A simple compound prioritization method for drug discovery considering multi-target binding.

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Drug discovery

Background:

• Active learning accelerates drug discovery by optimizing molecular properties.
• Prior methods focused on single property optimization, limiting drug candidate utility.
• Multiple properties, like binding affinity to various targets, are crucial for drug development.

Purpose of the Study:

• To develop and validate a multiobjective active learning method for ligand optimization.
• To efficiently handle multiple, computationally expensive molecular properties.
• To improve the efficiency of hit-to-lead and lead optimization in pharmaceutical research.

Main Methods:

• Introduced a novel multiobjective ligand optimization protocol.
• Applied active learning to simultaneously optimize distinct molecular properties.
• Validated the method retrospectively using docking scores.

Main Results:

• Achieved improved retrieval of top drug binders compared to greedy acquisition.
• Demonstrated efficient handling of multiple, expensive-to-compute molecular properties.
• Showed that fitting individual properties separately enhances prediction rank correlation.

Conclusions:

• The multiobjective active learning approach enhances drug discovery efficiency.
• This method effectively balances computational budgets across multiple critical properties.
• The workflow supports pharmaceutical research by optimizing compounds for multiple targets.