Developing structure-activity relationships for the prediction of hepatotoxicity

Area of Science:

• Toxicology
• Medicinal Chemistry
• Computational Chemistry

Background:

• Drug-induced liver injury (DILI) is a significant concern, leading to drug withdrawals.
• Understanding chemical structure-activity relationships (SARs) aids in early toxicity prediction.
• Existing toxicity data is often fragmented and unstructured in scientific literature.

Purpose of the Study:

• To explore the feasibility of collecting and utilizing published data to develop new SARs for hepatotoxicity.
• To expand the limited existing knowledge on SARs for liver toxicity.
• To create a searchable database of hepatotoxicity data.

Main Methods:

• Compiled hepatotoxicity data from literature reviews to build a structure-searchable database.
• Analyzed the database to identify chemical classes associated with liver toxicity.
• Searched literature for additional evidence and incorporated findings into the database.
• Developed SARs for 38 chemical classes based on data for over 1266 chemicals.
• Implemented SARs as structural alerts in Derek for Windows (DfW) for in silico predictions.

Main Results:

• Successfully developed SARs for 38 chemical classes, encompassing data from 1266 chemicals.
• Implemented these SARs as structural alerts within the DfW expert system.
• Evaluation showed 56% overall concordance, with 73% specificity and 46% sensitivity for DfW predictions.
• The approach demonstrated the derivation of SARs for complex endpoints from published data.

Conclusions:

• It is feasible to derive SARs for complex toxicological endpoints like hepatotoxicity from published literature.
• This approach supports in silico toxicity assessment of new chemical entities during drug development.
• The developed SARs, implemented as structural alerts, provide transparent and supported toxicity predictions.