AI-EV Bone Marrow Targeting: Integrative Multi-Dimensional Evidence Dataset for Ligand-Receptor Prioritization

This dataset contains the input data and computational outputs from an integrative multi-dimensional evidence scoring framework for prioritizing candidate ligand-receptor pairs involved in extracellular vesicle (EV) bone marrow targeting. The pipeline integrates evidence from 10 public databases (UniProt, PDB, AlphaFold DB, STRING, PubMed, InterPro, Human Protein Atlas, GTEx, PaxDb, DICE) to score 66 E. coli K12 bacterial outer membrane and adhesin proteins against 45 human receptors across 10 evidence dimensions: surface accessibility, protein abundance, receptor expression, membrane localization, functional relevance, known PPI, literature/patent support, PLM features, structural evidence, and computational feasibility. Results include 3,366 ranked ligand-receptor pairs (1,914 neutrophil + 1,452 HSPC), with top audited candidates: fimH (Type 1 pilus adhesin), ompA (Outer membrane protein A), and lpp (Braun lipoprotein). Input data were derived from public databases and curated literature. All source database access links and processing methods are documented in the accompanying DATA_SOURCES.md file.