[NoopDog]

Part of #399 (its assay terms only; Bionano, seqspec, chains and GRCm39 stay open there)

What changed

• Three assay terms in assay_type_enum (src/meta_disco/schema/classification.yaml, generated model regenerated):
  • Hi-C: EFO:0007693. No parent among our values; its EFO parents, the chromosome conformation capture terms, aren't values here.
  • Strand-seq: EFO:0008952. No parent among our values.
  • Iso-Seq, titled "Isoform sequencing": OBI:0002575, under RNA-seq. EFO has no long-read RNA term, so this uses OBI's; the term's description records why (see Assumptions).
• Translation rows (rules/value_map.yaml):
  • assay_type.hi_c now declares assay_type: Hi-C beside its existing modality.
  • assay_type.isoseq is authored for the first time: Iso-Seq, transcriptomic.
  • New assay_type.strandseq: Strand-seq, genomic.
• Slot map: ANVIL_HPRC's sample.strandseq column now gives assay_type from its name, as sample.hic already does.
  • ANVIL_HPRC's evidence is re-imported as generation 20261027T015110Z. The only difference from 20261012T192136Z is 1,728 new strandseq assay lines.
• Consistency rule assay_for_genomic admits Strand-seq.
• Lookup tables: the manifest survey's name tokens give hic and strandseq their terms, and HPRC validation maps isoseq to Iso-Seq.
• CLAUDE.md: the assay_type_enum note.
• Tests: pin each new term's id and parent, and check the new assays against the consistency rules. Tests that used Hi-C as their example of a value with no term now use CenSat.
• Reports: regenerated from a full run at this branch's head.

Why

AnVIL's HPRC submitter tables already state these assays: HPRC_R2's hic and kinnex tables, and the hic and strandseq columns of ANVIL_HPRC's sample table. Their files stayed unclassified for assay type only because our vocabulary had no term to translate them into. #399 recorded these as vocabulary decisions owed; this change makes the three assay decisions.

Assumptions I made

• Iso-Seq's id comes from OBI, not EFO. I compared how others name this data:

  • EFO has no long-read RNA term. A request filed 2026-10-06 ([NTR] Request for new long-read RNA, single-cell long-read and DNA methylation assay terms EBISPOT/efo#2842) asks for one, citing OBI:0002575, and is still open. EFO's nearest term is FL-cDNA (EFO:0003755), any sequencing of full-length cDNA. It is broader (SRA uses it for Sanger-era cDNA libraries too) and sits under EFO's RNA assay rather than RNA-seq. The no-context review found it; Dave kept OBI's term (2026-10-27), and the term's description records why FL-cDNA was not taken.
  • OBI's isoform sequencing is defined as sequencing a cDNA library enriched for full-length reads. That fits PacBio's Iso-Seq and Kinnex and names the method rather than a vendor's kit. Direct RNA sequencing has its own OBI term, so it stays separate.
  • ENCODE names the same data long read RNA-seq, under its own placeholder id NTR:0000445.

  Dave chose OBI's id over a term of our own with no id. The value is Iso-Seq and the title "Isoform sequencing", the same pattern as WGS / "Whole genome sequencing". If EFO imports the term, its meaning can move to EFO's id.

• Hi-C and Strand-seq sit at the top level, as Methylation array and Genotyping array do: none of their EFO parents is one of our values.

• Strand-seq's modality is genomic: it measures the DNA sequence and its structure, from template strands phased into haplotypes and inversions. MODAL's genomic has no child after Drop data_modality terms that name a method, not a biological signal (genomic.assembly and its siblings) #756. The reason argues from what is measured, not from the molecule sequenced, which the table already refuses as modality evidence for SRA's GENOMIC.

• Iso-Seq's row also declares transcriptomic. This agrees with inference's existing answer on all 230 files.

• Translation rows only. No inference rule claims these assays. The consistency linter reads inference output, so a Hi-C ↔ modality rule would never fire; that belongs with Run the consistency linter over the reconciled run, not only inference output #597.

• docs/manifest-survey.md is not regenerated. It still says Hi-C has no term, but regenerating it would also pull in unrelated changes from a newer manifest download (a duos_dataset_registration table, a file_path column).

• The evidence re-import writes into the shared data/ folder. Until this merges, a run from main in another clone would list strandseq in the review queue, because main has no row for it. The PRIMED change (Read a PLINK .psam's sample IDs as its fileset's samples: PRIMED's 11,414 genotype files #751) did the same.

How to verify

1. Checks: run make test-all lint-all format-check validate-metadata. Everything should pass.
2. The terms:
   • In src/meta_disco/schema/classification.yaml, find Hi-C (EFO:0007693), Strand-seq (EFO:0008952) and Iso-Seq (OBI:0002575, is_a: RNA-seq). Iso-Seq's description should say why the id is OBI's.
   • tests/test_assay_type_vocabulary.py pins all three.
3. The rows: make review-queue. In docs/review-queue-report.md, isoseq should no longer be queued, and the rows assay_type.hi_c, assay_type.strandseq and assay_type.isoseq should show what they declare.
4. The evidence: ls data/source_evidence/anvil/anvil15/ANVIL_HPRC/ should list 20261027T015110Z. Compared with 20261012T192136Z, its sample.ndjson should add exactly 1,728 lines, all assay_type strandseq, and the other three files should be identical.
5. Output (run 20261026_194027, classified at 6edf390, reconciled at e6bf23f; since then only reasons, descriptions and tests changed, so the inference output stands): run make classify, then compare its reconciled/reconcile_report.json values with main's run 20261026_120218 (main bdf43ce). You should see:
   • ANVIL_HPRC: assay_type Hi-C +768 and Strand-seq +1,728; data_modality genomic +1,728.
   • AnVIL_HPRC_R2: assay_type Hi-C +3,002 and Iso-Seq +230.
   • No other value moves, the conflict rate is unchanged, and the inference files at the run root are unchanged. The committed reports come from that run.

