First-Generation Epigenetic Clocks
First generation clocks have been the most widely studied to date in the studies of epigenetic aging. They are broadly generalizable to different populations, and their accuracy can be easily assessed. Despite only being trained on CA, the EAA calculated by these clocks has been implicated in the incidence of many diseases14. First generation clocks can be further classified into single and multi-tissue clocks. In this review, we will discuss in detail the first-generation clocks that have broad utility (e.g. the Hannum clock67, the Horvath pan-tissue53 and Skin & Blood clocks68, the PedBE clock69). The features of other clocks tailored to narrower use cases (e.g., those aimed at studying specific tissues or primarily focusing on an age group such as neonates), are summarized in Table 2 . The accuracy of first-generation epigenetic clocks is assessed in relation to CA, usually using Absolute Error (AE=|epigenetic-chronological age|) or Pearson’s correlation coefficient (r ) between EA and CA. The EA calculated by these clocks is correlated with CA but the deviation between the two has been shown to be informative of ‘biological capacity’ (e.g. physical fragility, disease susceptibility) in adults65. First-generation clocks represent facets of both chronological and biological aging; separating these components remains a major challenge65.
Bocklandt et al developed the first epigenetic clock in 2011, marking a milestone in the field. However, this clock was aimed specifically at saliva samples and due to the tissue specificity of DNAm has not been generalized to other sample types (Table 2) 57. Shortly thereafter, the Hannum67 clock – a broadly used blood epigenetic clock – was published. Using Illumina 450K array data from 656 samples (482 training set and 174 testing set) of whole blood (age range: 19-101years), the Hannum clock was developed in stages67. In the first step, ~70,000 age associated autosomal CpG sites were identified. Then, elastic net regression with bootstrapping was performed to build an epigenetic clock consisting of 71 CpGs, most of which were located close to genes implicated in age-related conditions67(Table 2 ). When applied to pediatric samples (CA <18years) the Hannum clock demonstrated low accuracy61, potentially due to a lack of pediatric samples in the dataset used to develop this clock61.