Add E2E test scenario: Kusako yeast heat shock RNA-seq reanalysis

Realistic scenario where an undergrad uses Claude Code to reanalyze
published RNA-seq data (PMID 32109230, GSE135568) for TOR pathway
genes not discussed in the original paper. Exercises the full chain:
PMC text mining → TogoID ID conversion → fetchngs → rnaseq.

All data paths verified against live APIs.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: inutano

docs/e2e-tests/001-kusako-yeast-rnaseq.md

@@ -0,0 +1,130 @@
+# E2E Test 001: Kusako — Yeast Heat Shock RNA-seq Reanalysis
+
+> **Test type:** End-to-end agent scenario
+> **Workflows:** fetchngs → rnaseq
+> **Organism:** *Saccharomyces cerevisiae* (R64-1-1)
+> **Data:** GSE135568 (PMID: 32109230)
+
+## Scenario
+
+Kusako is an undergraduate biology student working on a thesis about nutrient signaling in *Saccharomyces cerevisiae*. Her lab studies the TOR pathway, and she's interested in how TOR-regulated genes (e.g., *GAP1*, *MEP2*, *DAL5* — nitrogen permease genes) respond under heat shock. She found a paper about yeast heat shock transcriptomics, but the paper focuses on the RNA binding protein Mip6 and histone modifications — her target genes aren't discussed at all.
+
+She wants to reanalyze the same RNA-seq data to look at expression of her specific gene set. She has Claude Code but no bioinformatics experience.
+
+## Prompt
+
+```
+I'm studying TOR pathway genes in yeast for my thesis. I found this paper about
+heat shock response (PMID: 32109230) that has RNA-seq data — they looked at Mip6
+and histone modifications but I want to check what happens to nitrogen permease
+genes like GAP1, MEP2, and DAL5 under the same conditions. Can you get the data
+and run the analysis for me?
+```
+
+## Expected Agent Behavior
+
+### Step 1: Understand the request
+
+This is a secondary analysis: reuse published RNA-seq data to answer a different biological question. The student doesn't know about workflows, WES, or CWL — the agent handles everything.
+
+### Step 2: Read AGENTS.md
+
+Discover available workflows, the WES submission protocol, and provenance requirements.
+
+### Step 3: Find the data
+
+Read `docs/data-discovery.md`, then:
+
+1. Try TogoID: PMID 32109230 → SRA run accessions (no direct route exists)
+2. **Fall back to PMC full text**: fetch the paper via Europe PMC or NCBI E-utilities
+   - Paper: "A multi-omics dataset of heat-shock response in the yeast RNA binding protein Mip6" (PMC7046740)
+   - Scan Data Availability / Data Records section for archive identifiers
+   - Find: **GSE135568** (GEO series accession)
+3. Use TogoID to convert identifiers:
+   - `GET https://api.togoid.dbcls.jp/convert?ids=GSE135568&route=geo_series,bioproject&format=json`
+   - Result: **PRJNA559331**
+   - `GET https://api.togoid.dbcls.jp/convert?ids=PRJNA559331&route=bioproject,sra_run&format=json`
+   - Result: 67 SRA run accessions (SRR9929263–SRR9929329)
+4. Filter for RNA-seq only (the dataset also contains ChIP-seq):
+   - Query ENA: `study_accession="PRJNA559331"` with `library_strategy` field
+   - 24 RNA-seq runs, 36 ChIP-seq runs — select only RNA-seq
+5. Present the accessions to Kusako with metadata:
+   - Organism: *S. cerevisiae* BY4741
+   - Conditions: wild-type vs mip6Δ, 30°C vs 39°C (20 min and 120 min heat shock)
+   - Library: paired-end, Illumina HiSeq 2000
+   - Ask which samples she needs (all 24, or a subset for her comparison)
+
+### Step 4: Fetch raw data
