2026-organism-tractability

This repository is an archived code snapshot accompanying the paper “A tractability atlas for experimental organism selection”.

• Paper (DOI): 10.57844/arcadia-h2nn-w619
• Paper (publication page): The Stacks publication page
• Raw dataset (Zenodo): Zenodo dataset DOI
• Interface (searchable table): organism-tractability-data.arcadiascience.com
• Visual map: organism-tractability.arcadiascience.com

We plan to refresh the data periodically; this code repo is not intended to be updated.

What this code does

Given a list of organisms, this code fetches tractability features across four domains:

• Community
• Logistics
• Throughput
• Tooling

Features are defined in:

• src/organism_tractability/db/feature_metadata/feature_metadata.yaml

Setup

This repo uses uv and requires Python >=3.12.

Install uv using the official instructions: https://docs.astral.sh/uv/getting-started/installation/.

On macOS with Homebrew, for example:

brew install uv
uv sync
source .venv/bin/activate

The commands above assume a Unix-like shell (bash/zsh). You can also run commands without activating the virtual environment by prefixing with uv run.

API keys

Populate required keys in .env (see .env.example).

Source ID	Required key(s)
ncbi	NCBI_API_KEY, NCBI_API_EMAIL
atcc	FIRECRAWL_API_KEY
exa_answer	EXA_API_KEY
protocols_io	PROTOCOLS_IO_API_CLIENT_ACCESS_TOKEN
nih_reporter	None

Input CSV contract

The features pipeline reads a CSV with these columns:

• organism_scientific_name: e.g. Escherichia coli
• organism_id: taxonomy id (integer). UniProt taxonomy id == NCBI taxonomy id (taxid).

Example input file:

• input/example_organisms.csv

Run: fetch all features for all organisms

python -m organism_tractability.db.cli get-features \
  --input input/example_organisms.csv \
  --output output/features.csv

You can optionally restrict sources:

python -m organism_tractability.db.cli get-features \
  --input input/example_organisms.csv \
  --output output/features.csv \
  -s ncbi -s protocols_io

When restricting sources with -s/--source-ids, only credentials for the selected sources are required.

The implementation lives in:

• src/organism_tractability/db/features/pipeline.py (FeaturesPipeline.run_csv)

Output CSV contract

The pipeline writes one row per (organism, feature) pair.

Output columns:

• organism_id: taxonomy id
• feature_id
• source_id
• fetched_object: JSON string (the raw returned object)

Example output row (illustrative):

organism_id,feature_id,source_id,fetched_object
562,pubmed,ncbi,"{""search_url"":""https://pubmed.ncbi.nlm.nih.gov/?term=%22Escherichia%22%20AND%20%22coli%22&sort=date&ac=yes"",""count"":123456}"

For a committed example output file, see:

• output/features.csv

The Zenodo dataset linked above is the canonical reference output for this archived snapshot.

Run sources directly (per organism)

Each source can also be queried directly:

python -m organism_tractability.sources.cli get-ncbi -n "Escherichia coli" -i 562
python -m organism_tractability.sources.cli get-atcc -n "Escherichia coli"
python -m organism_tractability.sources.cli get-nih-reporter -n "Escherichia coli"
python -m organism_tractability.sources.cli get-protocols-io -n "Escherichia coli"
python -m organism_tractability.sources.cli get-exa-answer -n "Escherichia coli"

Sources

Source	Notes	Required API key(s)
NCBI	Entrez E-utilities searches across multiple NCBI databases	NCBI_API_KEY, NCBI_API_EMAIL
ATCC	Scrapes ATCC search + product pages	FIRECRAWL_API_KEY
Exa Answer	Web search + LLM answer w/ citations + confidence. Output is non-deterministic across runs/time.	EXA_API_KEY
protocols.io	Searches public protocols	PROTOCOLS_IO_API_CLIENT_ACCESS_TOKEN
NIH RePORTER	Searches NIH-funded projects	None