ProfhEX: Empowering Early Drug Discovery with Machine Learning-Based Target Profiling and Liability Prediction.

Area of Science:

• Computational chemistry
• Pharmacology
• Drug discovery and development

Background:

• Drug discovery is lengthy, costly, and faces high attrition rates due to safety, efficacy, and regulatory issues.
• Artificial intelligence (AI) methods accelerate drug discovery by predicting molecular activity and identifying drug targets.
• Existing ligand-based prediction tools have limitations in data coverage, target scope, and usability.

Purpose of the Study:

• To present an enhanced version of the ProfhEX platform for comprehensive drug-target activity profiling.
• To improve the scalability, user-friendliness, and predictive accuracy of computational drug discovery tools.

Main Methods:

• Developed an enhanced ProfhEX platform with 969 predictive models for 693 human targets.
• Trained models on over 5 million curated bioactivity data points.
• Evaluated predictive accuracy in real-world scenarios and compared against state-of-the-art tools.

Main Results:

• The enhanced ProfhEX platform demonstrates high predictive accuracy in prospective studies.
• ProfhEX surpasses existing tools in primary target prediction benchmarks.
• The platform provides comprehensive compound-target activity profiling.

Conclusions:

• ProfhEX is one of the largest and most accurate platforms for compound-target prediction.
• The enhanced platform supports early-stage drug discovery and improves target liability assessment.
• ProfhEX addresses limitations of current prediction tools, enhancing drug development efficiency.