feat: add experimental plant hosts

[kkyungseo]

변경 파일	내용
data/arabidopsis_thaliana_codons.json	신규 — 64 코돈, TAIR10 기반
data/solanum_lycopersicum_codons.json	신규 — 64 코돈, ITAG4.0 기반
data/lemna_minor_codons.json	신규 — 64 코돈
cli/main.py	HOST_MAP + click.Choice 에 4종 추가
current_parameter_registry.yaml	4개 host_profiles 등록 (status: experimental)
design_package.schema.json	hostProfile enum/oneOf 확장
tests/engines/profile/test_host_plants.py	신규 — 파라미터화 테스트 12개
CHANGELOG.md / docs/cli.md	문서 업데이트

JSON 파일 품질 검증(64 코돈, 아미노산별 frequency 합계 ≈ 1.0, preferred 코돈 = 최고빈도 코돈)은 로컬에서 통과했으며, ruff lint/format도 clean 상태입니다.

[vercel]

@KyungSeo116 is attempting to deploy a commit to the munkyukim86's projects Team on Vercel.

A member of the Team first needs to authorize it.

[munkyukim86]

Code review

Found 3 issues:

1. New hosts added to CLI but not to the REST API — api/optimize.py still has VALID_HOSTS = ["nbenthamiana", "by2"] and its own HOST_MAP that were not updated. Any POST /optimize request using arabidopsis, tomato, lemna, or wolffia will be rejected with HTTP 400. The BY-2 addition (commit 593a04e) set the precedent of updating both files together.

factorforge-cds/src/factorforge/cli/main.py

Lines 19 to 27 in e8d1514

	HOST_MAP = {
	"nbenthamiana": "nbenthamiana",
	"by2": "ntabacum",
	"arabidopsis": "arabidopsis_thaliana",
	"tomato": "solanum_lycopersicum",
	"lemna": "lemna_minor",
	"wolffia": "wolffia_globosa",
	}

2. load_golden_set() is hardcoded to nbenthamiana_golden_set.json — The function in utils.py has no host parameter and always loads the N. benthamiana golden set. ReverseTranslator uses this to compute golden_ref_weights (the CAI reference weights), so any optimization run against arabidopsis, tomato, lemna, or wolffia will silently score and select codons using N. benthamiana CAI weights, not the target host's. The reported CAI values for these hosts will be misleading.

factorforge-cds/src/factorforge/engines/profile/rules/reverse_translator.py

Lines 74 to 84 in e8d1514

	# Load golden set for CAI reference weights
	if golden_set_path is not None:
	self.golden_set_table: dict[str, Any] = self._load_codon_table(golden_set_path)
	else:
	try:
	self.golden_set_table = load_golden_set()
	except (FileNotFoundError, json.JSONDecodeError):
	self.golden_set_table = self.codon_table
	# Pre-compute relative adaptiveness weights from golden set (Sharp & Li 1987)
	self.golden_ref_weights: dict[str, float] = self._build_ref_weights(self.golden_set_table)

3. Wolffia codon table provenance mismatch — The registry registers wolffia with ncbi_taxonomy_id: 113308 (Wolffia globosa), but the source field states "proxy from NCBI GCF_029677425.1 (W. australiana)" — a different species with a different taxonomy ID. The codon table data and the declared organism do not match. Either the taxonomy ID should reflect W. australiana, or a note should make clear this is a cross-species proxy.

factorforge-cds/src/factorforge/registry/current_parameter_registry.yaml

Lines 186 to 200 in e8d1514

	owner_job: "024"
	source: "NCBI Taxonomy Browser NCBITaxon:3702; TAIR10 CDS annotation (NCBI GCF_000001735.4)"
	rationale: "Model dicot plant A. thaliana; experimental host profile based on Kazusa/TAIR10 codon usage."
	provenance: "Issue #24 plant host addition; codon table from Kazusa CodonUsage Database + TAIR10."
	visibility: public
	permission: publish_allowed
	tomato:
	display_name: "S. lycopersicum"
	scientific_name: "Solanum lycopersicum"
	ncbi_taxonomy_id: 4081
	ncbi_taxonomy_curie: "NCBITaxon:4081"
	status: experimental
	claim_level: experimental_setting
	evidence_status: experimental
	release_status: experimental

🤖 Generated with Claude Code

_{- If this code review was useful, please react with 👍. Otherwise, react with 👎.}

[munkyukim86]

Thanks for contributing this! Really appreciate you taking the time to add these plant hosts. Left a few notes above, once those are addressed, happy to get this merged.

[kkyungseo]

Thanks for the review! I'll work through the notes above and aim to have everything addressed by June 18 (KST).
I'll ping you here once it's ready for another look.

[munkyukim86]

Can't wait, thanks!