Bayesian networks represent the association between many variables using conditional probability distributions and rely on Bayes’ theorem to show how changes in one or more variables in the networks affect other variables [127]. In this way, they can be used prognostically to compute the probability of the outcome (say a specific fetal hemoglobin range) of an individual given his genetic profile. They can also be used diagnostically, to discover the genetic profiles that maximize the probability of a particular outcome, and therefore be used to study how patterns of behavior can interact with the genetic profiles to shape the phenotype [128]. In this way, they appear to be able to simultaneously cast light on novel biological findings and be a prognostic tool for personalized medicine [46]. The Bayesian network that we developed for the genetic dissection of stroke in sickle cell anemia (see Fig. 6) offers an example of the power of these models. The network captures the interaction between 31 SNPs in 12 genes that, together with fetal hemoglobin, modulate the risk for stroke. We showed the prognostic accuracy
