Medical Research Wiki Schema

Domain

Personal medical research knowledgebase focused primarily on ulcerative colitis, with room for related symptoms, comorbidities, interventions, mechanisms, medications, supplements, diet/lifestyle approaches, clinicians, and patient anecdotes.

This wiki is for organizing research and lived-experience reports. It is not medical advice and should support better conversations with clinicians.

Core Principles

Separate clinical evidence from patient anecdotes and personal notes.
Preserve raw sources immutably under raw/; synthesize in wiki pages.
Track provenance for important claims.
Mark uncertainty, contradictions, and safety concerns explicitly.
Prefer practical summaries that are readable by a human in Obsidian.
Every substantial update should add or update links, index entries, and log entries.

Directory Layout

raw/                         # Immutable sources
  papers/                    # PubMed papers, PDFs, preprints
  articles/                  # Web articles, explainers, blog posts
  clinical-guidelines/       # Guidelines from medical societies/hospitals
  notion-exports/            # Imported Notion notes/pages
  reddit/                    # Reddit threads/comments
  youtube/                   # YouTube transcripts/comments
  social/                    # X/Twitter, forums, other social sources
  personal-notes/            # User-provided notes, symptom logs, observations
conditions/                  # Top-level condition objects and condition folders
  ulcerative-colitis/        # Example condition object folder
    index.md                 # Main condition hub
    research.md              # Clinical and mechanistic research map
    methods.md               # Protocols, diets, supplements, lifestyle methods
    coaches-and-influencers.md # Clinicians, coaches, creators, communities
    personal-history.md      # Paul-specific timeline and observations
symptoms/                    # Symptom pages
interventions/               # Diet/lifestyle/protocol/intervention pages
mechanisms/                  # Mechanisms, hypotheses, biomarkers
medications-and-supplements/ # Drugs, supplements, dosing discussions
clinicians-and-sources/      # Doctors, researchers, trusted sites, channels
anecdotes/                   # Synthesized patient anecdote pages
comparisons/                 # Side-by-side analyses
queries/                     # Filed answers to substantial questions
journal/                     # Chronological research journal and weekly digests
_meta/                       # Maps, dashboards, maintenance reports
scripts/                     # Local helper scripts, including backups

Condition-First Notebook Model

Paul’s main medical issues should be treated as top-level objects in the notebook. Each major issue should have a condition folder with a main index.md hub plus supporting subpages.

Initial top-level condition objects:

conditions/ulcerative-colitis/
conditions/ulcerative-proctitis/ if it becomes useful to separate from broader UC.
conditions/silent-reflux/
conditions/high-cholesterol/
conditions/sleep-apnea/
Additional major issues from Guava/Notion as discovered.

Each condition object should answer:

What is the current understanding of the condition?
What is Paul’s personal history with it?
What objective data exists: labs, scopes, imaging, symptom logs, meds tried?
What healing/cure/remission hypotheses are being tracked?
Which coaches, clinicians, influencers, or protocols are relevant?
Which methods/interventions are promising, neutral, or risky?
What new findings from the journal should be promoted into the organized knowledgebase?

Use the old flat directories (interventions/, mechanisms/, clinicians-and-sources/, anecdotes/) for cross-condition canonical pages, but link them from the condition object pages. Do not let them replace the condition hub structure.

Journal Model

Use journal/ for chronological notes from recurring research, new internet scrubs, and interesting discoveries. The journal is the inbox and narrative layer; the condition pages are the organized knowledgebase.

Recommended paths:

journal/index.md
journal/weekly-digests/YYYY-MM-DD-weekly-medical-research-digest.md
journal/finds/YYYY-MM-DD-short-topic.md

Journal entries should include:

Date and source type.
Why it is interesting or hopeful.
Relevant condition object(s).
Evidence/anecdote/source class.
Whether it should be promoted into a condition page, method page, coach/source page, or mechanism page.
Links to raw sources.

Recurring cron jobs should append or create journal entries first, then promote high-value findings into organized pages when warranted.

Source Classes

Use these source_class values in frontmatter:

Clinical / scientific evidence:

clinical-guideline
systematic-review
meta-analysis
randomized-trial
phase-1-trial
observational-study
prospective-cohort
case-control-study
case-series
case-report
mechanistic-study
animal-study
mechanistic-review
clinical-review
review
hypothesis-paper
clinical-guidance
clinical-trial-registry

Clinician / patient / social / provenance:

clinician-article
clinician-video
clinician-email
clinician-voicemail
patient-blog
patient-anecdote
patient-resource
reddit-thread
youtube-transcript
youtube-comment
x-twitter-post
forum-thread
personal-note
notion-import
scite-search

Keep this list in sync with tools/validate_structure.py. Run python3 tools/validate_structure.py after adding pages; add a value here (and in the validator) before using it, to avoid taxonomy sprawl.

