Deep learning models for predicting Toxicity and Bioactivity of the chemical exposome: a case study for the Blood Exposome Database

Input datasets and result tables for the publication on predicting toxicity and bioactivity for the compounds in the blood exposome database. Table 2: GHS classification (make an extra column with training vs test tags also add compound properties to this table). CIDs in this table have been deduplicated. Table 3: Tox21 input data (see final output table and create a similar matrix just with train and test tags) Table 4a: Tox21 compound data (properties of compounds) Table 4b: Tox21 compound bioactivity data Tables 5: Tox21 assay list with AUC values and calculated optimal threshold for activity classification (47 assays show AUC over 0.80) Table6a: Tox21 predictions for Blood exposome (raw values) Table6b: Tox21 predictions for Blood exposome (binary values with optimal thresholds applied Note: Duplicate CIDs were included Table7a: GHS model predictions for blood exposome compounds (raw values) Table 7b: GHS model predictions for blood exposome compounds (binary values) Note: Duplicate CIDs were included Table 8 : Compound frequency by assay Table 9: Tox21 – Blood exposome-GHS overlap (Table goes to zenodo) (Should be based on CID no need to smiles standardization)