Skip to main content
Fig. 2 | Genome Medicine

Fig. 2

From: XGR software for enhanced interpretation of genomic summary data, illustrated by application to immunological traits

Fig. 2

Necessity of respecting ontology tree-like structure and of removing redundant non-structured pathways in enrichment analysis. This is demonstrated by analysing differentially expressed genes induced by 24-h interferon gamma in monocytes. The effect of taking ontology tree-like structure into account is demonstrated using Disease Ontology (DO) and the removal of redundant non-structured ontologies using Reactome pathways. a Side-by-side bar plots comparing the significant DO terms between the analysis without considering the tree structure (DO Tree(-)) versus the analysis considering the tree structure (DO Tree(+)). The horizontal dotted line separates commonly identified terms (top section) and redundant terms in the DO Tree(-) analysis. b DAG plot comparing commonly identified terms (coloured in cyan) and redundant terms from the DO Tree(-) analysis (coloured in light cyan). The term name (if significant) is prefixed in the form ‘x1-x2’. x1 represents ‘DO Tree (-)’ and x2 ‘DO Tree (+)’. The value of x1 (or x2) can be ‘1’ or ‘0’, denoting whether this term is identified (present) or not (absent). c The top pathway enrichments, with the redundant pathways to be removed indicated (X). d Illustrations of whether a less significant pathway B is redundant considering a more significant pathway A. Pathway B is counted redundant if it meets both criteria. Criterion 1: more than 90% of input genes annotated with pathway B are also covered by pathway A. Criterion 2: more than 50% of input genes annotated with pathway A are also covered by pathway B. Scenario 1 does not meet either criteria, scenario 2 meets both, and scenario 3 meets criterion 1 but not criterion 2. Notably, criterion 2 ensures the resulting pathways (as shown in scenario 3) are informative in capturing knowledge spheres of different granularities; otherwise, pathway B would be considered redundant in scenario 3, leading to loss of information. FDR: false discovery rate

Back to article page