Evidence Levels

Use evidence_level on synthesized pages:

high — multiple guidelines/systematic reviews or strong replicated clinical evidence
moderate — plausible clinical evidence but limitations remain
low — early, indirect, small, uncontrolled, or mostly mechanistic evidence
anecdotal — primarily patient reports/social data
speculative — hypothesis or weak inference
mixed — conflicting evidence or highly context-dependent
unknown — not yet assessed

Confidence Levels

Use confidence separately from evidence level:

high — well supported and consistently represented
medium — useful but incomplete or context-dependent
low — uncertain, sparse, or potentially biased

Safety Fields

When relevant, include:

safety_flags: [requires-clinician, medication-interaction, immunosuppression, pregnancy, liver-risk, kidney-risk, infection-risk, flare-risk, supplement-quality, unknown]
red_flags: []
clinician_questions: []

Frontmatter for Synthesized Pages

---
title: Page Title
created: YYYY-MM-DD
updated: YYYY-MM-DD
type: condition | symptom | intervention | mechanism | medication | supplement | clinician-source | anecdote-synthesis | comparison | query | dashboard | care-plan | journal-find | research-digest | weekly-digest
tags: [ulcerative-colitis]
source_class: []
evidence_level: unknown
confidence: medium
sources: []
contested: false
contradictions: []
safety_flags: []
---

Frontmatter for Raw Sources

---
title: Source Title
source_url: https://example.com
source_class: patient-anecdote
ingested: YYYY-MM-DD
sha256: BODY_SHA256
notes: []
---

Raw sources should not be edited after ingest except to add missing frontmatter. Corrections and synthesis belong in regular wiki pages.

Tag Taxonomy

Core condition tags:

ulcerative-colitis
ibd
flare
remission
gut-health
inflammation
immune-system
microbiome

Evidence/source tags:

clinical-evidence
guideline
paper
anecdotal-reports
social-media
notion-import
personal-note

Intervention tags:

medication
supplement
diet
lifestyle
sleep
stress
exercise
protocol

Meta tags:

open-question
needs-review
safety
contradiction
clinician-question

Add new tags here before using them to avoid tag sprawl.

Page Thresholds

Create a synthesized page when:

A condition, symptom, intervention, or mechanism is central to a source.
An intervention appears in 2+ sources or is important to the user’s situation.
A recurring anecdotal pattern appears across several reports.
A query would be painful to re-derive later.

Do not create pages for passing mentions unless they are clinically important or safety-relevant.

Anecdote Handling Rules

Anecdotes are useful but biased. For anecdote pages, track:

Where the anecdote came from.
Diagnosis certainty, if available.
What intervention was tried.
Dose/duration/timing, if available.
Reported benefit or harm.
Confounders: concurrent meds, diet changes, disease severity, placebo/nocebo, selection bias.
Whether reports cluster around a plausible pattern.

Never present anecdotes as proof. Label them as evidence_level: anecdotal unless corroborated elsewhere.

Medical Research Answer Format

When answering from the wiki, prefer:

Short answer
What the evidence says
What patient anecdotes suggest
Safety / reasons to be cautious
Open questions
Questions to ask a clinician
Sources consulted

Update Policy

When new information conflicts with existing pages:

Preserve both claims with dates and sources.
Prefer clinical guidelines over anecdotes for safety-critical decisions.
Mark contested: true and add contradictions if needed.
Add a note to log.md.

Link Conventions

Internal links use Obsidian [[wikilinks]]. To stay unambiguous under Obsidian’s default “shortest path when possible” resolver:

Never use ../ relative prefixes. Obsidian wikilinks do not truly resolve them; they only appear to work when a basename happens to be unique.
Unique basename → bare basename. e.g. [[central-theory]], [[distal-proctitis-onset|Distal / Proctitis-First Onset]].
Colliding basename → full vault path. There are several index.md files, so always write [[conditions/ulcerative-colitis/index|Ulcerative Colitis]], never [[index]] or [[../conditions/ulcerative-colitis/index]].
Give hub pages aliases: so human-friendly links like [[Ulcerative Colitis]] resolve to the hub, not to an unrelated same-named file.
Run python3 tools/fix_links.py (dry run) then --apply to normalize links, and python3 tools/check_links.py to catch dead/ambiguous/orphan links.

Maintenance

Read SCHEMA.md, index.md, and recent log.md before making updates.
Run python3 tools/check_links.py and python3 tools/validate_structure.py before and after structural edits; both exit non-zero on hard errors.
Update index.md for every created synthesized page.
Append every ingestion/query/update to log.md.
Back up the vault before bulk imports or structural edits.
Flag stale, low-confidence, and safety-relevant pages for review.