Chemoinformatic approaches for navigating large chemical spaces

Area of Science:

• Cheminformatics
• Computational Chemistry
• Drug Discovery

Background:

• Large chemical spaces (CSs) encompass diverse molecular collections, from traditional libraries to virtual spaces explored by AI.
• Navigating these vast CSs presents significant computational challenges due to their sheer size and complexity.

Purpose of the Study:

• To provide an overview of different types of large chemical spaces.
• To discuss suitable molecular representations and similarity metrics for large CS exploration.
• To summarize navigation strategies within generative models for CSs.

Main Methods:

• Review of existing chemoinformatic approaches for CS navigation.
• Discussion of molecular representations and similarity metrics.
• Analysis of navigation techniques employed by deep generative models.

Main Results:

• Different types of large CSs necessitate distinct chemoinformatic navigation methods.
• Efficient CS navigation requires scalable algorithms and intelligent molecule selection.
• Deep generative models show potential for implicit feature learning but require experimental validation.

Conclusions:

• The diversity of large chemical spaces demands specialized computational tools.
• Deep generative models offer a promising avenue for efficient CS exploration, pending experimental validation.
• Further research is needed to bridge the gap between virtual CS exploration and real-world chemical discovery.