Curated Datasets
| Dataset Name | Total # of Compounds | Data Type | Dataset Source | Dataset Description |
|---|---|---|---|---|
| Acute toxicity | 7332 | -log10 mol/kg-bw | Zhu et al. (DOI: 10.1289/ehp.0800471) | Publicly available rat oral acute toxicity data with LD₅₀ values. |
| Aquatic toxicity | 675 | -log10 of Conc. (µmol/L) | Klopman et al. (https://doi.org/10.1002/etc.5620190225) | Standardized 96-hour LC₅₀ toxicity data for Pimephales promelas (fathead minnow) using publicly available toxicology resources. |
| BBB (Blood Brain Barrier) |
438 | logBB | Wang et al. (https://doi.org/10.1007/s11095-015-1687-1) | Compounds with experimental logBB values were compiled from various public resources and curated using ChemAxon and CASE Ultra tools. |
| Bioavailability | 1141 | oral bioavailability (%F) |
Kim et al. (https://doi.org/10.1007/s11095-013-1222-1) Moda et al. (DOI: 10.1016/j.bmc.2007.08.060) |
Compiled from public databases and literature sources. Chemical structures were standardized, and reported %F values were harmonized to resolve discrepancies among sources. |
| Carcinogenicity | 342 | Binary, 0=Non-Carcinogen; 1=Carcinogen | Chung et al. (https://doi.org/10.1021/acs.est.3c00648) | 342 unique organic compounds (59 carcinogenic and 283 non-carcinogenic) from the EPA’s IRIS database, labeled as carcinogenic or noncarcinogenic based on oral slope factor (OSF), a quantitative measure for oral cancer risk. |
| Cosmetics | 4129 | Activity_Cosmetics (All chemicals fall under the cosmetic category) | Chung et al. (https://doi.org/10.1021/acs.est.3c00648) | Collected from COSMOS Cosmetics Inventory knowledge base. |
| DART (Developmental and Reproductive Toxicity) |
1999 | Binary Labels (1:Safe; 0: Teratogen); Activity_DART (1:Toxic; 0:Non-toxic) |
Ciallella et al. (https://doi.org/10.1021/acs.est.2c01040) Aljarf et al. (https://doi.org/10.1021/acs.jcim.2c00824) |
Collected from public databases and literature sources, including ECHA, ToxRefDB, and the Procter & Gamble dataset, using mammalian prenatal developmental toxicity studies involving oral or inhalation exposure. EmbryoTox dataset was curated from Australian Therapeutic Goods Administration (TGA) pregnancy classification data. |
| Drugbank | 8055 | Activity_Drugbank (All chemicals fall under the drug category) | Chung et al. (https://doi.org/10.1021/acs.est.3c00648) | Collected from DrugBank. |
| Endocrine disruption | 2103 | Agonist, Antagonist, Binding, and Uterotrophic class | Ciallella et al. (DOI: 10.1038/s41374-020-00477-2) | The dataset was compiled from CERAPP, which provides in vitro estrogen receptor agonism, antagonism, and binding data, and from EADB, which contains in vivo rodent uterotrophic assay data. |
| Hepatotoxicity | 5177 | Several classification endpoints for hepatotoxicity are provided at standard or dose-based thresholds. Activity is the main endpoint. Endpoints labeled H-* represent human data, and endpoints labeled PC-* represent preclinical data. The abbreviations used include: CC – clinical chemistry, HC – hepatocellular injury, HT – hepatotoxicity, HB – hepatobiliary injury, and MF – morphological findings. | Mulliner et al. (https://doi.org/10.1021/acs.chemrestox.5b00465) | Compiled from published literature, DrugBank, PharmaPendium, Leadscope, and internal 14–28 day rat study data. |
| High Production Volume chemicals | 1672 | Activity_HPV (All chemicals fall under the HPV category) | Chung et al. (https://doi.org/10.1021/acs.est.3c00648) | Collected from the U.S. EPA High Production Volume (HPV) Challenge Program chemical database. |
| Httk_ADME_Parameters | 1449 | Human CLint (intrinsic clearance) in µL/min/10^6 cells, Human Fu (fraction unbound), Rat CLint (intrinsic clearance) in µL/min/10^6 cells, and Rat Fu (fraction unbound) | High-Throughput Toxicokinetics (https://cran.r-project.org/web/packages/httk/index.html) | The HTTK dataset, developed by the U.S. EPA, contains high-throughput toxicokinetic data and models covering pharmaceuticals and environmental chemicals. It includes in vitro measurements like plasma protein binding and hepatic clearance rates, as well as species-specific physiological data such as tissue volumes and blood flow rates. |
| Natural products | 2479 | Activity_NP (All chemicals fall under the natural product category) | Chung et al. (https://doi.org/10.1021/acs.est.3c00648) | Curated from the Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP). |
| Pesticides | 1009 | Activity_Pesticides (All chemicals fall under the pesticide category) | Chung et al. (https://doi.org/10.1021/acs.est.3c00648) | Collected from literature and public databases, including the U.S. EPA CompTox Chemistry Dashboard. |
| Transporter-mediated toxicity (Classification data) | 10,875 |
Proteins: • P-glycoprotein (P-gp; MDR1) • Breast cancer resistance protein (BCRP) • Multidrug resistance–associated proteins (MRP1, MRP2) • Bile salt export pump (BSEP) Activity values: • Substrate and Inhibitor calls • Evidence of inhibition thresholds. |
Daood et al. (https://doi.org/10.1021/acs.molpharmaceut.5c01065) Sedykh et al. (https://doi.org/10.1007/s11095-012-0935-x) Zhao et al. (https://doi.org/10.1021/acsomega.7b00274) |
Consolidates curated transporter data across P-gp, BCRP, MRP1, MRP2, and BSEP. Records were standardized and manually reviewed for cited literature sources, noting details such as cell lines, substrates, substrate concentrations, and positive controls where available. Sources include large-scale ChEMBL bioactivity extractions, PubChem, and Metrabase data. High-confidence entries from the Intestinal Transporter Database are included for MDR1. Substrate/inhibitor calls are harmonized, and inhibition evidence follows study-specific thresholds (10 µM for P-gp/BCRP; 100 µM for BSEP). |
| Transporter-mediated toxicity (Regression data) | 2223 |
Proteins: • P-gp • BCRP • MRP1 Activity values: • -log(IC50) or pIC50 |
ChEMBL (doi: 10.1093/nar/gkad1004) Papyrus (https://doi.org/10.1186/s13321-022-00672-x) |
Curated in vitro chemical inhibition data for P-gp, BCRP and MRP1 were collected from ChEMBL and Papyrus. Records were standardized and activities were harmonized using available pChEMBL values. When duplicate entries were present for the same chemical, one representative entry was retained using the average reported activity value. |