PlantMetWiki Full Dataset: 421-species enzyme-reaction graph with node embeddings and link-prediction splits

The complete PlantMetWiki plant-metabolism knowledge graph (424 plant species), prepared for heterogeneous-graph link prediction. Task: predict (Protein, catalyzes, Interaction) edges ("this enzyme catalyses this reaction"), where the destination is always a Conversion-subtype node (the reaction itself). Three complementary train/val/test split strategies are included: a random 80/10/10 split, an organism-held-out split (tests cross-species generalisation), and a pathway-held-out split (tests generalisation to unseen biological contexts). Includes pre-computed node embeddings attached to the PyTorch Geometric HeteroData object as node features: MAP4 binary fingerprints for metabolites, ESM-C (esmc_300m) embeddings for proteins, and PlantCaduceus (l32) DNA embeddings for gene products. Two new precomputed reaction-chemistry files are now included as standalone .pt companions and auto-loaded by learnathon/dataset.py. See README.md inside the archive for the full file list, coverage numbers, and loading code, and LICENSE.md for provenance of each embedding's source sequences. v5 changes (2026-07-06 — two new precomputed files; nodes.tsv updated): (1) embeddings_conversion.pt (242 MB) added: precomputed MAP4 reaction fingerprints for all 19,927 Conversion nodes (90% coverage), 3072-dim format [mean(substrate_MAP4) ‖ mean(product_MAP4) ‖ diff] — an approximation of DRFP using MAP4 instead of Morgan fingerprints. Auto-loaded by learnathon/dataset.py and fulldata_baseline_train.py. Generated by scripts/compute_conversion_embeddings.py. (2) embeddings_ec.pt (93 MB) added: precomputed EC hierarchy one-hot features for all 84,470 Interaction nodes (68.2% of 19,927 Conversion nodes covered), 237-dim format [EC-L1 (8) ‖ EC-L2 (61) ‖ EC-L3 (168)]. Auto-appended by learnathon/dataset.py after conversion embeddings; used by fulldata_baseline_train.py when --ec-features is passed, skipping the 30–60 s rdflib parse. Generated by scripts/compute_ec_embeddings.py. (3) nodes.tsv updated: new ec_number column (e.g. 2.1.1.280) for all annotated Interaction nodes — EC readable directly from the table without rdflib. (4) Hard-negative sampling implemented: same-pathway negatives force the model to learn within-pathway biochemical specificity rather than exploiting the near-trivial pathway co-membership shortcut. (5) Baseline v11 result: test P-H@50=0.118 (20× random, +66% vs v4 baseline), using conversion embeddings. v4 changes (2026-07-03 — learnathon baseline updated to v6; data files unchanged): The graph (heterodata.pt) and splits (splits_taxa.pt) are identical to v3. Changes are in learnathon training code and documentation: (1) dataset.py adds disjoint_train_ratio=0.2 default — 80% of training positives enter the message-passing graph, 20% are supervision-only, preventing the 1-hop memorisation shortcut; (2) dataset.py removes (Interaction, catalyzed_by, Protein) from the MP graph by default (keep_catalyzed_by=False), closing the 2-hop P→C→P shortcut; (3) train.py default negative sampling changed from 5+5 (random+corrupt-protein) to 5+0 — corrupt-protein negatives hurt P-H@50 when used aggressively (see README §Baseline); (4) learnathon metrics clarified: P-H@50 is the primary leaderboard metric, CP-AUC/CP-AP are secondary diagnostics, standard AUC/AP (random negatives, always ≥0.95) are omitted from reporting; (5) is_about edges confirmed load-bearing — ablation (v7) that removed them caused test P-H@50 to drop below random (0.003) and AUC to invert to 0.188. Baseline results: val P-H@50=0.124 (21× random), test P-H@50=0.071 (12× random), val CP-AUC=0.817 (emergent, no CP training). v3 changes (graph redesign): (1) generic DataNode/External nodes eliminated; (2) Inhibition/Stimulation/ComplexBinding-subtype nodes dropped; (3) TranscriptionTranslation collapsed into a direct (GeneProduct, encodes, Protein) edge. v2 changes: target edge collapses through Catalysis to Conversion; embedding-coverage-aware taxa split. See README.md inside the archive for full changelog. This is the full-graph counterpart of the 2-species Learnathon subset (10.5281/zenodo.20736061) — same task and file format, complete species coverage instead of an A. thaliana / G. max slice.