GO terms are organized as a directed acyclic graph, in which a directed edge denotes a parent-child semantic relationship. A parent term might be significantly enriched only because it contains all the genes of a significantly over-represented child term. Consequently, the list of enriched GO terms is often too long and contains redundant terms, which hinders effective interpretation. Therefore, clusterProfiler integrates a simplify function to eliminate such redundant GO terms. This function employs the GOSemSim17 package to calculate semantic similarities among enriched GO terms using multiple methods based on information content or graph structure. Highly similar GO terms (e.g., >0.7) will be removed by applying the simplify function to retain a representative term (e.g., the most significant term). The following example shows an ORA on Biological Process (BP) to identify significant BP terms associated with the differentially expressed genes (DEGs). The geneList dataset, which contains fold change of gene expression levels between breast tumor and normal samples and is provided by the DOSE package, was used in this example. The DEGs were identified by a criterion of fold change >2.
