High-Impact Genome Editing Literature Tracker

A production-oriented literature triage pipeline for genome editing labs.

This project scans a curated high-impact journal pool plus a watchlist of key corresponding / senior authors, filters papers using a boolean gate, ranks the matched papers, and writes a reader-friendly review card into Notion.

For every selected paper, the tracker produces:

• a 5-line review in Korean by default
• Why It Matters for the lab
• one best fast-follower
• a concrete first experiment
• a fast-follower score / rank
• a one-line share blurb

---

What this tracker is optimized for

This is not a generic paper summarizer.

It is designed for labs that care about:

• genome editing broadly
• CRISPR / Cas systems
• base editing / prime editing
• TALEN / zinc-finger / programmable nuclease platforms
• editor engineering, off-target behavior, specificity, and structural biology
• delivery and compact editor strategy
• mitochondrial editing and mitochondrial biology as an important subdomain

The ranking is now genome-editing-first. mtDNA and mitochondria remain important, but they are treated as part of the broader landscape rather than the only priority axis.

---

Core design

1) Two collection streams

The tracker collects from both:

A. Journal watch

• Nature
• Nature Biotechnology
• Nature Methods
• Nature Genetics
• Nature Medicine
• Nature Neuroscience
• Nature Structural & Molecular Biology
• Nature Chemical Biology
• Nature Biomedical Engineering
• Nature Cell Biology
• Nature Metabolism
• Nature Communications
• Science
• Science Translational Medicine
• Cell
• Molecular Cell
• Neuron
• Cell Metabolism
• Cell Stem Cell
• Nucleic Acids Research

B. PI / watch-author stream Recent papers are also pulled for the following author watchlist:

• David R. Liu
• Jennifer A. Doudna
• Feng Zhang
• Jay Shendure
• Omar O. Abudayyeh
• Jonathan S. Gootenberg
• Patrick D. Hsu
• Samuel H. Sternberg
• Caixia Gao
• Wensheng Wei
• Sangsu Bae
• Hyongbum Henry Kim

This makes the tracker robust even when a directly relevant paper appears outside the journal preset.

---

How the pipeline works

1. Collect

• High-impact journal query via Crossref
• Watch-author query via Crossref query.author

2. Enrich

PubMed enrichment is optional but recommended. When available, the tracker adds:

• PMID
• abstract
• MeSH / keyword metadata
• author list
• corresponding-author candidates inferred from affiliation emails

3. Boolean gate

A paper passes if:

• relevant terms appear in the title or keyword list, or
• strong editor terms appear in the abstract, or
• the paper matches the watch-author logic

The gate is intentionally broad enough to catch:

• genome editing core papers
• editor engineering / specificity / off-target papers
• delivery and compact-editor papers
• mitochondrial editing papers
• selected mitochondrial biology papers

4. Rank

Only gated papers are ranked.

The ranking is now organized around four buckets:

• Editing core
• Editor engineering
• Delivery / translation
• Mitochondrial bonus

This avoids overfitting to mtDNA while still rewarding papers that matter for mitochondrial editing.

5. Generate a reading card

The LLM writes:

• 5-line review
• why it matters to the lab
• technical takeaway
• best fast-follower
• first experiment
• FF score / rank
• one-line share blurb

6. Save to Notion

Each reviewed paper is written as a Notion page under the configured database.

---

Recommended operating mode

Weekly mode (recommended)

For this source universe, weekly review is more stable than daily review.

Recommended settings:

• days_back = 7
• min_gated = 5
• max_days_back = 35
• expand_step_days = 7
• llm_limit = 10

This gives a dense enough set for a team digest or lab meeting discussion.

On-demand mode

You can still run the pipeline manually after a major paper drop, conference, or relevant preprint-to-paper transition.

---

Reader-facing language

Reader-facing long-form output is Korean by default (READER_LANGUAGE=ko).

This applies to:

• 5-line Review
• Why It Matters
• Best Fast-Follower
• Share Blurb
• page body section headings and explanations

Structured metadata stays mostly stable / English-friendly so Notion filtering remains reliable.

---

Quick start

1) Create the environment

sudo apt update
sudo apt install -y python3 python3-venv python3-pip

cd cns_mt_tracker_high_impact
python3 -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install -r requirements.txt
cp .env.example .env

2) Fill .env

Minimum required:

OPENAI_API_KEY=...
NOTION_API_KEY=...
NOTION_DATABASE_ID=...

Recommended defaults:

OPENAI_MODEL=gpt-5.4-mini
READER_LANGUAGE=ko
TIMEZONE=Asia/Seoul

DAYS_BACK=7
LLM_REVIEW_LIMIT=10
MIN_GATED_PAPERS=5
MAX_DAYS_BACK=35
EXPAND_STEP_DAYS=7

CROSSREF_ROWS_PER_JOURNAL=50
CROSSREF_ROWS_PER_AUTHOR=10
JOURNAL_PRESET=cns_high_impact
ENABLE_PUBMED_ENRICHMENT=true

CROSSREF_MAILTO=your_email@example.com
NCBI_EMAIL=your_email@example.com
NCBI_API_KEY=

3) Prepare Notion

Share the original Notion database with your integration, then run:

python scripts/bootstrap_notion.py --apply

4) Weekly dry-run

./scripts/run_weekly.sh --dry-run

5) Weekly real run (writes to Notion)

./scripts/run_weekly.sh

---

CLI examples

Weekly dry-run with explicit args

python scripts/run_daily.py \
  --journal-preset cns_high_impact \
  --days-back 7 \
  --min-gated 5 \
  --max-days-back 35 \
  --expand-step-days 7 \
  --llm-limit 10 \
  --dry-run

Skip PubMed enrichment

python scripts/run_daily.py --skip-pubmed --dry-run

Run only flagship journals

python scripts/run_daily.py --journal-preset cns_main --dry-run

---

Notion field guide

Reader-facing fields

• Paper: paper title
• Journal: journal name
• Published: publication date
• Paper Keywords: paper-reported or extracted keywords
• Gate Reason: why the paper passed the initial filter
• 5-line Review: fast skim summary
• Why It Matters: lab relevance
• Best Fast-Follower: best follow-up idea
• FF Rank: actionability bucket
• DOI/URL: canonical link
• Status: reading / discussion state

Operational fields

• Lane: Genome editing core / Editor engineering / Delivery & translation / Mitochondrial editing / Mitochondrial biology / General biology
• Prescore: internal ranking score among gated papers
• LLM Priority: LLM-assigned importance
• Matched Keywords: additional matched terms used in ranking
• Watch Authors: matched watchlist PI names
• Watch Basis: author-list / last-author / PubMed email heuristic basis
• Corresponding Authors: corresponding-author candidates inferred from PubMed affiliation emails
• Source: metadata source
• PMID: PubMed identifier
• Discuss This Week: auto-checked for stronger candidates

---

Customization

Add more gate keywords

Edit:

• app/keywords.py

Useful sections:

• GENOME_EDITING_GATE_TERMS
• ENGINEERING_GATE_TERMS
• DELIVERY_GATE_TERMS
• MITO_EDITING_GATE_TERMS
• MITO_BIO_GATE_TERMS

Add more watched authors

Edit:

• WATCHED_AUTHOR_ALIASES in app/keywords.py

Change ranking behavior

Edit:

• RANK_BUCKETS in app/keywords.py
• synergy logic in app/prescore.py

Change output language

Edit .env:

READER_LANGUAGE=ko

or

READER_LANGUAGE=en

---