Development of a microRNA-Based age estimation model using whole-blood microRNA expression profiling

Abstract

Age estimation is a critical aspect of human identification. Traditional methods, reliant on morphological examinations, are often suitable for living subjects. However, there are relatively few studies on age estimation based on biological samples, such as blood. Recent advancements have concentrated on DNA methylation for forensic age prediction. However, to explore further possibilities, this study investigated microRNAs (miRNAs) as alternative molecular markers for age estimation. Peripheral blood samples from 127 healthy individuals were analyzed for miRNA expression using small RNA sequencing. Lasso regression selected 103 candidate miRNAs, and Shapley additive explanations (SHAP) analysis identified 38 key miRNAs significant for age prediction. Five machine learning models were developed, with the elastic net model achieving the best performance (MAE of 4.08 years) on the testing set, surpassing current miRNA age estimation results. Additionally, we observed significant changes in the expression levels of miRNAs in healthy individuals aged 48–52 years. This study demonstrated the potential of blood miRNA biomarkers in age prediction and provides a set of miRNA markers for developing more accurate age prediction methods.