You are here:
Open Source QSAR Models For pKa Prediction Using Multiple Machine Learning Approaches
Citation:
Mansouri, K., N. Cariello, A. Korotcov, V. Tkachenko, W. Casey, N. Kleinstreuer, D. Allen, Chris Grulke, AND A. Williams. Open Source QSAR Models For pKa Prediction Using Multiple Machine Learning Approaches. Presented at The American Society for Cellular and Computational Toxicology (ASCCT), Bethesda, MD, September 11, 2018. https://doi.org/10.23645/epacomptox.7157207
Impact/Purpose:
Poster presented at American Society for Cellular and Computational Toxicology (ASCCT) meeting in September 2018. An automated QSAR data preparation workflow was applied to a public data set of 7912 chemicals, created three data subsets, Acidic, Basic and Combined. The best models were compared and benchmarked with two commercial predictors showing different levels of concordance. The models and source codes will be available for download and use. This modeling effort will help provide predicted pKa values for all ionizable chemicals in the EPA DSSTox database.
Description:
The logarithmic dissociation constant, pKa, strongly influences a chemical’s pharmacokinetic and biochemical properties. pKa is an important parameter for physiologically based pharmacokinetic (PBPK) modeling, in vitro to in vivo extrapolation (IVIVE), and predicting tissue:plasma partition coefficients. Commercial software tools such as ACD/Labs and ChemAxon predict the pKa of individual ionization sites independently of chemical class. However, current publicly available pKa models are limited to certain chemical classes. Here we provide free, open-source, fast, and reliable options for predicting pKa for heterogeneous chemical classes. The best models were compared and benchmarked with two commercial predictors showing different levels of concordance. The models and source codes will be available for download and use. This modeling effort will help provide predicted pKa values for all ionizable chemicals in the EPA DSSTox database. Predictions will be available on the EPA’s CompTox Chemistry Dashboard (https://comptox.epa.gov).
URLs/Downloads:
DOI: Open Source QSAR Models For pKa Prediction Using Multiple Machine Learning Approaches
PKA ASCCT POSTER.PDF (PDF, NA pp, 1410.368 KB, about PDF)