Age estimation using bloodstain miRNAs based on massive parallel sequencing and machine learning: A pilot study

Abstract

Age estimation is one of the most important components in the practice of forensic science, especially for body fluids or stains at crime scenes. Recent studies have focused on the application of DNA methylation for chronological age determination in the field of forensic genetics. However, the amount of DNA and the complex bisulfite conversion process make applying this method in trace or degraded samples difficult. MicroRNAs (miRNAs), a group of small noncoding RNAs, have great potential in forensic science due to their antidegradation property and tissue specificity. Certain miRNAs are highly age-related and may have potential utility in age prediction. In this study, the expression profile of miRNAs from blood samples was explored using massive parallel sequencing; age-related miRNAs were subsequently selected for age prediction. We then established age prediction models for bloodstains based on six age-related miRNAs using seven machine learning models. Results revealed that the mean absolute error (MAE) was 5.52 and 7.46 years in male and female bloodstain samples, respectively, using the AdaBoost algorithm. This pilot study demonstrates the possibility of performing forensic age prediction using miRNAs and may provide useful information in future case investigations.

Keywords: Age estimation; Bloodstain; Machine learning; microRNA.