PlantMetWiki Embedding Inputs: resolved sequences, SMILES, EC numbers and organism IDs for 424-species plant metabolism graph

Resolved raw inputs for all embedding-generation scripts used in PlantMetBench / PlantMetWiki. Downloading this archive lets you reproduce the node embeddings (embeddings_protein.pt, embeddings_gene.pt, embeddings_metabolite.pt) without re-running the API scraping pipelines (UniProt, NCBI nuccore/eutils, PubChem, MetaNetX, KEGG). The archive contains six TSV files (tab-separated, UTF-8): FileRowsDescriptionprotein_sequences.tsv8,445node_id → amino-acid sequence, resolved from UniProt REST API / RCSB PDBgene_sequences.tsv1,359node_id → DNA sequence (GenBank/EMBL), resolved via NCBI nuccore/eutilsmetabolite_smiles.tsv4,529node_id → canonical SMILES, resolved via PubChem, MetaNetX chem_prop.tsv, and KEGG mol filesmetabolite_inchikeys.tsv4,234 validnode_id → InChIKey, derived from SMILES via rdkit (295 SMILES contain R-group wildcards [*] for macromolecule attachment points that are valid for MAP4 fingerprinting but cannot produce a standard InChIKey)conversion_ec.tsv13,585node_id → EC number (e.g. 2.1.1.280), parsed from reactions.ttl via rdflib; covers 68.2% of the 19,927 Conversion nodesorganism_ncbi.tsv424node_id → NCBI taxon ID, extracted from NCBITaxon IRIs in nodes.tsv These inputs feed the following embedding scripts in the companion code repository: scripts/learnathon_embed_proteins.py / scripts/fulldata_embed_proteins.py (ESM-C 300M per-sequence embeddings), scripts/learnathon_embed_genes.py (PlantCaduceus l32 DNA embeddings), scripts/learnathon_embed_metabolites.py (MAP4 fingerprints), scripts/compute_ec_embeddings.py (237-dim EC hierarchy one-hot features), and scripts/compute_conversion_embeddings.py (DRFP-approximation reaction fingerprints). The 295 metabolites without InChIKeys are macromolecule fragments that carry R-group wildcards ([*]) in their SMILES — they are valid MAP4 substrates and are included in metabolite_smiles.tsv with an empty inchikey field in metabolite_inchikeys.tsv. Companion datasets: full graph with embeddings (10.5281/zenodo.20847651), 2-species Learnathon subset (10.5281/zenodo.20736061).