Class imbalance learning with Bayesian optimization applied in drug discovery.

Area of Science:

• Computational chemistry
• Artificial intelligence in drug discovery
• Machine learning for pharmacology

Background:

• Machine intelligence (MI) offers potential for cost reduction in drug development.
• A key challenge in MI is the trade-off between interpretability and predictive performance.
• Machine learning models are interpretable but less performant than deep learning models.

Purpose of the Study:

• To propose a novel pipeline, Class Imbalance Learning with Bayesian Optimization (CILBO), to improve machine learning model performance.
• To address the performance limitations of interpretable machine learning models in drug discovery.
• To provide an alternative approach to accelerate drug discovery and reduce costs.

Main Methods:

• Development of the Class Imbalance Learning with Bayesian Optimization (CILBO) pipeline.
• Implementation of a machine learning model using the CILBO pipeline for predicting antibacterial candidates.
• Comparative analysis of the CILBO model's performance against a deep learning model.

Main Results:

• The CILBO pipeline successfully enhanced the performance of machine learning models.
• The developed model demonstrated comparable antibacterial prediction performance to a leading deep learning model.
• The CILBO pipeline offers a viable alternative for accelerating drug discovery screenings.

Conclusions:

• The CILBO pipeline effectively improves machine learning model performance in drug discovery.
• This approach provides a competitive alternative to deep learning models for drug activity prediction.
• CILBO facilitates faster preliminary screenings and lowers overall drug development costs.