Maintenance, update and further development of EFSA's Chemical Hazards: OpenFoodTox 2.0

The present document provides the summary of the activities undertaken during the fourth of the framework contract (OC/EFSA/SCER/2018/01) to maintain, update and further develop the OpenFoodTox database (“OpenFodTox 2.0”). OpenFoodTox has been developed to map hazards data published in EFSA documents (opinions, statements, and conclusions) on risk assessment of food and feed. The repository holds summary data on identification of chemicals, document descriptors, hazard identification, and hazard characterisation. Within OpenFoodTox 2.0, the collection and entry of all hazard data assessed by EFSA scientific panels was performed according to the existing data model. During the fourth year, 141 new substances and 304 new hazard assessments (collected from new 188 EFSA documents) were added to the database. The OpenFoodTox 2.0 now includes more than 10800 assessments for over 5890 chemicals (each assessment potentially including multiple (eco‐) toxicity endpoints and/or hazard/risk characterisation data) extracted from the screening of about 2435 documents (opinions, statements, conclusions) published by EFSA from 2020 to August 2022. To maintain the database, its data model was further expanded to incorporate new data types, including physicochemical properties (OHT 1 to 23‐5) and toxicokinetic data (OHT 58). Overall, OpenFoodTox was enriched with physicochemical properties and/or ADME/PK/TK data (including quantitative values for key toxicokinetics parameters, e.g., Cmax, AUC, T‐half and T‐max) for 1,263 substances from 664 EFSA outputs. In total, 16,611 database records with experimentally‐derived physicochemical properties data were added to the database for 969 substances (collected from 605 EFSA outputs) as well as 7,699 ADME/PK/TK database records for 852 substances (from 577 documents) and are available for download under https://zenodo.org/records/8120114. Using VEGA predicted values of physiochemical properties and toxicity endpoints were added for 3980 compounds. The flexible summary wad tested on a data rich scenario using the data on conazoles. Finally, new QSAR models and a new read‐across tools (VERA) are presented here.