Interview-Driven Workflow Construction and Repair in Galaxy

Candida auris SCF1 Adhesin RNA-seq Workflow Case Study

This directory contains the artifact bundle generated by the PAPER → GALAXY pipeline (pipeline-paper-to-galaxy) from the Galaxy Workflow Foundry, run against Santana et al. 2023, Science (PMID 37769084).

Status: partial. Phases 1–10 completed; the workflow is concrete and passes every static gate. Phase 11 (run-workflow-test) failed twice and phase 12 was not reached, so this bundle has no passing end-to-end execution. See CASE_STUDY.md §7 for both failures and the identified fix. Contrast with ../ebola-rnaseq, which reached 100% pass.

See CASE_STUDY.md for the phase audit, methodology and results narrative.

Quick Links

Generated Galaxy Workflow: galaxy-workflow.gxwf.yml — 27 steps, 9 inputs, 16 outputs
Annotated draft (reasoning intact): galaxy-workflow-draft.gxwf.yml — the same workflow plus 850 comment lines of provenance that draft-extract strips
Workflow Tests: galaxy-workflow.gxwf-tests.yml — 2 cases, 172 assertions
Test Plan: galaxy-test-plan.yml
Test Data: test-data/ — 200k-pair fixtures for case 1 (60 MB) and test-data/smoke-25k/ for case 2 (6.7 MB). All 24 verified against the SHA-1s recorded in the test file.
Test Data Refs: test-data-refs.json — pinned ENA/NCBI URLs, md5s, subsetting recipe
Validation Report: galaxy-workflow-validation-result.json — 20 validated / 7 skipped / 0 fail
Execution logs (both failures): planemo-smoke.attempt1.log, planemo-smoke.log, planemo-smoke.json, planemo-smoke.html
Obligations ledger: open-requirements.ledger.yml — 28 entries, 15 open / 13 resolved
Foundry feedback ledger: foundry-feedback.ledger.yml — 61 entries of defects/gaps found in the Foundry by using it

Re-running Verification

1. Static validation (all currently green)

gxwf validate galaxy-workflow.gxwf.yml \
  --cache-dir <a writable tool cache>        # WITHOUT --cache-dir this reports 0 validated / 27 skipped at rc 0
gxwf validate-tests galaxy-workflow.gxwf-tests.yml --workflow galaxy-workflow.gxwf.yml
foundry validate-galaxy-workflow-test-plan galaxy-test-plan.yml

2. Planemo execution — NOT yet passing

The recommended next attempt uses biocontainers, as ../ebola-rnaseq did, which avoids the conda environment build that failed here:

planemo test --biocontainers \
  --galaxy_branch release_26.1 \
  --galaxy_python_version 3.12 \
  --test_index 2 \
  --test_output tool_test_output.html \
  --test_output_json tool_test_output.json \
  galaxy-workflow.gxwf.yml

--galaxy_branch release_26.1 is mandatory: sample_sheet:paired support lands in Galaxy 25.1, and branches ≤ 25.0 cannot run this workflow at all. --test_index 2 selects the 25k smoke case, which exercises the full topology in a fraction of the time.

3. Regenerating the fixtures

Both sets are included, but they are also fully regenerable and pinned — stream each ENA FASTQ from the URL in test-data-refs.json and take head -n 800000 (200k pairs, test-data/) or the first 100,000 lines of that (25k pairs, test-data/smoke-25k/). Verify against the recorded md5_uncompressed — hash the uncompressed file, since gzip output is not byte-stable across versions. The SHA-1s in galaxy-workflow.gxwf-tests.yml check the compressed files as shipped.