+
+Read `workflows/fetchngs/agent.yaml`:
+- Prepare input YAML with confirmed SRA accessions
+- Submit fetchngs to WES, poll until complete
+- Retrieve FASTQ output paths
+
+**For testing:** Use a subset of 2–4 samples to keep runtime short (e.g., 1 wild-type 30°C + 1 wild-type 39°C).
+
+### Step 5: Resolve references
+
+Read `references/README.md` + `references/genomes.yaml`:
+- Look up *S. cerevisiae* R64-1-1 (Tier 1 in catalog)
+- rnaseq needs: genome FASTA + GTF + STAR index
+- Check iGenomes: `s3://ngi-igenomes/igenomes/Saccharomyces_cerevisiae/Ensembl/R64-1-1/Sequence/STARIndex/`
+- Download pre-built STAR index, OR run `prepare-references` with `build_star: true`
+
+### Step 6: Run RNA-seq quantification
+
+Read `workflows/rnaseq/agent.yaml`:
+- Prepare inputs: FASTQs from step 4, genome FASTA + GTF + STAR index from step 5
+- Select aligner: STAR (default for model organisms with good annotation)
+- Submit to WES, poll until complete
+- Retrieve gene-level counts and MultiQC report
+
+### Step 7: Retrieve provenance
+
+`GET /runs/{run_id}/ro-crate` for both fetchngs and rnaseq runs.
+Save RO-Crate metadata.
+
+### Step 8: Present results in student-friendly terms
+
+- Extract expression levels of GAP1, MEP2, DAL5 from the gene count matrix
+- Compare across conditions (30°C vs 39°C heat shock)
+- Explain what the counts mean (higher = more transcripts = more active)
+- Point to MultiQC report for data quality
+- Note that this is raw counts, not differential expression — suggest DESeq2 as next step if she sees interesting patterns
+
+## Verified Data Path
+
+This test scenario has been verified end-to-end:
+
+| Step | Input | Method | Output | Verified |
+|------|-------|--------|--------|----------|
+| PMC lookup | PMID 32109230 | Europe PMC / NCBI efetch | PMC7046740 (open access) | Yes |
+| Text mining | PMC7046740 full text | Scan Data Records section | GSE135568 | Yes |
+| ID conversion | GSE135568 | TogoID geo_series→bioproject | PRJNA559331 | Yes |
+| ID conversion | PRJNA559331 | TogoID bioproject→sra_run | 67 SRR accessions | Yes |
+| RNA-seq filter | 67 runs | ENA API library_strategy filter | 24 RNA-seq runs | Yes |
+| Genome lookup | *S. cerevisiae* | references/genomes.yaml | R64-1-1, Ensembl 115 | Yes |
+| STAR index | R64-1-1 | iGenomes S3 | Available | Yes |
+
+## Success Criteria
+
+- [ ] Agent fetches PMC full text and extracts GSE135568 from the paper
+- [ ] Agent converts GSE135568 → PRJNA559331 → SRR accessions via TogoID
+- [ ] Agent filters to RNA-seq runs only (24 of 67)
+- [ ] Agent presents sample metadata and asks Kusako to confirm
+- [ ] fetchngs completes and produces FASTQ files
+- [ ] Yeast reference resolved from genomes.yaml without manual intervention
+- [ ] rnaseq completes with STAR alignment and gene quantification
+- [ ] Agent reports expression values for GAP1, MEP2, DAL5
+- [ ] RO-Crate retrieved and valid for both workflow runs (Dataset + ComputationalWorkflow + CreateAction)
+
+## Test Subset (for CI/quick validation)
+
+For automated testing, use 2 samples to keep runtime under 30 minutes:
+
+| Run | Condition | Genotype |
+|-----|-----------|----------|
+| SRR9929263 | 30°C (control) | BY4741 wild-type |
+| SRR9929265 | 39°C 20 min (heat shock) | BY4741 wild-type |
+
+These two samples provide the minimal comparison: control vs heat shock in wild-type, which is what Kusako needs to see if her target genes respond to heat stress.