Toxicity to various tissues, organs or systems is a common reason for drug failure. There are many in vitro assays to detect toxicity well before a compound reaches pre-clinical testing, but compound synthesis itself can be a costly undertaking.
Virtual screening techniques are now routinely applied to identify compounds with potential undesirable or ineffective characteristics prior to incurring costs associated with synthesis or in vitro testing. Predicting chemical- or drug-induced toxicity using computational modelling techniques provides a valuable insight into possible liability and enables early stage pipeline prioritisation. Use of these models in industries such as the cosmetics and personal care markets is especially important where preclinical toxicity testing in animals is banned in a growing number of countries throughout the world.
chemTox is a new workflow solution which predicts several important toxicity parameters directly from chemical structure, without requiring compound synthesis, physicochemical data or toxicity data. In addition to Ames mutagenicity and rat acute dose LD50 (intravenous or oral), chemTox also predicts aqueous solubility . It is a custom-built node for the open-source KNIME analytics platform, which executes models in a workflow-based environment. Required input is chemical structure in a standard format such as SMILES, mol or sdf. The results are delivered as a KNIME data table, which can be saved as a file or database, or used as input into ongoing workflow processes. The models themselves are based on quantitative-structure property relationships (QSPRs) that calculate toxicity properties based on known responses of existing compounds. By using a compound’s molecular structure to gain insight into its toxic potential, medicinal chemists can work to optimise physical properties to mitigate risk and maximise therapeutic effect earlier in the drug discovery process.