Using Bayesian networks constructed on gene expression data we attempted to find the molecular underpinnings of asthma. For this discussion we define a Bayesian network as a probabilistic graphical model that represents a set of random variables (gene expression in this case) and their conditional dependencies (edges) via a directed acyclic graph (see Text S1). The workflow is illustrated in Figure S2. We first identified each gene from the GABRIEL GWAS dataset and collected all genes within 3 edges of that gene on the Bayesian Networks and found the largest coherent sub-network (i.e. the largest network with an observable path between every gene in the network) that contained the highest proportion of GABRIEL genes (Figure S3). We choose to use 3 edges so that we had enough genes to lead to a reasonable biological annotation of the genes in the network, i.e. limiting to 2 edges is too restrictive and results in a small number of relevant genes; on the other hand, using 4 edges leads to broad gene sets and makes it hard to pinpoint pathways. From this sub-network,
