This file contains value of the model training set confusion matrix, domain of applicability evaluation based on training set to predicted chemicals structural similarity, and 75th percentile bioactivity index values for each QSUR model. This dataset is associated with the following publication: Phillips, K., J. Wambaugh, C. Grulke, K. Dionisio, and K. Isaacs. High-throughput screening of chemicals as functional substitutes using structure-based classification models. GREEN CHEMISTRY. Royal Society of Chemistry, Cambridge, UK, 19: 1063-1074, (2017).

There are five different files for this dataset: 1. A dataset listing the reported functional uses of chemicals (FUse) 2. All 729 ToxPrint descriptors obtained from ChemoTyper for chemicals in FUse 3. All EPI Suite properties obtained for chemicals in FUse 4. The confusion matrix values, similarity thresholds, and bioactivity index for each model. 5. The functional use prediction, bioactivity index, and prediction classification (poor prediction, functional substitute, candidate alternative) for each Tox21 chemical. This dataset is associated with the following publication: Phillips, K., J. Wambaugh, C. Grulke, K. Dionisio, and K. Isaacs. High-throughput screening of chemicals as functional substitutes using structure-based classification models. GREEN CHEMISTRY. Royal Society of Chemistry, Cambridge, UK, 19: 1063-1074, (2017).

Additional details used in the methods are found in the MS Word file “S1_Dawson et al._Supporting_Information.docx”. The MS Excel file “S2_Dawson et al. Supporting Information.xlsx” contains datasets and graphical results. The Excel file sheets are as follows: S2.1 illustrates Clint hepatic flow calculations, S2.2 - 5 include training and test data sets; S2.6-7 include figures illustrating Clint model selection criteria and assemblages of model descriptors; S2.8 includes confusion matrices for evaluation Clint model, S2.9-10 include figures illustrating fup model selection criteria and assemblages of model descriptors (with ranges); S2.11 includes tables of model assessments of the Clint test set, S2.12 includes information relevant to BER calculations for the ToxCast test set, S2.13 includes information relevant to BER calculations for Tox21 chemicals, and S2.14 provides information on different transformations for fup. This dataset is associated with the following publication: Dawson, D., B. Ingle, K. Phillips, J. Nichols, J. Wambaugh, and R. Tornero-Velez. Designing QSARs for Parameters of High-Throughput Toxicokinetic Models Using Open-Source Descriptors. ENVIRONMENTAL SCIENCE & TECHNOLOGY. American Chemical Society, Washington, DC, USA, 55(9): 6505, (6517).

Here we describe the development of an automated KNIME workflow to curate and correct errors in the structure and identity of chemicals using the publically available PHYSPROP physico-chemical properties and environmental fate datasets. The workflow first assembles structure-identity pairs using up to four provided chemical identifiers, including chemical name, CASRNs, SMILES, and MolBlock. Problems detected included errors and mismatches in chemical structure formats, identifiers, and various structure validation issues, including hypervalency and stereochemistry descriptions. Subsequently, a machine learning procedure was applied to evaluate the impact of this curation process. The performance of QSAR models built on only the highest quality subset of the original dataset was compared to the larger curated and corrected data set. The latter showed statistically improved predictive performance. The final workflow was used to curate the full list of PHYSPROP datasets, and is being made publically available for further usage and integration by the scientific community. This dataset is associated with the following publication: Mansouri, K., C. Grulke, A. Richard, R. Judson, and A. Williams. (SAR AND QSAR IN ENVIRONMENTAL RESEARCH) An automated curation procedure for addressing chemical errors and inconsistencies in public datasets used in QSAR modeling. SAR AND QSAR IN ENVIRONMENTAL RESEARCH. Taylor & Francis, Inc., Philadelphia, PA, USA, 27(11): 911-937, (2016).

Here we describe the development of an automated KNIME workflow to curate and correct errors in the structure and identity of chemicals using the publically available PHYSPROP physico-chemical properties and environmental fate datasets. The workflow first assembles structure-identity pairs using up to four provided chemical identifiers, including chemical name, CASRNs, SMILES, and MolBlock. Problems detected included errors and mismatches in chemical structure formats, identifiers, and various structure validation issues, including hypervalency and stereochemistry descriptions. Subsequently, a machine learning procedure was applied to evaluate the impact of this curation process. The performance of QSAR models built on only the highest quality subset of the original dataset was compared to the larger curated and corrected data set. The latter showed statistically improved predictive performance. The final workflow was used to curate the full list of PHYSPROP datasets, and is being made publically available for further usage and integration by the scientific community. This dataset is associated with the following publication: Mansouri, K., C. Grulke, A. Richard, R. Judson, and A. Williams. (SAR AND QSAR IN ENVIRONMENTAL RESEARCH) An automated curation procedure for addressing chemical errors and inconsistencies in public datasets used in QSAR modeling. SAR AND QSAR IN ENVIRONMENTAL RESEARCH. Taylor & Francis, Inc., Philadelphia, PA, USA, 27(11): 911-937, (2016).

