Illustration of different dependency types between variables in a Bayesian Network structure.

<p><i>Pairwise relevance relations</i>: Direct causal relevance (e.g., Y1 and SNP1 have common edge), Transitive causal relevance (e.g., there are two directed paths between Y3 and SNP5), Confounded relevance (e.g., Y2 and SNP3 have a common ancestor SNP1), Association (e.g., Y1 and SNP1, because SNP1 is directly related to Y1; Y3 and SNP5, because SNP5 is transitively related to Y3; Y2 and SNP3, because they are in a confounded relation), Pure interactionist relevance (e.g., Y1 and SNP7 have common child), Strong relevance (e.g., Y1 and SNP1, because SNP1 is directly related to Y1; Y1 and SNP7, because they are in pure interaction). <i>Relevance of variable sets</i>: Strong relevance (e.g., the variable set consisting of Y2's parents, its children, and the other parents of its children is {Y1, SNP9, Y3, SNP7}). <i>Relevance for multiple target variables</i>: Strong relevance to one or more targets (e.g., the variable set consisting of {Y1,Y2,Y3}'s parents, its children, and the other parents of its children is {SNP1, SNP4, SNP7, SNP9}). <i>Red nodes</i>: potential target variables, <i>Green nodes</i>: SNP variables.