Goal

Create a 2024–2025 PETase variants & effects evidence table using PubMed and Scholar Gateway connectors, with strict, citable provenance.

When to use

• The user asks for PETase or PET hydrolase variant scouting, benchmarking, or zero-shot hypothesis generation focused on 2024–2025.

Operating rules

1. Always use connectors (PubMed first; Scholar Gateway for Wiley/paywalled content). Prefer PMC for full text when available.
2. Restrict the search window to 2024-01-01 .. 2025-10-21.
3. For each paper, extract mutations, assay comparability, performance endpoints, thermostability, expression notes, rationale, and citations.
4. Deduplicate by DOI; include PMID/PMCID when present.
5. Use the Output schema exactly (see resources/output-schema.csv).

Query seeds

(See resources/query-strings.md for adaptable queries.)

Output format

• Primary: table matching the schema (CSV in a fenced code block).
• Secondary: ranked shortlist (5–10 variants) + zero-shot hypotheses.

Steps

1. Search PubMed with Boolean + fielded queries; filter 2024–2025; sort by date.
2. For promising records, fetch abstracts/full text (PMC if available). Cross-check in Scholar Gateway to access Wiley content.
3. Extract data into the schema; convert/standardize units carefully.
4. Deduplicate by DOI; verify all rows have at least one citation (PMID/PMCID/DOI).
5. Produce the CSV table, then a ranked shortlist and 2–3 hypotheses.

Safety & provenance

• Cite rigorously (PMID/PMCID/DOI). No hallucinated values.
• If a metric is missing, leave it blank.
• Do not train or store proprietary datasets; this is a literature-only workflow suitable for Zero‑Shot reasoning.