Skip to main content

Connect biological knowledge

This part treats GO, KEGG, Reactome, disease, MeSH, pathway collections, and custom annotations as knowledge sources for one common analysis system. The analysis engine from Part 1 remains the same while the vocabulary, annotation coverage, update schedule, and biological interpretation change.

A gene set is not necessarily a pathway

The collections tested in this part split into two kinds, and the difference is what the analysis is allowed to conclude:

  • Descriptive collections — GO terms, pathways, disease and MeSH concepts — answer what these genes do or what they are associated with. The answer describes the input.
  • Pointing collections — transcription factor targets, microRNA targets, protein complexes, drugs, cell-type signatures — answer which upstream participant or state could have produced these genes. The answer proposes a mechanism to be tested.

Both are run with the same enricher() or GSEA() call, and both return the same result object; only the reading of a term changes. Upstream regulator and participant enrichment is the chapter for the second kind, and is the route to take when the goal is a hypothesis about what drives the list rather than a description of it.

How to choose a knowledge source

Need Start with
Broad ontology-based functional coverage GO
A hypothesis about the upstream regulator, complex, drug, or cell state behind the list Regulator and participant enrichment
Curated pathways, modules, or compounds KEGG or Reactome
Disease or phenotype concepts Disease enrichment
Literature-oriented biomedical concepts MeSH
A project-specific, unsupported, or multi-resource collection Universal enrichment and GSON
Genomic regions that must first be mapped to genes Genomic coordination enrichment

The shared knowledge contract

Every source should make its term-to-gene mapping, term names, gene identifier type, gene set identifier domain, organism, release, and filtering rules explicit. A two-column TERM2GENE table is enough for an exploratory analysis; a versioned GSON object is preferable when the knowledge will be reused, shared, combined, or cited, because it carries these declarations as slots rather than leaving them in the notebook that produced it.

The domain chapters use the same pattern:

biological question → source and identifier contract → ORA/GSEA
                   → result object → visualization → limitations

Knowledge sources are not interchangeable evidence. GO terms, KEGG pathways, disease terms, and WikiPathways entries can differ in granularity and coverage. Combined plots should keep the source identity visible.