DNA mixtures in forensic investigations: The statistical state of the art

Forensic science has experienced a period of rapid change because of the tremendous evolution in DNA profiling. Problems of forensic identification from DNA evidence can become extremely challenging, both logically and computationally, in the presence of complicating features, such as in mixed DNA trace evidence. Additional complicating aspects are possible, such as missing data on individuals, heterogeneous populations, and kinship. In such cases, there is considerable uncertainty involved in determining whether or not the DNA of a given individual is actually present in the sample. We begin by giving a brief introduction to the genetic background needed for understanding forensic DNA mixtures, including the artifacts that commonly occur in the DNA amplification process. We then review different methods and software based on qualitative and quantitative information and give details on a quantitative method that uses Bayesian networks as a computational device for efficiently computing likelihoods. This method allows for the possibility of combining evidence from multiple samples to make inference about relationships from DNA mixtures and other more complex scenarios.