Ensaya: An ensemble age model for prediction of chronological age in adolescents and young adults
Forensic Science International: Genetics · Published 30 June 2026
In plain language
Our DNA carries chemical markers that change predictably with age, and a simple blood sample can be used to read them. This study tested the existing "epigenetic clocks" on 687 samples from 12 groups of people aged 12 to 25, then combined the three best into a new model called Ensaya. It estimated age to within about one year on average, and stayed reliable across laboratories and testing methods, which matters most around the legal age of 18.
Abstract
DNA methylation-based models for chronological age estimation have, for over a decade, been among the most important methods for age prediction in humans. While large-scale data have enabled the development of numerous models, their accuracy varies substantially depending on choice of machine learning algorithm, preprocessing pipeline, age range, tissue type, and population characteristics. Accurate age estimation in adolescents and young adults is particularly relevant in legal, investigational, and biomedical contexts. Here, we systematically evaluated blood-based DNA methylation age predictors in individuals aged 12–25 years and developed Ensaya (Ensemble age model for prediction of chronological age in adolescents and young adults), an ensemble model optimized for this age range. We assembled a validation dataset of 687 whole blood or peripheral blood mononuclear cell samples from 12 cohorts, profiled using Illumina 450 K, EPIC, and EPIC v2 arrays. We compared existing clocks and assessed robustness across cohorts, platforms, normalisation procedures, technical replicates, sex, and circadian timing. Among individual models, PAYA, cAge, and Garma showed the strongest performance and were combined into the Ensaya ensemble. In validation, Ensaya achieved improved accuracy, with a median absolute deviation of 0.79 years and a mean absolute error of 1.04 years, along with reduced age-dependent bias. It also demonstrated high technical reproducibility and stable performance across platforms and normalisation methods. These findings show that ensemble-based modelling improves age estimation in adolescents and young adults, supporting Ensaya as a practical blood-based predictor.
Authors
Kristina Stenløkk, Pål Skage Dahlberg, Veslemøy Rolseth, Herman Bernitz, Mohamed Yasin Gamieldien, Mona Nahas Gholmieh, Liam Robinson, Antoine Saadé, Andre Uys, Øyvind Bleka
Citation
Stenløkk K, Dahlberg PS, Rolseth V, Bernitz H, Gamieldien MY, Gholmieh MN, Robinson L, Saadé A, Uys A, Bleka Ø (2026) Ensaya: An ensemble age model for prediction of chronological age in adolescents and young adults. Forensic Science International: Genetics. Vol. 86, 103579. https://doi.org/10.1016/j.fsigen.2026.103579
Abstract and PDF reproduced under the CC BY 4.0 licence. Summary written by Dental Experts Center.
