Hi #CE8SSJV3N #C0493FQQZ3P, I am writing to inquiry about stageR and DEXSeq. I apply stageR after DEXSeq for differential transcript usage testing.

library(stageR)

stageRObj <- stageRTx(pScreen = pScreen, pConfirmation = pConfirmation,
                      pScreenAdjusted = FALSE, tx2gene = tx2gene)

stageRObj <- stageWiseAdjustment(stageRObj, method = "dtu", alpha = 0.05)

stageR_results <- getAdjustedPValues(stageRObj, order = FALSE,
                                     onlySignificantGenes = TRUE)

# Extract significant genes and transcript groups
sig_genes_stageR <- unique(stageR_results$geneID[stageR_results$gene < 0.05])
sig_tg_stageR <- stageR_results$txID[stageR_results$transcript < 0.05]

# Filter original DEXSeq table
dxr_stageR <- dxr_original_df[
  dxr_original_df$gene_id %in% sig_genes_stageR &
  dxr_original_df$transcript_group_ids %in% sig_tg_stageR, ]

Problem: Some transcript returned by stageR have padj > 0.05 in the original DEXSeq output table. Question: Should I re-filter again based on DEXSeq2 padj, or is stageR's adjustment sufficient? From what I understand, stageR should already handle multiple testing properly — is that correct? Thank you for your consideration. Best, Nhu

Since stageR is not part of the pipeline (and I'm not familiar with stageR), you might find better answers on some other forum that is more specialized in R and statistics, e.g. Bioconductor has one. Regarding different p-value results, I would guess that DEXseq and stageR likely apply different tests, maybe even ask different questions, so different p-values are likely normal. Regarding the question if stageR p-values are corrected for multiple testing, I would say that the function name

getAdjustedPValues

suggests that they have been adjusted (there might be a separate function to get the unadjusted/raw p-values). The definitive answer would come from looking at the source code.

I appreciate your supports @Thomas Danhorn.