EVisionary: A Harmonized Multi-Omics Extracellular Vesicle Metadata Resource

The EVisionary Framework provides a rigorously harmonized, provenance-aware Extracellular Vesicle (EV) metadata resource. This dataset was created to address pervasive metadata fragmentation, semantic inconsistencies, and structural conflations across legacy EV repositories (Vesiclepedia, ExoCarta, and EV-TRACK). Data Structure & ContentsThe unified master table contains 258,460 provenance-tagged EV metadata records, consolidated from 713,667 source rows through explicit source-to-canonical field mapping, rule-based normalization, and provenance-preserving deduplication (overall retention 36.2%). The dataset integrates four jointly queryable molecular cargo classes — protein (207,623 records), mRNA (26,701), miRNA (16,131), and lipid (2,896) — together with broad species coverage (401 distinct labels, predominantly Homo sapiens) and study-level reporting metadata. To ensure programmatic scalability and privacy-preserving local execution, the dataset is provided in a compressed columnar Apache Parquet format (unified_EVmetadata_keyB.parquet), optimized for DuckDB and pandas analytics. Key Features• Provenance preservation: Every record retains explicit source attribution; deduplication uses a seven-field key that removes within-repository duplicates while keeping cross-repository records distinct and traceable.• Semantic normalization: Case-insensitive, word-boundary-stabilized matching resolves lexical fragmentation (e.g., human vs. Homo sapiens; mrna vs. mRNA) that otherwise distorts exact-match queries.• Multi-cargo integration: Four molecular classes are jointly queryable, with complementary source coverage (protein and miRNA concentrated in Vesiclepedia lipid in ExoCarta), demonstrating that no single repository provides complete molecular coverage.• Auditable & reproducible: Each transformation is recorded, and an automated cross-script consistency check verifies identical counts across the entire pipeline. UsageThis dataset is the analytical foundation for the article "EVisionary: a provenance-aware framework for federated harmonization and querying of extracellular vesicle repositories." For the harmonization pipeline, validation and audit scripts, and the browser-based query interface, please refer to the linked GitHub repository (https://github.com/Sogandste/EVisionary). License: MIT