This study extends the current use of Bayesian networks, a graphical and numerical representation that enables reasoning about uncertainty, by incorporating the effects of allelic dependencies in paternity calculations.
The networks presented in this paper provide graphical representations for various paternity cases and also provide a computational alternative to sometimes laborious hand calculations; however, additional functionality must be provided before Bayesian networks are to remain valuable tools for the forensic scientist. The most likely scenarios in which Bayesian networks would prove invaluable are complex paternity cases with several observed and unobserved founders in a pedigree. These are the cases where hand calculation is nearly impossible, requiring a computational tool. Unfortunately, these are also the cases in which the methods as proposed in this paper break down. Additional research is needed regarding the use of HUGIN's API, which is a promising solution to many of the problems encountered in the current study. Another disadvantage of the methods presented in this paper are the limiting assumptions. Specifically, assuming the population allele frequencies and theta values are known is particularly disconcerting. This is rarely the case. An area for future research should be the investigation of a relaxation of these assumptions while allowing for uncertainty in their values. This study explored two paternity examples: the most common scenario in which DNA evidence was available from the alleged father, the mother, and the child; and a more complex case in which DNA was not available from the alleged father, but was available from the alleged father's brother. Object-oriented networks were built, using HUGIN, for each example, which incorporate the effects of allelic dependence caused by evolutionary relatedness. 2 tables, 10 figures, and 31 references