Lectin L3b: full-text literature pass (report and data)

[hyphaltip]

Literature pass (task L3b) for the lectin class spec (PR #78). Report and data only. The spec edits that came from it are already on lectin-class-spec-r5.

• Report: docs/reports/2026-10-10-lectin-l3b-literature.md (per paper: PMID, DOI, access route, full text read yes/no, key facts with quotes, effect on spec F14/F17/F20/D7/D9).
• Data: docs/reports/data/lectin_l3b/ (292 PubMed hit titles; SHA-256 of archived texts). Publisher texts are not committed (kept in _workdir/curation_texts_archive/lectin_l3b/).
• Full text read for 14 papers (Kerr 2016, Houser 2013, Richard 2018, Duca 2022, Martínez-Alarcón 2021, Fernandes 2017, Oliveira 2017, Pitangui 2021 and 2023, Tronchin 2002, and others). Abstract only for 7. Not read: Ecp6, MoCVNH, BbLec1, Emp46/47, Avr4, Kuboi, Houser 2015, Lehot, Dussouy.
• Sakai 2019 was NOT read in the PDF (publisher 403; no PMC copy). The text came from a summarising fetch of the journal page.
• PubMed: 292 unique hits from 3 queries; 32 read at abstract or full-text level; 260 by title only. Emmonsia/Emergomyces coverage is weak.
• Kerr 2016 genus names (Trichophyton, Microsporum, Arthroderma) are verified in the full text (Europe PMC XML); the claim is a BLAST presence statement, not a tested orthology.
• No fungal lectin gene found for Coccidioides, Histoplasma or Blastomyces. The L3b stop rule is met (paracoccin; Histoplasma activity with no sequence). Default D9 (no custom model, report the gap) is consistent with the data; no performance data supports it.

Do not merge: needs owner review.

🤖 Generated with Claude Code