Skip to contents

Builds a professionally formatted reactable table to display protein differential expression results. Works both in Shiny (it returns a reactable object) and standalone via results_list_widget().

Usage

results_list_reactable(
  data,
  protein_quant = NULL,
  comparisons = NULL,
  ain = NULL,
  alpha = 0.05,
  lfc_thr = 0,
  page_size = 15,
  height = 720,
  show_missing = TRUE,
  element_id = NULL,
  selection = NULL,
  searchable = TRUE
)

Arguments

data

Data frame or path to a TSV/CSV/Parquet file with DE results. Required columns: Protein.IDs, Gene.Names, logFC, P.Value, adj.P.Val, Change, Comparison. Optional: Assay, and the four missingness columns MissGlobal, MissComp, MissCND1 and MissCND2, which are shown only if all four are present

protein_quant

Data frame (preprocessing$protein_quant) or path to a Protein_QUANT_*.tsv file. If not NULL, the Description and Quant_Pepts columns are added.

comparisons

Vector of comparisons to include (NULL = all of them)

ain

Vector of assays to keep (NULL = all of them)

alpha

Significance threshold used to highlight the FDR (default: 0.05)

lfc_thr

log2 fold-change threshold (default: 0, reserved for future use)

page_size

Rows per page (default: 15)

height

Table height in pixels (default: 720)

show_missing

Show the missingness percentage columns (default: TRUE)

element_id

Element ID for the Reactable JS API (default: NULL)

selection

Selection type: "single", "multiple", or NULL (default: NULL)

searchable

Enable the internal search box (default: TRUE)

Value

A reactable object

Examples

data(nadia_dia)
res <- process_proteomics(nadia_dia, verbose = FALSE)

# protein_quant adds the Description and Quant_Pepts columns
tbl <- results_list_reactable(
  res$DEPs_results,
  protein_quant = nadia_dia$protein_quant,
  comparisons   = "B-A"
)
class(tbl)
#> [1] "reactable"  "htmlwidget"