MASTER PROMPT — Web of Biocatalysis → Pillar 02 (Design & Engineering) → Overview
Role You are writing the Overview for Pillar 02 inside a framework called the “Web of Biocatalysis.” This Overview is a high-level map, not a handbook chapter.
Pillars (context)
Discovery & Sourcing (find parents) · 2) Design & Engineering (change the enzyme) · 3) Assays & Analytics · 4) Media & Microenvironment · 5) Immobilization & Materials · 6) Expression & Formats · 7) Cascades & Routes · 8) Reactors, Flow & In-Line.
Scope of Pillar 02 We change amino-acid sequence (hence folded structure/dynamics) to meet process goals: activity (kcat, KM, kcat/KM), selectivity (ee/er, regio/chemo), stability (ΔTm, T50, half-life), tolerance (solvent/pH/salt), and substrate scope. Core approaches: rational design, semi-rational focused libraries, directed evolution; computation acts as a prior; scalable screens enable search. Interfaces only (do not drift): 1,3,4,5,6,7,8.
What you will receive (in this message) A foundation JSON with: definition (seed scope text), core_review_bundles (context), evidence_list (S1..Sn items with title/year/doi/snippet), and evidence_block (S-mapping). Optional: metrics/methods/open-question candidates.
HARD RULES
Evidence
Ground every factual sentence in the supplied evidence; end with one or more tags [S#]. Chain if needed: [S3][S7].
Use only DOIs present; no external facts; no placeholders.
No decade generalities (“over decades”), no “first/earliest/began/invented by” unless literally in a snippet.
Definition & opening
Start with a true definition of this pillar in your own words: “Protein engineering here means intentional modification of amino-acid sequences to alter folded structure and function so catalysts meet process goals.” Ground it in one or more high-level evidence items (reviews/overviews), not Wikipedia. End with [S#].
Examples & metrics policy
Ban singletons: Do not name specific enzymes (e.g., phytase, nitrilase, GlcDH) or quote single, old case metrics in the Overview.
Prefer platform-level examples that generalize (e.g., CAST focused saturation; picodroplet million-member screens; growth-amplified clonal assays).
Include at most one concrete metric only if it summarizes a generalizable pattern across items (e.g., “million-member library”), not a one-off ΔTm. If unsure, omit metrics.
ML sentence policy (neutral)
If the evidence supports ML-guided design, state how it fits (as a prior, rounds ↓, scope ↑) with [S#].
If the evidence set does not substantiate ML, write one neutral clause: “ML-guided strategies are not substantiated by this evidence set.” (Optionally append one [S#] if any ML-adjacent item exists.)
Do not use emoji or warning symbols.
Out-of-scope guardrails
Do not present immobilization, solvent choice, reactor tricks, or expression formats as design tactics; refer to them only as interfaces if needed and only in one compact sentence, grounded by [S#] if invoked.
Tone & style
Compact, precise, non-hyped; connective prose (no bullets).
Avoid filler verbs like “tailors,” “makes waves,” “revolutionizes.”
No cookbook “use X when you want Y”; prefer synthesis about how approaches relate.
Form & length
Output exactly two paragraphs, each ≤140 words. No headings, no bullets, no extra commentary.
Every sentence ends with [S#] (except the neutral ML clause if it has truly zero supporting items).
ALGORITHM (follow step-by-step)
Parse JSON → gather definition, core_review_bundles, evidence_list, map S1..Sn. Identify high-level sources suitable to ground definitions (reviews/overviews).
Build a tactics shortlist only from present evidence:
focused saturation / CAST / active-site windows (look for “CAST”, “site-saturation”, “combinatorial”),
scalable screens (microdroplets, picoliters, “million-member,” “growth-amplified”),
directed evolution at industrially relevant criteria; promiscuity/metagenomic exploration (only if present). Exclude named enzyme anecdotes.
Paragraph 1 (“What & How”):
One sentence: definition + process goals. [S#]
One sentence: name the three approaches and their roles. [S#]
One sentence: 1–2 platform-level examples from the shortlist (e.g., CAST; million-member droplets; growth-amplified clonal assays). [S#][S#]
Paragraph 2 (“Synthesis & Trajectory”):
One sentence: explain the continuum (stabilize/bias scaffold → focused libraries → iterate with scalable screens). [S#]
One sentence: acknowledge enabling platforms (droplets/metagenomic picodroplets/growth-amplified) if present. [S#]
One sentence: ML policy as above (neutral clause if unsupported).
One closing sentence: articulate what this layered playbook unlocks (e.g., robust, selective, process-fit catalysis), grounded in general evidence. [S#]
QA before output:
No enzyme names; no single-case ΔTm; no immobilization as a design method.
Each sentence ends with [S#] (except neutral ML clause if truly zero support).
Each paragraph ≤140 words.
LEXICAL GUARDRAILS (avoid)
“tailors,” “makes waves,” “over decades,” “game-changing,” “revolutionary,” “in the modern era,” “in recent years” (unless in evidence).
OUTPUT
Two paragraphs only. No headings. No bullets. No commentary. Keep it cohesive and readable, yet fully evidence-tagged.