Development of a Profiler for Facile Chemical Reactivity Using the Open-Source Konstanz Information Miner

Sanjeeva Joseph Wijeyesakere, Daniel M. Wilson, Raja S. Settivari, Tyler R. Auernhammer, Amanda K. Parks, Mary Sue Marty · Applied In Vitro Toxicology · 2018

It is prudent to assess the chemical reactivity of compounds during early stages of product development, given that formation of covalent adducts between electrophilic compounds and biological nucleophiles serves as the molecular initiating event in a range of adverse outcome pathways (AOPs). In vitro assays such as the direct peptide reactivity assay (DPRA) have been developed to assess facile reactivity. However, these approaches still require test material and laboratory resources, are less conducive to rapid high-throughput screening than computational approaches, and have been developed with a primary focus on a single AOP (e.g., skin sensitization). Given the known set of specific mechanisms that drive facile chemical reactivity between electrophilic compounds and biological nucleophiles, we developed an in-house profiler to identify reactive moieties using an open-source Konstanz Information Miner (KNIME) workflow. This covered the chemical space for five known classes of chemical facile reactivity: (1) Michael acceptor, (2) Schiff-base formation, (3) acylation, (4) nucleophilic aromatic substitution (SNAr), and (5) bimolecular nucleophilic substitution (SN2) reactions. When we assessed our profiler by screening a database of 162 chemicals with in vitro DPRA study data, it showed high sensitivity (100%) and specificity (74%), yielding an overall accuracy of 89%. The positive and negative predictive values for facile reactivity were 84% and 100%, respectively, providing a conservative screening approach, where potentially positive compounds can be further scrutinized, whereas negative compounds can be given lower priority for reactivity concerns. We anticipate these findings will have broad applicability to various toxicity endpoints.

Read the paper · More papers on PaperTik