gsea software validation
The usage snippet shows the supported vegas()-to-gsea() mapping. Focused deterministic tests cover each supported dispatcher route.
These recorded simulation results assess selected behavior on ten fixed datasets from the shared design. They do not revalidate the statistical method or cover every setting. The use snippets show the current package interfaces. This report step summarizes retained artifacts and does not rerun analyses. The gsea guide describes current inputs and options.
Use
library(gsea)
evidence <- vegas(statistics, LD_Glist, sets = genes,
metadata = metadata,
control = list(tail_method = "saddlepoint"))
input <- gsea_input(evidence)
fit <- gsea(input$stat, pathways, method = "preranked")Competitive, MAGMA and Bayesian pathway fits require a separately prepared gene covariance. VEGAS does not infer cross-gene independence.
Results
| Method | Enriched p < 0.05 | Background p < 0.05 |
|---|---|---|
| ora | 15/30 | 0/30 |
| preranked | 30/30 | 1/30 |
| competitive | 30/30 | 1/30 |
| magma | 26/30 | 0/30 |
The scalar BayesC pathway analysis reports posterior inclusion probabilities on a separate scale:
| Pathway | Mean PIP |
|---|---|
| enriched | 0.828 |
| background | 0.066 |
| mixed | 0.048 |
The controlled gene-tail calculation was available for 29945/30000 requested tests; the documented saddlepoint route was available for 30000/30000. Its moment fallback was used 12 times.
The enriched and background sets are artificial and fixed before analysis. Counts are descriptive across dependent traits and pathways; they are not power or false-positive-rate estimates. PIPs and p-values have different interpretations.