BRM5012 Chap.5 Functional Enrichment and Ranked Gene Sets
Functional Enrichment and Ranked Gene Sets
Functional enrichment organises long differential-expression outputs into testable biological hypotheses. This chapter contrasts over-representation analysis of a thresholded gene list with gene set enrichment analysis of a complete ranking.
It explains background selection, separate directional lists, Wald statistic ranking, leading genes, database differences and how to compare tools transparently without searching until a preferred pathway appears.
What this chapter covers
- 01
Selected lists for ORA
- 02
Background gene universe
- 03
Directional subsets
- 04
Full rankings for GSEA
- 05
Wald statistic
- 06
Leading genes
- 07
Database and algorithm differences
- 08
Biological interpretation
Worked example · free
Rank genes with the Wald statistic
- +1Calculate Gene A as 1.8 divided by 0.9, giving a Wald statistic of 2.0.
- +1Calculate Gene B as 1.2 divided by 0.3, giving a Wald statistic of 4.0.
- +1Rank Gene B above Gene A because its estimated change is more precise relative to uncertainty.
- +1Keep the complete tested list and declared comparison direction for GSEA.
Key terms
- Gene universe
- The set of genes that were genuinely available for selection and form the background for ORA.
- Wald statistic
- The estimated log fold change divided by its standard error, combining direction, size and uncertainty.
- Leading genes
- The subset of ranked genes contributing most strongly to a gene set enrichment signal.
- Orthogonal analysis
- A complementary method or evidence source used to test whether an interpretation is robust to approach.
Functional Enrichment and Ranked Gene Sets FAQ
Why does ORA need a background list?
ORA compares the observed overlap with what would be expected from genes that could have been selected. Using genes that were never measured or tested changes the expected frequency and can distort enrichment. Use the actual tested universe after relevant filters.
Why should GSEA use a complete ranking?
GSEA detects coordinated shifts across the ordered result, including genes that do not cross an individual significance threshold. Filtering first discards that structure. Rank the full tested list with the stated statistic and preserve direction.
How should I compare enrichment tools?
Record identical inputs, backgrounds and settings, then compare leading genes, direction, database versions and biological themes. Exact pathway names can differ because databases and algorithms differ. Agreement is useful only when the comparison is transparent.
Assessment move
Create two input cards: ORA receives a selected directional list plus a tested-gene background, while GSEA receives the complete ranking. Practise explaining why those inputs answer different questions. Recompute Wald statistics and verify sign before exporting a list. For every enriched term, identify the leading genes, direction and relationship to the original research question.
Group overlapping ontology labels instead of treating every name as independent evidence. When comparing tools, keep inputs and backgrounds fixed, record database versions and examine whether the same biological process is supported by different gene members. End enrichment answers with a validation step, such as an independent dataset, targeted assay, perturbation or proteomic comparison.
This keeps pathway results in their proper role as structured hypotheses rather than causal proof. Construct one synthetic differential-expression table and derive both enrichment inputs from it. For over-representation analysis, state the significance and direction rule, split positive and negative lists when appropriate, and define the background as genes that were actually tested after measurement filters.
Recalculate how overlap would look if every annotated gene were used, then explain why inaccessible genes distort the expected universe. For ranked analysis, remove the selected-list threshold and calculate several Wald statistics from log2 change and standard error. Check the sign against the contrast before sorting the complete list. Practise explaining why a precise moderate change can outrank an uncertain large change.
When results arrive, open the pathway membership rather than stopping at the term name. Mark the leading genes, their directions and whether a small shared core generates several related labels. Build a term map that groups overlapping biological ideas while preserving distinct evidence.
Compare two tools with the same input, background, direction and database date; differences can then be discussed in terms of curation or algorithm rather than hidden setup changes. Record unmapped identifiers and database versions because absent annotation limits what the analysis can discover.
Write a three-sentence interpretation: the coordinated process, the gene-level evidence that drives it, and the connection to the original comparison. Follow it with a competing explanation and an independent validation. Suitable next steps include another cohort, a targeted assay, proteomic comparison or perturbation.
The exercise is complete only when the pathway result returns to the research question without being promoted to causal proof.
Working through Functional Enrichment and Ranked Gene Sets in BRM5012? Sia is AskSia’s AI Biology tutor — ask any BRM5012 Functional Enrichment and Ranked Gene Sets question and get a clear, step-by-step explanation grounded in how BRM5012 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
