circrna live audit (2026-08-23)
Original workflow (not an upstream port) · Repo: oxo-flow-circrna
(Traitome org, entry file circrna.oxoflow) · Rating:
live-verified (data/pipelines.json).
Four independent circRNA callers (CIRIquant, CIRCexplorer2, find_circ,
circRNA_finder) + fastp QC + ensemble aggregation (≥2 callers agree).
Reference indexes (bowtie2/STAR/bwa/hisat2) and conda envs build
automatically on first run; samples auto-discovered from raw/.
Live re-verification (2026-08-23)
Engine: latest main (post-v0.14.1) · Box: tx-ubuntu · Mode: real CLI
run, not dry-run · -j 2 --keep-going · exit 0, one round, zero
fixes.
- 4 reference indexes auto-built (1.2kb synthetic genome, seconds)
- 9 rules all passed: fastp → 4 callers → multiqc → aggregate → aggregate_dataset → report; 12 output files verified
- Real signal: find_circ detected the planted BSJ
chr1:520-719 + 20 reads circ_000001— exactly matching the fixture's designed e3→e2 back-splice position - CIRIquant ran for real: 19 candidate splice-signal reads → second-scan PEM 0 → 0 circRNA (below the statistical threshold of the tiny fixture)
- circRNA_finder produced real STAR Chimeric.out.junction / SJ.out.tab outputs; no candidates passed its filters
Honest coverage note
All 4 callers executed for real. 1/4 detected the planted BSJ; 3/4
reported 0 circRNAs because the mini fixture (200 read pairs, 1kb
genome) sits below their detection thresholds — a statistical limit
of the fixture, not a tool or port defect. The report rule's guard
echo (WARNING not created) is rule-internal logic; exit 0 is
correct.
Usability note
Entry file is circrna.oxoflow, not the main.oxoflow convention —
auto-discovery still finds it (alphabetically first *.oxoflow), but
renaming to main.oxoflow would align with the other community
repos. Pending owner decision.