Fig. 1.

The posterior probabilities for each mutation to be associated interactively with indinavir treatment. The Upper shows the posterior probabilities using prior one, which assumes that it is equally likely (1/3) for a mutation to be unassociated, individually associated, and interactively associated with the drug treatment. The Lower shows the posterior probabilities using a more stringent prior (prior two) assuming that only two makers are expected to be associated with the drug, either individually or interactively.