9 featureMSEA Analysis

9.1 Step 4 — Run enrichment analysis

perform_fmsea_analysis() ranks features by the condition statistic, computes a running enrichment score for each metabolite set, and estimates significance via permutation testing. Results are returned as a fmsea S4 object.

# Supported metabolite set databases: KEGG, SMPDB, IMETPD, Reactome, Wikipathway
fmsea_result <- perform_fmsea_analysis(
  pathway_database,           # metabolite set database object (from Step 1)
  annotation_table,           # from process_annotation_table() in Step 3
  ranking_table,              # from process_annotation_table() in Step 3
  threads          = 6,
  min.compounds.num = 15,
  max.compounds.num = 300,
  id.col           = "KEGG_ID",
  perm.num         = 10000,
  seed             = 123,
  fdr.thr          = 0.05,
  max.iter.num     = 3,
  verbose          = TRUE
)

# Significant pathways are stored in:
fmsea_result@significant_modules

9.2 Parameters

Parameter	Description	Default
pathway_database	Metabolite set database object. Its database_info$source must be one of "KEGG", "SMPDB", "IMETPD", "Reactome", "Wikipathway"	—
annotation_table	Output from process_annotation_table() — links features to candidate metabolites	—
ranking_table	Output from process_annotation_table() — provides the ranking statistic per feature	—
threads	Number of CPU threads for parallel computing. NULL runs serially	NULL
min.compounds.num	Minimum pathway size; smaller pathways are excluded from testing	15
max.compounds.num	Maximum pathway size; larger pathways are excluded from testing	300
id.col	Compound ID column used to match features to pathway members, e.g. "KEGG_ID" for KEGG or "HMDB_ID" for HMDB-based databases	"KEGG_ID"
perm.num	Number of permutations for building the null distribution. Higher values give more stable p-values at the cost of runtime	1000
seed	Random seed for reproducible permutation testing	123
fdr.thr	FDR threshold; pathways with FDR below this value are kept in significant_modules	0.05
max.iter.num	featureMSEA re-weights annotations based on significant pathways and re-runs enrichment iteratively; this caps the number of iterations	3
verbose	Print progress messages during the run	TRUE

Note: Increasing perm.num improves p-value precision but increases runtime proportionally. 1000 permutations gives a good balance for exploratory analysis; use 10,000+ for publication-quality results.