/propose

/propose — Research Foundation Document

Synthesizes accumulated literature notes into a structured **Research Foundation Document (RFD)**.

The RFD serves as the standardized upstream input for any downstream protocol design skill (e.g., clinical-cohort-protocol-designer, translational-study-blueprint, statistical-analysis-plan-writer). It can also be used independently as a research proposal outline.

> ⚠️ **Literature Integrity (LITERATURE_HARD_RULES.md applies throughout)** > Every citation in the RFD must correspond to a paper in `reading_list.md` marked as read `[x]`. > Never fabricate PMIDs, DOIs, author names, citation counts, or study findings. > If a claim requires literature support that the user has not yet read, insert a gap marker: > `[GAP: supporting literature needed — suggested search: <keywords>]`

---

Steps

1. Collect Source Material

**[Tier C / B]** Read `config.json` to get `data_dir`. Then read:

• `memory/MEMORY.md` — research direction, framework, Reading Progress
• `data_dir/notes/reading_list.md` — complete reading record
• All read paper notes (`.md` files marked `[x]`): extract **Section III (Transferable Elements)**, **Section IV (How to Use)**, and **Discussion Log** entries
• Most recent recap snapshot in `data_dir/notes/recaps/` (if available)

**[Tier A]** Ask the user to paste their latest Session Card. Optionally ask them to paste key note content from important papers.

---

2. Show Summary and Ask Clarifying Questions

Display a materials overview:

> **You currently have:** > - N papers analyzed: [list with one-line core contribution each] > - Theoretical framework: [extracted from MEMORY.md — or "not yet defined" if absent] > - Key discussion insights: [top 3 most actionable entries from all Discussion Logs] > > **I will now guide you through building your Research Foundation Document.** > We will go section by section. After each section, you can revise before we proceed. > Type "skip" to skip a section, or "back" to revise a previous one.

Then ask the following questions. **Only ask questions whose answers cannot be found in MEMORY.md or the notes.** Skip any that are already answered.

**Q1 — What is the core question your research aims to answer?** Use the format: [Population] + [Exposure / Predictor / Intervention] + [Outcome]. Examples:

• "In glioma patients, does low SNAI3-AS1 expression predict sensitivity to ferroptosis inducers?"
• "In AD patients with high lncRNA ceRNA network activity, is cognitive decline faster?"

**Q2 — What study design do you lean toward?** (Can be uncertain)

• Mechanistic exploration (in vitro / in vivo experiments)
• Bioinformatics analysis (transcriptomics / multi-omics)
• Retrospective cohort (existing data)
• Prospective cohort (requires enrollment)
• Machine learning prediction model
• Systematic review / meta-analysis

**Q3 — What data or samples do you have or can realistically access?** (Can be uncertain)

• Public databases (GEO, TCGA, CGGA, UK Biobank, etc.)
• Institutional EHR / medical records
• Collected tissue or blood samples
• No data yet — still in planning
• Other

---

3. Build the RFD Section by Section

After each section, pause and ask: "Does this look accurate? Any changes before we continue?" Accept "continue" or direct corrections before moving on.

---

**Section 1 — Study Population & Clinical Context**

Describe:

• Target disease and clinical setting
• Study population characteristics (informed by Q2/Q3 answers)
• How prior literature defines this population

For each claim, cite the source paper from the reading list. If no source is available, label it `[GAP: literature needed]`.

---

**Section 2 — Focused Research Question**

Structured format (PECOT):

| Element | Definition | Source | |---------|-----------|--------| | **P** Population | [specific description] | [Author_Year or GAP] | | **E/X** Exposure / Predictor | [specific description] | [Author_Year or GAP] | | **C** Comparator / Baseline | [if applicable] | [Author_Year or GAP] | | **O** Outcome | [specific description] | [Author_Year or GAP] | | **D** Design | [from Q2 answer] | — | | **T** Time horizon | [if applicable] | [Author_Year or GAP] |

Every cell in the Source column must either name a paper from the reading list or state `[Assumption — literature support needed]`.

---

**Section 3 — Theoretical Framework & Mechanistic Basis**

• ASCII diagram of the framework (key nodes + directional arrows)
• Each node labeled with the supporting paper(s)
• Nodes without literature support labeled `[GAP: node not yet supported]`

Example format:

lncRNA SNAI3-AS1
    │ (suppresses)           [Zheng_2023_SNAI3-AS1]
    ▼
SND1 / m6A pathway
    │ (destabilizes)         [Zheng_2023_SNAI3-AS1]
    ▼
Nrf2 mRNA stability
    │ (reduces)              [GAP: mechanism link needs literature]
    ▼
Ferroptosis sensitivity      [Hu_2020_TMZ_Resistance]

---

**Section 4 — Methodological Precedents**

Drawn from Section III (Transferable Elements) of the paper notes:

| Method | Source Paper | Transferable Operation | |--------|-------------|----------------------| | LASSO feature selection | Yang_Pan_2021 | Applied to ferroptosis-related gene set; identical approach applicable here | | WGCNA co-expression | Tan_2023 | Hub gene identification in AD; adapt for glioma lncRNA network |

Only methods from papers in the reading list may appear here.

---

**Section 5 — Identified Research Gaps**

Drawn from Section II (Paper Highlights) — specifically the critique of existing limitations — and the framework completeness check from the last `/recap`:

• Gap 1: [description] → How your study addresses this
• Gap 2: [description] → How your study addresses this
• ...

---

**Section 6 — Literature Source Index**

| Paper | Supports Section | Citation Role | |-------|-----------------|---------------| | Author_Year_Title | Section 2 / 3 / 4 | Population definition / Mechanistic support / Method precedent |

⚠️ This table mu