Propositions used to assess the value of forensic DNA mixtures in an FSWG-ISFG interlaboratory comparison

DNA interpretation relies on the evaluation of results under at least two mutually exclusive propositions. This study evaluates the application of the International Society for Forensic Genetics (ISFG) recommendations concerning the formulation of such propositions across 15 laboratories in six countries and examines how evaluations are conducted when two persons of interest are considered. They were asked to assess results considering propositions about the source of the DNA during an interlaboratory comparison organized by the French Speaking Working Group of the ISFG. This article focuses on a DNA mixture from a mock case involving a complainant and two persons of interest. Seven of the eight ISFG recommendations were applicable to this interlaboratory comparison, with six being implemented by all laboratories. However, when two persons of interest were submitted for comparison without further case information, only seven laboratories followed Recommendation 3 by assigning a different likelihood ratio (LR) to each potential contributor. One of them used multiple propositions (more than two mutually exclusive propositions) and considered that each person, in turn, could or could not be the source of the DNA with or without the other person. The eight remaining laboratories assigned only one LR considering that both persons were contributors, or neither. As stated in the ISFG recommendations, such a practice should be avoided as it could lead to an overestimation of the LR for one of the contributors. We also demonstrate the effect of considering, or not, the presence of the DNA of persons whose contribution to the DNA mixture was not disputed (i.e., "conditioning" the DNA results on the DNA profiles of the undisputed source) on the LR. When conditioning is applied in ground truth experiments, the results provide stronger support for the proposition known to be true compared to the alternative. More precisely, the LR increased by a factor of 100–10’000 when conditioning, depending on the laboratory. The LRs assigned in two real cases are presented to illustrate the need to consider new information, such as the presence of a potential contributor to a DNA mixture, when evaluating results, and multiple propositions when several persons of interest are considered. It can significantly change the LR value.