Thesis title: Validation of an Internal Method for the interpretation of mixed genetic profiles and validation of a Probabilistic Genotyping Software within the Forensic Genetics Laboratory of the Forensic Scientific Police Service
The present doctoral research focuses on the validation of an internal method for the analysis of mixed DNA profiles and the implementation of probabilistic genotyping software (PGS) within the Forensic Genetics Laboratory of the Forensic Science Police Service in Rome. The project is a response to the growing need to extend the laboratory's ISO/IEC 17025 accreditation—previously limited to single-source DNA profiles—to the analysis of mixed profiles, which are increasingly common in forensic casework due to advances in DNA recovery from degraded or low-template biological samples.
The research was developed in two principal phases. The first concerned the validation of the complete analytical workflow for the interpretation of mixed genetic profiles involving two and three contributors. In the laboratory, ad hoc mixtures were prepared in the laboratory using buccal swabs from selected individuals, ensuring defined allelic compositions and mixture ratios. These samples were analysed using the GlobalFiler™ PCR Amplification Kit, capillary electrophoresis on the Applied Biosystems 3500 Genetic Analyzer, and GeneMapper® ID-X software, in accordance with internal procedures. The study also verified whether analytical thresholds values validated for single profiles could be extended to mixed profiles. The findings corroborated the adequacy of pre-existing thresholds that had been validated, encompassing both low-template thresholds (LT-DNA) and conventional samples thresholds (cDNA).
The subsequent phase pertained to the validation of probabilistic genotyping software, with the objective of complementing the classical, expert-based interpretation of mixed profiles with an objective, likelihood ratio (LR)–based probabilistic assessment. Following a comparative evaluation between semi-continuous and continuous models (specifically LRmixStudio, STRmix™, and EuroForMix), the latter was identified as the most appropriate for the laboratory's operational needs. EuroForMix was consequently subjected to a thorough internal validation process in accordance with the SWGDAM recommendations (2015) with a view to evaluating its performance and limitations. Validation datasets included laboratory-generated mixtures, in silico samples derived from the PROVEDIt database, and real forensic cases, representing the full range of complexities encountered in practice.
Validation tests examined reliability, precision, sensitivity, and specificity. Repeated analyses demonstrated high reproducibility of LR values. Sensitivity assessments confirmed the software’s robustness even in cases of partial profiles or stochastic effects for true contributors, while specificity testing verified that non-contributors consistently yielded LR values below 1, thus minimizing false inclusions.
The results of the sensitivity and specificity test demonstrated that the EuroForMix software is well calibrated.
This study shows that although PGS tools should be considered as complementary aids rather than substitutes for expert judgment, the integration of experimental validation and probabilistic approaches can improve the quality and transparency of forensic genetic analyses, thanks to a better and more accurate interpretation of complex genetic profiles.