Skip to content

Fix BeatAML drug response sample mapping - #516

Merged
sgosline merged 1 commit into
mainfrom
fix-beataml-drug-mapping
Oct 20, 2026
Merged

sgosline merged 1 commit into
mainfrom
fix-beataml-drug-mapping

Conversation

@jjacobson95

@jjacobson95 jjacobson95 commented Oct 19, 2026 •

Copy link
Copy Markdown
Collaborator

Fixes how BeatAML dose-response rows are matched to specimens. generate_raw_drug_file merged the raw inhibitor table onto the sample map on dbGaP sample ids and assigned the merged labId back by position. Rows without an RNA-seq id matched on the missing key, which misaligned that assignment, so some drug response values were matched to the wrong specimen.

Changes:

  • Map RNA-seq and DNA-seq ids to labId and cohort with key lookups that skip missing keys.
  • Resolve raw inhibitor names that carry an alias in parentheses (e.g. "Afatinib (BIBW-2992)") to the existing drugs-file entry.
  • Use one improve_sample_id per labId in the experiments step, so duplicated sample rows no longer duplicate dose data.

Before: 309 specimens, 139 drugs, 32,211 sample/drug pairs, Spearman 0.47 against the published BeatAML AUCs. After: 631 specimens, 162 drugs, 62,707 pairs, Spearman 0.99.

Closes #514

Map dbGaP sample ids to labId and cohort with key lookups instead of
merges, resolve inhibitor names with parenthetical aliases, and use one
improve_sample_id per labId in the experiments step.
@jjacobson95 jjacobson95 added this to the 2.4 new build milestone Oct 19, 2026
@jjacobson95 jjacobson95 self-assigned this Oct 19, 2026
@jjacobson95
jjacobson95 requested a review from sgosline October 19, 2026 20:33
@sgosline
sgosline merged commit fc0b779 into main Oct 20, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

BeatAML had a long-term drug mapping issue

2 participants