Supporting data for: End-to-end mapping of membrane transport from chemical structure to microorganisms

This dataset accompanies ChemProFlow, a tool that maps membrane transport from chemical structure to microorganisms. Given a compound (SMILES), ChemProFlow predicts whether it is likely to be transported, assigns transport mechanisms using the Transporter Classification Database (TCDB), and retrieves the microorganisms whose genomes encode the matching transport systems. It provides: Datasets and models Training datasets: substrate–transporter associations built from TCDB, Rhea, UniProt and ChEBI, and unlabeled natural-product-like compounds from PubChem. Trained models: graph neural networks (D-MPNN) that predict transport ability and transport mechanism (TC-IDs), with cross-validation folds, checkpoints, decision thresholds and training logs. Case study data on pyoverdine siderophores and their outer-membrane receptor. A catalog of microorganisms from BioCyc, linked to TCDB transporters by reciprocal best hits. Source data for figures Figure2.xlsx Figure3.xlsx