Computational profiling of bioactive compounds using a target-dependent composite workflow.

Area of Science:

• Computational chemistry
• Chemoinformatics
• Pharmacology

Background:

• Computational target fishing aids in understanding drug mechanisms, predicting side effects, and drug repurposing.
• Current methods often rely on single computational approaches for target-ligand association prediction.

Purpose of the Study:

• To present an automated workflow integrating multiple computational methods for enhanced target fishing.
• To improve the accuracy and scope of identifying novel target-ligand associations.

Main Methods:

• Developed an automated workflow combining four ligand-based methods (SVM classification, SVR affinity prediction, nearest neighbors interpolation, shape similarity) and two structure-based methods (docking, pharmacophore matching).
• Applied ligand and target property checks to guide the sequential application of these methods.
• Validated the workflow's accuracy in identifying known targets for 189 clinical candidates.

Main Results:

• The workflow achieved 72% accuracy in identifying the primary targets of clinical drug candidates.
• Two novel off-target interactions were proposed and experimentally validated.
• Rolofylline was confirmed to inhibit phosphodiesterase 5, and PF-2545920 showed strong binding to the cysteinyl leukotriene type 1 receptor.

Conclusions:

• The integrated computational workflow offers a robust and accurate approach for target fishing.
• This method facilitates the discovery of novel drug-target interactions and potential therapeutic applications.
• Experimental validation confirmed the predictive power of the workflow for both on-target and off-target activities.