Comparison of Bayesian and maximum-likelihood phylogenetic approaches in two legal cases involving accusations of transmission of HIV

We investigated two cases of alleged criminal transmission of HIV-1 using Bayesian and maximum-likelihood phylogenetic approaches to determine whether the inference method used influenced the outcome in these cases. In the first case, Bayesian methods were used to reexamine gag and env sequences from an earlier investigation in which the HIV-1 strains infecting one of several contacts could not be linked phylogenetically to that of the accused despite strongly suggestive epidemiological evidence. In the second case, maximum-likelihood and Bayesian inference methods were used to investigate the relatedness of gag and env sequences from HIV-1 strains infecting a man accused of intentionally transmitting the virus to several contacts. Bayesian analysis of HIV-1 strains from the first case confirmed earlier results obtained by maximum-likelihood analysis. A monophyletic cluster linking viruses from the accused and three of his direct and indirect contacts was supported, but a linkage between these viruses and a fourth epidemiologically linked contact could not be demonstrated. In the second case, a strong virological link between the accused and two of his contacts, and the absence of links with four other contacts, was confirmed by both maximum-likelihood and Bayesian inference methods. It is important that phylogenetic programs applied in a legal setting are conservative in their outcome. Although Bayesian methods offer computational tractability for large data sets and complex evolutionary models, this study demonstrates they do not assist when clear linkages between viruses are demonstrated using maximum-likelihood methods.