While STR profiling remains the primary tool for forensic identification, it is limited when no reference sample is available. In such cases, predicting demographic characteristics — including biogeographic ancestry, age, and externally visible traits — from crime scene DNA can provide valuable investigative leads. However, obtaining multiple types of information from a single limited sample remains challenging, as each characteristic typically requires separate marker sets and sufficient DNA input. To address this, we developed two targeted capture panels — one for 917 CpG methylation markers and one for 601 SNPs using Twist hybrid capture approach. 

The methylation panel includes markers for body fluid identification, tissue identification, age estimation, and lifestyle prediction using both published and in-house models. The SNP panel includes ancestry- and phenotype-informative markers from ForenSeq, Precision ID, Snipper, HirisPlex-S, and the VISAGE Enhanced Tool. Both panels were evaluated through MiSeq sequencing of 96 samples across seven body fluid and tissue types (using 200 ng gDNA input for the methylation panel), and 101 blood DNA samples from 18 population groups (using 50 ng gDNA input for the SNP panel).

The methylation capture panel achieved a mean coverage of 330×, with 915 of 917 targets covered at ≥10×. Body fluid and tissue identification were accurate across most sample types. Age estimation was performed using tissue-appropriate models. Though array- and SBE-based models demonstrated high correlations, MPS-based models consistently achieved the best performance with the lowest MAEs (2.27–3.83 years in blood, r ≥ 0.95; 2.46–3.36 years in semen, r ≥ 0.97). Lifestyle prediction achieved 86.7–93.3% accuracy for smoking and BMI classification. 

The SNP capture panel covered 600 of 601 targeted markers with a mean coverage of 474×, and locus and heterozygous balance were both close to 1.0. Ancestry analysis across multiple platforms (Snipper, ForenSeq, Precision ID, VISAGE, and SNU platform) clearly separated major continental populations, although some overlap was observed near population boundaries. HirisPlex-S phenotype predictions for hair and eye color were largely consistent with the expected phenotypic distributions of each population group, whereas skin color predictions showed broader distributions, especially in East and Southwest Asian populations, suggesting the need for identification of more population-specific markers.

Both panels showed stable performance across diverse body fluids and tissues, with high coverage and consistent results. Targeted capture–based workflows offer an efficient means of generating leads from unidentified samples. As forensic casework becomes more diverse, these panels would strengthen DNA-based investigative support.

Authors

• Leejae Cho (Seoul National University College of Medicine, South Korea)
• Moon Hyun So (Seoul National University College of Medicine, South Korea)
• Hwan Young Lee (Seoul National University College of Medicine, South Korea)