Machine learning techniques and drug design.

Area of Science:

• Computational Chemistry
• Medicinal Chemistry
• Pharmacology

Background:

• Drug design is a complex process necessitating hybrid methodologies.
• Machine learning techniques (MLT) have emerged as powerful tools in this field.
• The application of MLT in drug design has seen a notable increase in recent decades.

Purpose of the Study:

• To review and critically evaluate various MLT applicable to drug design.
• To compare the performance of MLT against traditional statistical methods.
• To highlight the current state and future trends of MLT in medicinal chemistry.

Main Methods:

• Review of selected MLT: self-organizing maps, multilayer perceptron, Bayesian neural networks, counter-propagation neural networks, and support vector machines.
• Comparative analysis of MLT performance versus classical statistical methods like partial least squares and multiple linear regression.
• Exploration of MLT applications in quantitative structure-activity relationship (QSAR) modeling and virtual screening.

Main Results:

• MLT demonstrate significant advantages compared to classical statistical methods in drug design tasks.
• Support vector machines are particularly prominent in current medicinal chemistry studies.
• MLT-derived models show promise for constructing reliable QSAR models and aiding in the discovery of new chemicals.

Conclusions:

• MLT are valuable tools for drug design, offering enhanced capabilities for modeling and prediction.
• Future trends indicate a growing role for MLT in predicting pharmacokinetic and toxicity properties, thereby reducing clinical trial failures.
• The critical evaluation presented underscores the potential of MLT to revolutionize drug discovery and development.