Category evidence

Explore the gene sets and genes supporting this category.

Set × leading-edge gene connections

Select a set or gene to highlight its connections.

Positive NES Negative NES · dots show GSEA leading-edge membership, not complete set membership.
Positive gene DE effect Negative gene DE effect Missing DE; numerical DE FDR is shown separately.

Sorts apply before the display limits. Recurrence counts use every significant unique set in this category, not just visible sets. Matching direction does not make NES and gene log2FC interchangeable; opposing genes are never hidden. Display limits do not remove downloadable rows. No combined gene score is calculated.

Complete evidence tables

Data, methods & interpretation

These TSVs cover the full selected analysis/collection/tier, not just the current display. The standardised DE table is stored once in genes.tsv; category connections reference exact gene symbols. ORA overlaps are not mixed into GSEA leading edges.

Leading-edge Jaccard = intersection / union of unique leading-edge genes. Overlap tables report both set sizes, intersection, union and NES direction. The default displayed_leading_edge_overlaps.tsv covers only the default displayed significant sets, not all mapped sets. Select a set above to compare it with every mapped set and download those rows. An optional all_leading_edge_overlaps.tsv is produced only when full all-pair export is explicitly requested. It is NA when either membership is empty or unrecorded, not evidence of dissimilarity. Complete-set overlap, if supplied, follows the same explicit pair scope in a separate export and must not be interpreted as leading-edge overlap. Overlap is descriptive redundancy, not causal evidence.

“Non-evaluable” means valid NES and adjusted P value were unavailable or the set was ineligible. “No significant sets” and absent leading-edge genes are explicit states, not evidence that a category is inactive. The assigned category is semantic metadata; inspect the original sets and gene-level measurements.

Gene sets in this category

Original gene-set NES and FDR. The figure shows significant sets; complete states remain in the data table.

Category significance and support

Hommel is a multiple-testing correction: robust Hommel closed testing accounts for the many gene sets assessed together. It limits false positives by controlling the chance of even one false positive within one analysis (familywise error), assuming valid input P values.

What does the adjusted category P mean?

The adjusted category P produced by Hommel tests whether none of the evaluable gene sets in this category is enriched. A small adjusted P gives evidence that at least one is enriched. The stars on the figure use this adjusted P: * for P ≤ 0.05, ** for P ≤ 0.01, *** for P ≤ 0.001, and **** for P ≤ 0.0001. Only the strongest level reached is shown. No stars means P > 0.05 or not evaluable; the category card distinguishes these cases.

What does minimum support mean?

The category card also shows d/N: among N eligible gene sets with a usable raw P value, at least d can be supported as enriched with 95% simultaneous confidence. Here is an invented example, not a study result: 5/20 means at least 5 of the 20 sets, not exactly 5; more may be enriched. The percentage is the same minimum expressed as a fraction, not an additional test. If d=0, no positive minimum is supported; this does not mean no enrichment. If N=0, the category cannot be evaluated. Check evaluated / total coverage in the card.

The adjusted P and d/N answer different questions: evidence for at least one enriched set versus a simultaneous lower bound on how many are enriched. Neither measures effect size, proves the whole biological process is activated, establishes direction, or tests a difference between conditions. The separate NES panel describes observed directions among individually significant sets. Overlapping gene sets are not independent mechanisms.

Method: robust Hommel closed testing. Goeman et al. (2019), doi:10.1093/biomet/asz041.