Probabilistic Approach to Generating MPOs and Its Application as a Scoring Function for CNS Drugs

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Multiparameter optimization (MPO) scoring functions are vital for guiding molecule design in medicinal chemistry.
• Existing MPO methods can be complex, involving numerous parameters.

Purpose of the Study:

• To introduce and evaluate a novel probabilistic MPO (pMPO) scoring function.
• To assess the utility of pMPO for designing central nervous system (CNS) drugs and predicting blood-brain barrier penetration.

Main Methods:

• Developed a pMPO approach combining statistically determined empirical boundaries with probability distributions of desired molecules.
• Defined desirability functions using a minimal set of parameters.

Main Results:

• The pMPO method demonstrated accuracy comparable to Bayesian models in distinguishing desired from undesired molecules using a test set of orally approved drugs.
• The approach offers enhanced human interpretability compared to traditional Bayesian models.
• The pMPO method's applicability to blood-brain barrier penetrant drugs was also demonstrated.

Conclusions:

• The pMPO scoring function provides a powerful and interpretable tool for drug design.
• This method effectively balances predictive power with a reduced parameter set.
• pMPO shows promise for optimizing the design of CNS-penetrant drugs.