Hepatocyte Differentiation Protocol Database: Structured Protocols Extracted from PMC Open-Access Literature

A structured database of 878 hepatocyte differentiation protocols extracted from 544 PubMed Central open-access papers using an LLM-based text mining pipeline. The dataset covers iPSC-derived, ESC-derived, and directly reprogrammed hepatocyte protocols. Files included: protocols.db — SQLite database containing paper metadata, triage classifications, extracted protocols (cell source, culture system, differentiation stages with growth factors/small molecules/duration, endpoint assessments), citation graph, GEO accessions, RNA-seq metadata, and expression values. protocols_export.xlsx — Multi-sheet Excel workbook summarizing protocols, stages, expression trajectories, and RNA-seq metadata. triage_results.jsonl — LLM triage classifications (category, confidence, reasoning) for each paper. Protocols were extracted using a three-pass architecture (structure identification, detailed extraction with tool calling, supplement enrichment) with gpt-4o-mini at temperature 0. Average extraction confidence: 0.753. For the extraction pipeline, see: https://github.com/taiwong2/Hep-Diff-Mining