Predicting Physical Appearance from Low Template: State of the Art and Future Perspectives

The document reviews advancements in forensic DNA phenotyping (FDP) from low-template DNA (LT-DNA), emphasizing the role of machine learning (ML) in predicting physical traits like eye, hair, and skin color. While tools like HIrisPlex-S achieve high accuracy for eye color, challenges persist with skin tone and complex traits in diverse populations. A study with 155 samples showed that DNA concentrations above 10 pg/μL improve profile informativeness. Emerging technologies like massively parallel sequencing enhance LT-DNA analysis, though ethical concerns remain. The document stresses the need for diverse datasets and improved models to enhance predictive accuracy and forensic applicability, highlighting the importance of ethical considerations and transparency in forensic applications.