SMILE: Systematic Microbiome Intelligence for Lost Ecosystems - Ancient Oral Microbiome Database v1.0

SMILE (Systematic Microbiome Intelligence for Lost Ecosystems) is a curated relational database of ancient oral microbiome data derived from published archaeogenomic literature. Version 1.0 integrates data from 45 peer-reviewed publications (2014–2024), spanning 1,414 samples, 16,196 microbiome records, and 2,150 distinct taxa. Temporal coverage extends from approximately 102,000 BP (Pesturina Cave, Serbia — Neanderthal era) to the medieval period, with near-global geographic coverage (-175 to 178° longitude, -34 to 79° latitude). The corpus includes records for Bacteria (16,120), Archaea (53), Fungi (16), and Viruses (7), with 1,401 human host samples and 13 non-human host samples (bear, gorilla, reindeer). Authentication metric records (n=212) and methodological metadata entries (n=42) are included to support reproducibility and downstream filtering. Data were extracted using an LLM-assisted pipeline (Anthropic API: Haiku for pre/post-processing stages, Sonnet for structured extraction) applied to primary literature, with multilingual extraction coverage including English, Japanese, French, and Russian sources. All extraction scripts are available at https://github.com/satoru-bio/smile-pipeline. The database is structured as a PostgreSQL 16 relational schema with PostGIS extension. This deposition includes a full SQL dump (restorable to any PostgreSQL 16+ instance), six CSV table exports for tool-agnostic access, a dataset summary JSON, and a README with restoration instructions. Limitation: 262 records identified as figure-only in source publications remain undigitised in v1.0 and are documented as a known gap. These are distributed across approximately 20 DOIs and will be addressed in a subsequent release. This dataset supports research in palaeomicrobiology, ancient DNA, evolutionary medicine, antimicrobial resistance baseline reconstruction, and biosecurity preparedness.