Reader-local minimum-peptides-per-protein filter. Counts the number of distinct ‘peptide_col’ values per ‘protein_col’ in ‘data’ and keeps only the rows whose parent protein reaches ‘nr_peptides’. The reader supplies the columns because only the reader knows which column holds the stripped (unmodified) peptide sequence.
filter_by_peptide_count(data, protein_col, peptide_col, nr_peptides = 1)
data
|
long-format quant table (before ‘setup_analysis’) |
protein_col
|
name of the parent protein column (single column) |
peptide_col
|
name of the (stripped) peptide column to count (single column) |
nr_peptides
|
minimum number of distinct peptides per protein (>= 1) |
Filtering the quant/peptide table is sufficient: ‘ProteinAnnotation’ is right-joined onto the ‘LFQData’ protein set (it seeds ‘row_annot’ from the ‘LFQData’ proteins and left-joins annotation onto it), so the annotation, the logged protein summary, and IBAQ all follow the filtered quant automatically – see the "ProteinAnnotation is annotation, not a filter" invariant in ‘AGENTS.md’. Apply this to the long table before ‘setup_analysis()’ / ‘LFQData$new()’.
A no-op when ‘nr_peptides <= 1’ (or ‘NULL’), so default runs are unaffected.
‘data’ with the rows of under-supported proteins removed
library("prolfquapp")
d <- data.frame(
prot = c("A", "A", "A", "B", "C", "C"),
pep = c("p1", "p2", "p2", "p1", "p1", "p2")
)
# A has 2 distinct peptides, B has 1, C has 2
nrow(filter_by_peptide_count(d, "prot", "pep", 2)) # drops B[1] 5nrow(filter_by_peptide_count(d, "prot", "pep", 1)) # keeps all[1] 6