Arabidopsis Metal Homeostasis Knowledge Graph

Which transcription factors bind near iron-homeostasis genes, and how do their motifs and protein interactions connect? This project brings those records together in Neo4j, using public Arabidopsis thaliana data.

Data

• Araport11: gene coordinates and symbols, pinned to the 2024-10-01 snapshot.
• DAP-seq, GSE60141: transcription-factor binding peaks.
• JASPAR CORE 2026: Arabidopsis transcription-factor motifs.
• InterATOME 1.1: the AI-1 yeast two-hybrid subset from the ORFeome-scale interactome screen.

The verified full build contains 33,318 genes, 598 motifs, 6,438 Y2H pairs and 1,172,835 putative DAP-seq target relationships. All four validation suites passed. Downloads stay out of Git; source versions, checksums, citations and reuse terms are in data provenance and the input snapshot.

Run locally

You'll need Python 3.11+ and Docker Compose. From the repository root:

python -m venv .venv
source .venv/bin/activate
python -m pip install -e ".[dev]"
cp .env.example .env
docker compose up -d --wait
plant-kg pipeline --profile fixture

In PowerShell, use .venv\Scripts\Activate.ps1 and Copy-Item .env.example .env for activation and copying. The fixture runs offline and is for testing only.

Open Neo4j Browser with user neo4j and password plantgraph-local. Keep these defaults local. docker compose down stops the service without deleting its data.

For the public datasets, switch to a separate volume so synthetic fixture records don't mix with real evidence. Allow extra disk space and network access:

docker compose down
docker compose --project-name plant-kg-full up -d --wait
plant-kg pipeline --profile full

Stop that instance with docker compose --project-name plant-kg-full down.

Explore and check

Start with the metal regulatory neighborhood, motif-supported regulators or convergent evidence queries.

plant-kg validate checks identifiers, coordinates, provenance and relationship integrity. Unique constraints protect node keys; loading the same tables twice doesn't duplicate the graph. GitHub Actions runs pytest against an isolated Neo4j service. For local tests, see Contributing.

Reading the evidence

DAP-seq measures binding in vitro. A peak within 2,000 bp upstream to 500 bp downstream of a strand-aware transcription start site creates a putative target, not a validated regulatory relationship. Motif links require exact Araport symbol matches, so some aliases are missed.

Y2H edges describe binary assay observations at the gene-locus level, without isoform, tissue or condition specificity. The four iron-homeostasis seed genes are starting points for queries, not a complete metal-homeostasis catalog.

Architecture · Data dictionary · Citation · MIT license. The license covers code and documentation; upstream data keep their own terms.

Package installation and release checks are in the release guide.