The interpretation of complex DNA profiles remains a major challenge in forensic genetics, especially when dealing with mixed and low-template samples, such as touch DNA. This study aimed to comparatively evaluate the performance of different probabilistic genotyping software (PGS) for the interpretation of complex DNA mixtures, under both controlled laboratory conditions and in more realistic trace DNA scenarios.

The study was structured in two complementary phases. Firstly, reference DNA mixtures composed of known contributors at defined ratios were generated in the laboratory to characterize the analytical behavior of each PGS under controlled conditions. This preliminary phase was designed to evaluate sensitivity to minor contributors, accuracy of contributor proportion estimates, and deconvolution performance, while also providing a framework for interpreting more complex samples. In the second phase, touch DNA samples were collected from selected car surfaces, after repeated manipulation under simulated usage conditions, to reproduce realistic patterns of DNA contributor transfer, persistence, and overlap.

Semi-continuous and continuous PGS (LRmix Studio, DNAStatistX, and EuroForMix) were compared in terms of likelihood ratio (LR) estimation, contributor proportion assessment, and deconvolution capability. Under ideal conditions, continuous models showed high accuracy in contributor proportion assessment, with a strong linear correlation between expected and observed values (r=0.99, R²=0.99) and a relatively low mean absolute deviation (3.79%). Continuous models generally showed enhanced discriminative performance, producing more stable and extreme LR-values for correct contributor hypotheses compared to semi‑continuous models in both study phases.

Systematic differences were also observed among continuous PGS, including slight variations in contributor proportion estimates and a tendency for DNAStatistX to provide more conservative assessments under low-template and highly unbalanced conditions (adjusted p<0.0001). In parallel, the informativeness of touch DNA profiles was influenced by sample-related factors, particularly the sampled surface and the degree of handling. Specifically, LR-values were significantly higher for medium- and high- contact surfaces compared to low- contact ones (adjusted p<0.0001). Within this context, minor contributors became progressively more difficult to resolve as mixture imbalance increased; however, generally, differentiation between major and minor contributors remained achievable and supported a consistent interpretation of contributor dynamics.

Overall, this study provides a comparative methodological framework for evaluating PGS across increasing levels of mixture complexity and highlights the importance of integrating controlled baseline data with context-dependent interpretation in forensic genetics. This supports a more critical and informed application of PGS in the interpretation of complex touch DNA evidence.

Authors

• Arianna Delicati (Legal Medicine Unit and Unit of Biostatistics, Epidemiology and Public Health, Department of Cardiac, Thoracic, Vascular Sciences and Public Health, University of Padua, Italy)
• Dolores Catelan (Unit of Biostatistics, Epidemiology and Public Health, Department of Cardiac, Thoracic, Vascular Sciences and Public Health, University of Padua, Italy)
• Beatrice Marcante (Legal Medicine Unit and Unit of Biostatistics, Epidemiology and Public Health, Department of Cardiac, Thoracic, Vascular Sciences and Public Health, University of Padua, Italy)
• Pamela Tozzo (Legal Medicine Unit, Department of Cardiac, Thoracic, Vascular Sciences and Public Health, University of Padua, Italy)
• Luciana Caenazzo (Legal Medicine Unit, Department of Cardiac, Thoracic, Vascular Sciences and Public Health, University of Padua, Italy)