ToxProfiler: Toxicity-target profiler based on chemical similarity

Abstract

Identifying the ability of a chemical to interact with toxicity targets, such as proteins in an adverse outcome pathway, is an essential step in drug discovery and risk assessment. Computational approaches to screen for chemical-toxicity target interaction can serve as a rapid alternative to traditional in vitro/in vivo methods. In this work, we have developed a chemical-similarity based protocol that predicts the potential of a chemical to interact with 64 established toxicity targets. In particular, we created a chemogenomics database from public data sources to identify target representatives, i.e., chemicals that are known to interact with the selected targets. We evaluated the performance of 2D and 3D similarity approaches in correctly ranking known interacting compounds using an external evaluation set from ChEMBL database. We found that the 2D approach outperforms the 3D approach in target prediction. Here, we developed a publically available toxicity profiler website (https://toxpro.bhsai.org/) using 2D similarity-based screening approach that allows user to obtain toxicity target profile for a set of query compounds. We utilized the profiler to screen 649 known acute and highly toxic chemicals with a Globally Harmonized System (GHS) score of less than 2. In this set, acetylcholinesterase was the most frequently occurring target underlying toxicity. The developed toxicity profiler tool provides a rapid means to screen for mechanisms underlying chemical toxicity.