Refined Group-Level False Discovery Rate (FDRr) Calculation
refined_gfdr.RdThis function computes refined False Discovery Rate (FDR) estimates at the protein
group level using a competitive approach. It extends refined_fdr by handling
multiple proteins per identification in the target-decoy competition.
Arguments
- data
A data frame containing identification data. It must include:
proteinMaster: the main protein identifier for the group.proteinRefs: all proteins associated with the identification, separated by ";".isDecoy: logical indicating whether the identification is a decoy.A score column used for ranking identifications.
- score
The column name of the score used to rank the identifications. This should be an unquoted column name.
- lower_better
Logical; if TRUE, lower scores indicate better identifications (default
TRUE).- affix
String indicating the suffix/prefix used to identify decoy entries (default
"_REVERSED").
Value
A data frame containing the original data along with additional columns:
- FDRn
Normal FDR estimation as cumulative minimum (q-value).
- FDRp
Picked FDR estimation as cumulative minimum (q-value).
- FDRr
Refined FDR estimation as cumulative minimum (q-value).
- to
Target-only identifications count at the protein group level.
- do
Decoy-only identifications count at the protein group level.
- td
Count of identifications with equal target and decoy scores.
- tb
Target-best identifications count at the protein group level.
- db
Decoy-best identifications count at the protein group level.
Details
The function first expands multiple protein references per identification into individual rows, removes decoy affixes, and then determines which protein references are in competition with the proteinMaster. It calculates the FDR estimates by summing target and decoy identifications across regions (to, do, td, tb, db), following the competitive approach described in the cited paper.