回到作品集

提示词库

从确定问题到核验终稿,共八个步骤。以一个连续的研究主题为例,提供可以复用的提示词。

下载全部提示词 TXT ↓英文提示词 · 8 个步骤

从你自己的问题开始。示例主题是 CRISPR-Cas 复合物的分子动力学。请替换主题、时间范围和方括号内的输入,保留证据核查要求;使用任何结论之前,回到原始来源核验。

工具名称沿用原始工作流程。你可以使用偏好工具中的同类功能,具体菜单与可用功能可能变化。

01

明确范围

原流程建议工具 · frontier LLM

I'm planning a review article and want to sharpen its scope before I start searching.

Working topic: molecular dynamics (MD) simulations of CRISPR-Cas effector
complexes, focused on what they reveal about conformational mechanisms —
target recognition, R-loop formation, conformational gating, and cleavage
activation.

Help me by:
1. Proposing 3 candidate research questions at different scope levels
   (broad / focused / narrow), each phrased as a single answerable question.
2. For the focused option, drafting explicit inclusion and exclusion criteria:
   date range, simulation types to include (classical all-atom, enhanced
   sampling, QM/MM, coarse-grained), which Cas families are in vs out of scope,
   and what counts as in-scope vs merely adjacent.
3. Listing 4–6 sub-themes a thorough review of this question must cover.

Ask me clarifying questions first if anything about the scope is ambiguous.
02

建立全局认识

原流程建议工具 · Perplexity · Academic Focus + Deep Research

Turn on Academic Focus and run it in Deep Research mode. Treat the output as a map, not a draft — click and verify the sources it cites.

Act as a research analyst. Give me a structured, cited overview of the use of
molecular dynamics simulations to study CRISPR-Cas effector complexes from
2018 to 2026.

Cover:
- Which Cas systems MD has been applied to (Cas9, Cas12 family, Cas13, etc.)
  and what mechanistic questions each line of work addressed
- The dominant simulation approaches — classical all-atom MD, enhanced sampling
  (metadynamics, umbrella sampling, replica exchange), QM/MM, coarse-grained —
  and the force fields used for protein–nucleic acid complexes
- Landmark papers and their most-cited findings
- Open questions and known methodological limitations

For every claim, cite a peer-reviewed source with a working link. Explicitly
flag where the evidence is thin, preliminary, or contested.
03

检索与筛选

原流程建议工具 · Elicit

Search question — paste into Elicit's main query box

What do molecular dynamics simulations reveal about the conformational
mechanisms of CRISPR-Cas effector complexes during target recognition,
R-loop formation, and cleavage?

Screening criteria — apply while reviewing the result list

INCLUDE: peer-reviewed primary studies, 2018–2026, that run MD (any flavor)
on a CRISPR-Cas effector–nucleic acid complex and report a conformational or
mechanistic result.

EXCLUDE: pure experimental papers with no simulation; docking-only or static
structure-prediction papers with no dynamics; reviews; Cas systems outside
[your chosen families]; simulations under [e.g., 50 ns] with no enhanced
sampling.
04

提取证据

原流程建议工具 · Elicit columns · SciSpace · Scite

Elicit custom columns — add these so the corpus drops into a comparable grid

- Cas effector studied
- Simulation method + force field
- Timescale (ns/µs) and system size (atoms)
- Enhanced sampling used? (which)
- Key conformational finding
- Validated against experiment? (how)
- Stated limitations

Per-paper deep read — paste into SciSpace / any chat-with-PDF tool with the paper open

Summarize this paper's MD methodology in 5 bullets: system built, force field,
solvation and ions, ensemble + timescale, and enhanced-sampling method (if any).
Then state the single main conformational conclusion, and quote the specific
result — with its numbers — that supports it.

Do not infer beyond the text. If a detail isn't stated, write "not reported."

Reception check — paste into Scite for any finding you plan to lean on

Has the finding that [e.g., "the REC2 domain gates HNH activation in Cas9"]
been supported, mentioned, or contradicted by later studies?
05

综合与提纲

原流程建议工具 · frontier LLM

Feed it your finished extraction table. This is where you decide the argument — the model only organizes.

Below is my extraction table from [N] papers on MD simulations of CRISPR-Cas
complexes.

[PASTE TABLE]

Using ONLY the information in this table:
1. Group the papers into 4–6 coherent themes and name each theme.
2. For each theme, state the consensus finding and note any disagreements
   between papers.
3. Identify 2–3 genuine gaps the table reveals — questions that none of these
   papers answer.
4. Propose a section-by-section outline for a review built around these themes.
   Under each section, list which papers (first author + year) support it.

Do not add any paper, fact, or citation that is not in the table. If something
is missing, say so rather than filling it in.
06

逐节起草

原流程建议工具 · frontier LLM · sources-only

Draft section by section. Paste in only the verified sources for that section.

Draft the "[SECTION NAME — e.g., Enhanced-sampling approaches to R-loop
formation]" section of my review. Target ~[600] words.

Rules:
- Use ONLY the sources pasted below. Introduce no other papers, facts, or
  citations.
- Cite as [Author, Year], matching the provided sources exactly.
- Preserve every number, unit, and statistical value exactly as written —
  never round, rephrase, or "improve" them.
- Where sources disagree, present the disagreement; do not smooth it over.
- Academic register. No filler, no "in conclusion."

Sources:
[PASTE each verified paper or your notes, tagged with Author + Year]
07

建立核验清单

原流程建议工具 · frontier LLM

The model can't verify citations — it just extracts them so you can check each one by hand.

Here is a section of my review draft. Extract every factual or numerical claim
and the [Author, Year] it is attributed to. Output a table with columns:
Claim | Cited source | Verified? (leave this column blank for me).

Do NOT judge whether any citation is correct — you cannot verify sources. Only
pull out the claims and their attributed sources so I can check each manually.

[PASTE draft section]

Then open every DOI yourself and fill the "Verified?" column.

08

学术语言润色

原流程建议工具 · frontier LLM · or Writefull / Paperpal

Line-edit the following for academic journal register. Tighten wordiness, fix
grammar, and sharpen vague phrasing (e.g., replace bare "significant" with the
specific statistical claim). Do NOT change any technical claim, number,
citation, or the argument.

Show the edited text, then a short list of the substantive changes you made.

[PASTE text]
← 回到往昔