Figure 2

Pruning methods used to decrease the complexity of a result graph. Illustration of the different pruning methods implemented to reduce the complexity of the graph-based output. For all cases, edges are directed and equivalent, even if they represent different relationships. Arrows are shown where relevant. (a) The Zero Degree Vertex Pruner excludes nodes that have neither incoming nor outgoing edges. These normally appear once the other pruners have been applied. (b) The Root Children Pruner removes higher levels of the ontology that are deemed irrelevant to the enrichment analysis as they represent very general classes of objects (such as “chemical entity” or “ion”). (c) The Molecule Leaves Pruner removes all nodes of the graph that do not have any incoming edges and represent a distinct small molecule – and not a class of small molecules. A small molecule is defined as distinct if the node representing it has an InChI line notation. (d) The High p-value Branch Pruner looks for branches of the graph where only nodes with high p-value (>0.05) are present and removes those branches. (e) The Linear Branch Collapser Pruner finds linear stretches (nodes have only one incoming and one outgoing edge) and collapses them, leaving the nodes that surround those stretches connected.