Differential Compound Prioritization via Bidirectional Selectivity Push with Power.

Area of Science:

• Computational chemistry
• Drug discovery
• Machine learning

Background:

• Effective in silico compound prioritization is crucial for identifying drug candidates.
• Current methods often focus on single-target activity, neglecting compound selectivity.
• Selectivity is vital for minimizing undesired side effects in drug development.

Purpose of the Study:

• To present a novel machine learning-based differential compound prioritization method (dCPPP).
• To improve the simultaneous ranking of active and selective compounds.
• To enhance the identification of promising drug candidates with minimal side effects.

Main Methods:

• Developed a differential compound prioritization method (dCPPP) using machine learning.
• Implemented a bidirectional selectivity push strategy to prioritize selective compounds.
• Determined push powers based on ranking differences across multiple bioassays.

Main Results:

• The dCPPP method effectively ranks active compounds.
• dCPPP preferentially ranks selective compounds higher than baseline models.
• Experimental results show significant improvement in prioritizing selective compounds.

Conclusions:

• The novel dCPPP method enhances compound prioritization by considering both activity and selectivity.
• This approach aids in identifying safer and more effective drug candidates.
• dCPPP offers a significant advancement over existing computational prioritization techniques.