Computer models predict physicochemical properties
Scientists from the National Toxicology Program and the U.S. Environmental Protection Agency, along with their collaborators, have built new computer models that use molecular structures to estimate the physicochemical features of a wide range of chemicals. The models may provide useful tools for researchers to rapidly assess toxicity of these chemicals to humans.
The movement of substances in the body, such as pesticides, food additives, and fragrances, is determined by their physicochemical properties. Because measuring the movement of these substances in living systems is time-consuming and costly, the researchers in this study developed quantitative structure-property relationship models.
The scientists collected and curated data sets from an open source to train and validate models that predicted six key properties: solubility, melting point, boiling point, vapor pressure, octanol-water partition coefficient, and bioconcentration factor. Different modeling methods, from linear regression to machine learning, were used and their predictions compared. The new models performed better than the existing platform, and the estimated values were well matched to the experimental data, according to the authors. (QX)
Citation: Zang Q, Mansouri K, Williams AJ, Judson RS, Allen DG, Casey WM, Kleinstreuer NC. 2017. In silico prediction of physicochemical properties of environmental chemicals using molecular fingerprints and machine learning. J Chem Inf Model 57(1):36–49.