The MS Excel file (Dawson et al S2 Supporting information.xlsx) contains multiple sheets containing the training sets, test sets, and predictions for intrinsic metabolic clearance (Clint), fraction unbound in plasma (fup), and bioactivity-exposure ratios (BER), for ToxCast and pharmaceutical-like chemicals. The Word file (Dawson et al S1 Supporting Information.docx) provides additional supporting information on assembly of the training and test sets for Clint, fup, and BER. The data dictionary describes the terms used in the supporting information, S1 and S2. This dataset is associated with the following publication: Dawson, D., B. Ingle, K. Phillips, J. Nichols, J. Wambaugh, and R. Tornero-Velez. Designing QSARs for Parameters of High-Throughput Toxicokinetic Models Using Open-Source Descriptors. ENVIRONMENTAL SCIENCE & TECHNOLOGY. American Chemical Society, Washington, DC, USA, 55(9): 6505-6517, (2021).

The MS Excel file (Dawson et al S2 Supporting information.xlsx) contains multiple sheets containing the training sets, test sets, and predictions for intrinsic metabolic clearance (Clint), fraction unbound in plasma (fup), and bioactivity-exposure ratios (BER), for ToxCast and pharmaceutical-like chemicals. The Word file (Dawson et al S1 Supporting Information.docx) provides additional supporting information on assembly of the training and test sets for Clint, fup, and BER. The data dictionary describes the terms used in the supporting information, S1 and S2. This dataset is associated with the following publication: Dawson, D., B. Ingle, K. Phillips, J. Nichols, J. Wambaugh, and R. Tornero-Velez. Designing QSARs for Parameters of High-Throughput Toxicokinetic Models Using Open-Source Descriptors. ENVIRONMENTAL SCIENCE & TECHNOLOGY. American Chemical Society, Washington, DC, USA, 55(9): 6505-6517, (2021).

The dataset has all of the information used to create and evaluate 3 independent QSAR models for the fraction of a chemical unbound by plasma protein (Fub) for environmentally relevant chemicals. In vitro plasma protein values for 1245 pharmaceuticals and 406 ToxCast chemicals were collected from the literature (Obach 2008, Zhu 2013, Wetmore 2012, Wetmore 2015). The 21 descriptors calculated by MOE that were used in the models are included, as is an acid/base/neutral/zwitterions classification based on ionization percentages calculated in ADMET Predictor. Finally, the dataset includes the in silico Fub predictions for each chemical from the constructed k-nearest neighbor, support vector machine, and random forest QSAR models, as well as a consensus (average) prediction. This dataset is associated with the following publication: Ingle, B., R. Tornero-Velez, J. Nichols, and B. Veber. Informing the Human Plasma Protein Binding of Environmental Chemicals by Machine Learning in the Pharmaceutical Space: Applicability Domain and Limits of Predictability. Journal of Chemical Information and Modeling. American Chemical Society, Washington, DC, USA, 56(11): 2243-2252, (2016).

We have performed high-throughput experimental estimates of five physicochemical properties for a set of 200 chemicals to evaluate the consistency with previous measurements, factors impacting consistency and experimental success, and the applicability domain of the new data in relation to previously measured data and predictive models. This dataset is associated with the following publication: Nicolas, C., K. Mansouri, K. Phillips, C. Grulke, A. Richard, A. Williams, J. Rabinowitz, K. Isaacs, A. Yau, and J. Wambaugh. (ENVIRONMENTAL SCIENCE and TECHNOLOGY) Rapid Experimental Estimates of Physicochemical Properties to Inform Models and Testing. ENVIRONMENTAL SCIENCE & TECHNOLOGY. American Chemical Society, Washington, DC, USA, 636: 901-909, (2018).

Observed fup data obtained from the literature; SMILES structures obtained from Pubchem SDF files. This dataset is associated with the following publication: Yun, Y.E., R. Tornero-Velez, T. Purucker, D. Chang, and A.N. Edginton. Evaluation of Quantitative Structure Property Relationship algorithms for predicting plasma protein binding in humans. Computational Toxicology. Elsevier B.V., Amsterdam, NETHERLANDS, 17: 100142, (2021).