kfukuta/wf-transcriptomes-v202
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Set conv.em to 1e-4 as documented

Browse Source
This commit is contained in:
Sam Nicholls 2026-05-19 09:29:48 +01:00
parent 0bd6f118b5
commit 051e60792c
No known key found for this signature in database
1 changed files with 4 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
bin/workflow_glue_r/R/bambu.R
View File

@ -277,7 +277,10 @@ bambu_build_args <- function(args, reads, annotation_obj, discovery, quant) {
if (quant) { if (quant) {
# Chunked quant re-estimates degradation bias per chunk, which changes # Chunked quant re-estimates degradation bias per chunk, which changes
# the EM inputs and breaks equivalence with unchunked bambu quant. # the EM inputs and breaks equivalence with unchunked bambu quant.
bambu_args$opt.em <- list(degradationBias = FALSE) # Use bambu's documented EM convergence threshold here. The looser
# implemented default (1e-2) leaves small deterministic chunk-vs-global
# quant discrepancies in the split workflow.
bambu_args$opt.em <- list(degradationBias = FALSE, conv = 1e-4)
} }
bambu_args bambu_args