#399's definition of done, for the three tokens this covers:

• Each token is in exactly one recorded state, mapped to a term this work adds: steps 2 and 3. hic and Hi-C → Hi-C, isoseq → Iso-Seq, strandseq → Strand-seq.
• Each term this work adds is a schema change reviewed on its own: step 2. The terms are in the schema diff, apart from the rows.
• The file counts are re-measured against the current catalog: step 5. Hi-C 3,770 as Vocabulary decisions: the closed set of declared terms our schema has no word for #399 measured, Iso-Seq 230, and Strand-seq 1,728 (not 1,344: the column also reaches the HPRC_PLUS Strand-seq files).
• A declined token makes no claim and stays in the review queue: not exercised. This change declines no token.

Review findings declined (Dave, 2026-10-27), recorded in a PR comment:

• The schema change is not a separate commit. Dave reviewed the three terms before any rule was written, and they are in their own two files in the diff.
• The HPRC evidence fixture is pinned to one measured generation, so it gains no strandseq line.
• Tests do not restate what bundled rows declare, or the slot map's spans, per the project's rule on tests.
• docs/manifest-survey.md is not regenerated (see Assumptions).

🤖 Generated with Claude Code

https://claude.ai/code/session_01VF7tbFa5ZTmYWZ6JWKiBBS
🐦 Built with cc-claude-tools

[Copilot]

🔵 Needs a closer look

Final human review should confirm the ontology choices and externally stored evidence generation’s provenance.

No open issues 🎉

What changed in this PR

Adds Hi-C, Strand-seq, and Iso-Seq to Meta-disco’s assay vocabulary, addressing the assay-term portion of #399 through submitter-evidence translations rather than new inference rules.

Changes:

• Defines the three assay terms and ontology identifiers, with Iso-Seq under RNA-seq.
• Updates evidence routing, translations, consistency checks, and lookup tables.
• Adjusts vocabulary tests and refreshes reports showing 5,728 newly classified assay values.

File	Description
tests/​test_value_map.py	Updates vocabulary-gap fixtures and remaining queue expectations.
tests/​test_source_evidence.py	Uses CenSat as the raw vocabulary-gap example.
tests/​test_rule_engine.py	Replaces the Hi-C state-claim example with CenSat.
tests/​test_hprc_validation.py	Expects canonical Iso-Seq spelling.
tests/​test_consistency.py	Checks Strand-seq and Iso-Seq modality compatibility.
tests/​test_claim_record.py	Removes Hi-C from a vocabulary-gap comment.
tests/​test_assay_type_vocabulary.py	Pins new terms’ ontology identifiers and parents.
src/​meta_disco/​validation_maps.py	Maps HPRC Iso-Seq spellings to the canonical term.
src/​meta_disco/​sources/​anvil_slot_map.yaml	Routes the Strand-seq column to assay evidence.
src/​meta_disco/​schema/​classification.yaml	Defines the new assay terms and their rationale.
src/​meta_disco/​schema/​classification_model.py	Updates the generated assay enum and documentation.
src/​meta_disco/​rules/​value_map.yaml	Declares assay and modality translations.
src/​meta_disco/​rules/​consistency_rules.yaml	Allows Strand-seq for genomic data.
src/​meta_disco/​manifest_survey.py	Recognizes Hi-C and Strand-seq name tokens.
schema/​tests/​test_output_validation.py	Updates the unrestricted raw-evidence example.
docs/​unprocessable-report.md	Refreshes the run reference and failure examples.
docs/​rules-report.md	Reports new declarations and application counts.
docs/​review-queue.html	Refreshes the queue and authored translation display.
docs/​review-queue-report.md	Removes translated Iso-Seq values from the queue.
docs/​reconcile-report.md	Records classification gains and evidence counts.
docs/​consistency-report.md	Refreshes the inference consistency run reference.
CLAUDE.md	Documents the expanded assay vocabulary.

_{Review effort: Balanced}

[Copilot]

🟢 Approval recommended

Static review supports the consistency of these additive schema, mapping and report changes; no blocking issues remain, though project checks were not executed.

No open issues 🎉

_{Review effort: Balanced}

[NoopDog]

Findings from the no-context review with no review thread, and what became of each:

Fixed (e6bf23f):

• Iso-Seq's description, the enum's and CLAUDE.md named no EFO term. They now name FL-cDNA (EFO:0003755) as EFO's nearest and say why it was not taken. Dave kept OBI:0002575.
• Hi-C's and Strand-seq's descriptions now list their direct EFO parents.
• The hi_c, isoseq and strandseq reasons were made accurate.
• validation_maps' module docstring was fixed.
• assay_for_genomic now notes that Strand-seq can't reach it yet.
• Two synthetic test rows no longer declare Hi-C as genomic.

Declined (Dave, 2026-10-27):

• The schema terms aren't a separate commit. They were reviewed with Dave before any rule was written, and they sit in their own two files.
• The HPRC fixture gets no strandseq line. The fixture is pinned to the measured 20260920 generation, which has 41 keys and 43 strings.
• Mutations to row declarations and to the slot-map span survive. Catching them would mean restating bundled rows in tests, which the project's rule on tests forbids.
• docs/manifest-survey.md is left stale. Regenerating it would pull in unrelated manifest drift.

🐦 Built with cc-claude-tools

[Copilot]

🟢 Approval recommended

Static review supports the consistency of the schema, mappings, tests, and reported deltas, with no blocking issues identified; automated checks were not run.

No open issues 🎉

_{Review effort: Balanced}