Source metadata and access conditions were checked on 2026-09-30. Exact machine
inputs are declared in config/sources.toml; this document explains why each source
is used and what the project does to it.
- Upstream biological files are downloaded into
data/raw/, which is ignored by Git. - The downloader writes a local
download-manifest.jsonwith the retrieval time, source URL, byte size, and SHA-256 digest of each artifact. - Normalized full-data outputs under
data/processed/are also ignored. - Fixture relationships, coordinates, and scores are synthetic and source-shaped. The small public JASPAR bHLH34 profile is included with attribution under CC BY 4.0.
- Publications and source accessions must be cited in downstream analyses. The repository MIT license does not relicense source data.
- Owner/distributor: TAIR/Phoenix Bioinformatics through its official Zenodo community
- Input:
Araport11_GFF3_genes_transposons.20241001.gff.gz - Release: pinned quarterly snapshot dated 2024-10-01
- Source URL: https://zenodo.org/records/17371665/files/Araport11_GFF3_genes_transposons.20241001.gff.gz?download=1
- Record DOI: https://doi.org/10.5281/zenodo.17371665
- Published MD5:
c72c998d26ef28180409370d74229a92 - Citation: Cheng C-Y et al. (2017), The Plant Journal. https://doi.org/10.1111/tpj.13415
- Reuse status: the Zenodo record is public but does not state a reuse license. Public access is not treated as permission to relicense the file. It is fetched for local processing and is not redistributed.
The adapter loads exact GFF3 gene features, decodes attributes, retains genomic
coordinates and strand, and keys records by AGI locus. Transcript and child features
are excluded. Araport coordinates are one-based and closed.
The pinned snapshot contains occasional legacy bytes in annotation text. Undecodable
UTF-8 bytes are retained as explicit \xNN escapes in normalized text fields.
- Owner/distributor: O'Malley et al.; NCBI Gene Expression Omnibus
- Input:
GSE60141_RAW.tar, containing processed narrowPeak files - Accession: GSE60141, the DAP-seq subseries within superseries GSE60143
- Source URL: https://ftp.ncbi.nlm.nih.gov/geo/series/GSE60nnn/GSE60141/suppl/GSE60141_RAW.tar
- Citation: O'Malley RC et al. (2016), Cell. https://doi.org/10.1016/j.cell.2016.04.038
- Reuse status: publicly archived by NCBI GEO. NCBI/GEO policies and any submitter rights apply. The archive is downloaded for local processing and is not redistributed.
The full pipeline selects native-genome DAP-seq peak files and excludes _colamp-
ampDAP-seq files. NarrowPeak intervals are zero-based and half-open; the adapter
converts the start to a one-based interval before overlap comparison. A peak supports
a PUTATIVE_DAP_TARGET edge when it overlaps the target gene's strand-aware window
from 2,000 bp upstream through 500 bp downstream of the transcription start site.
Qualifying peaks are aggregated by TF and target gene.
This transformation is a project inference. DAP-seq is an in vitro binding assay; neither a peak nor promoter proximity proves transcriptional regulation in vivo.
- Owner/distributor: JASPAR consortium
- Input: detailed REST API records for CORE plant matrices filtered to NCBI taxonomy ID 3702
- Release: JASPAR 2026
- Source URL: https://jaspar.elixir.no/api/v1/matrix/?collection=CORE&tax_group=plants&tax_id=3702&version=latest&page_size=1000
- Release site: https://jaspar.elixir.no/
- Citation: Ovek Baydar D et al. (2026), Nucleic Acids Research. https://doi.org/10.1093/nar/gkaf1209
- License: Creative Commons Attribution 4.0 International
- Redistribution policy: downloaded API output is not committed; attribution and license metadata are retained in the project documentation.
The downloader follows API pagination and fetches matrix detail records using at most
four simultaneous requests. tax_id=3702 selects Arabidopsis and version=latest
selects the non-redundant profiles. The live API serves the current JASPAR release;
the recorded SHA-256 and retrieval date identify the exact local snapshot. The adapter
keeps versioned matrix IDs, retains Arabidopsis records, derives a deterministic
consensus from each PFM, and emits a gene-to-motif link only when the JASPAR name
exactly matches an Araport symbol after case folding. It does not use fuzzy matching,
so valid aliases may remain unlinked.
- Owner/distributor: Dario Monachello and Claire Lurin; Recherche Data Gouv
- Dataset: InterATOME protein-protein interactions, Public, version 1.1
- Dataset DOI: https://doi.org/10.15454/5I2RO1
- Input file:
Public_raw_datathrough data-file API identifier100278 - Source URL: https://entrepot.recherche.data.gouv.fr/api/access/datafile/100278
- Published MD5:
8748bc11b954531f1516fccd4835fde8 - Related publication: Arabidopsis Interactome Mapping Consortium (2011), Science. https://doi.org/10.1126/science.1203877
- License: Etalab Open Licence 2.0, identified by the repository as compatible with CC BY 2.0
- Redistribution policy: fetched on demand rather than vendored; dataset authors, DOI, version, and license are attributed.
The public InterATOME collection includes systematic large-scale Arabidopsis Y2H
maps. This project keeps rows marked AI1, strips transcript suffixes, requires both
endpoints to resolve to Araport loci, canonicalizes endpoint order, and collapses
reciprocal duplicates. Reported homodimer evidence is retained as a self-loop. The
graph stores publication PMID:21798944 and assay Y2H on each relationship.
config/metal_focus.tsv is a small, hand-curated set of example query seeds. Each
row includes an evidence URL and the explicit note that it is not a comprehensive
annotation set. It is not presented as a fifth public dataset and should not be used
for enrichment statistics or completeness claims.
Run plant-kg download, then archive the generated download manifest alongside the
analysis outputs. Compare its SHA-256 values before reusing an existing raw-data
directory. A changed digest can reflect a changed upstream response even when the URL
is unchanged and should trigger review before graph regeneration.