Curated PFAS adsorption database mined from the literature for activated carbon, ion-exchange and nonionic resins, and cyclodextrin polymers

Literature-derived database of per- and polyfluoroalkyl substance (PFAS) adsorption on activated carbon, ion-exchange and nonionic resins, and cyclodextrin polymers. Data were mined from 177 screened studies with an LLM-based workflow (classification, enumeration, and extraction with GPT-4.1), then merged, harmonized, converted to a common distribution coefficient (Kd), quality-controlled, and reviewed against the source articles. Contents: records_all: all 11,448 standardized extracted records with the processing outcome of each record final_database: the 3,776 equilibrium records (89 studies, 128 PFAS) used for machine-learning modeling pfas_descriptors: PFAS identifiers, SMILES, classes, and RDKit descriptors commercial_adsorbent_properties, included_studies, extraction_fields: Supplementary Tables 3, 5, and 6 of the accompanying article data_dictionary: description and unit of every column Each table is provided as CSV and as a sheet in PFAS_adsorption_database.xlsx. See README.md for the record flow, conventions, and changes from v1.0.0. This dataset accompanies the article “Unlocking Heterogeneous PFAS Adsorption Literature for Data-driven Adsorbent Selection” (Yan, Ransom, Pitera, Sanders and Helbling). Code: https://doi.org/10.5281/zenodo.22821995