MetaGNN data v1: processed inputs and per-experiment artefacts

This deposit contains the processed input tensors, model checkpoints, and multi-seed prediction trees that accompany the MetaGNN framework. The companion code repository is at https://github.com/thiptanawat/MetaGNN-Framework. Contents (one tarball per subset): recon3d — Recon3D bipartite graph artefacts (SBML, stoichiometric matrix, metabolite features) tcga_crc_220 — 220-patient TCGA-CRC processed inputs tcga_crc_624 — 624-patient TCGA-CRC processed inputs tcga_brca — TCGA-BRCA processed graph + checkpoint (cross-cancer) tcga_luad — TCGA-LUAD processed graph + checkpoint (cross-cancer) hma_labels — HMA 11-tissue-union activity pseudolabels cptac_brca — CPTAC-BRCA same-platform discriminator inputs metabric — METABRIC microarray cross-consortium discriminator inputs depmap — DepMap CRISPR-essentiality + tissue GEMs training_logs — multi-seed prediction trees and fold-level CSVs Each subset unpacks into shared_data/<subset>/ in the GitHub repo. Verify integrity against the SHA-256 manifest SHA256SUMS.txt. Raw TCGA sequencing data are subject to dbGaP restrictions and are not redistributed here; they are fetched on demand by the download scripts in the GitHub repository.