MACAW: An Accessible Tool for Molecular Embedding and Inverse Molecular Design.

Area of Science:

• Computational Chemistry
• Synthetic Biology
• Drug Discovery

Background:

• Synthetic biology and organic chemistry require advanced tools for targeted molecule synthesis.
• Existing methods often introduce uninformative dimensions, hindering accurate property prediction.

Purpose of the Study:

• To introduce Molecular AutoenCoding Auto-Workaround (MACAW), a novel computational tool.
• To enable the generation of molecules with specific, desired properties.
• To facilitate inverse molecular design for various applications.

Main Methods:

• MACAW embeds molecules into a smooth, multidimensional numerical space.
• It uses these embeddings as features for accurate molecular property prediction.
• A generative algorithm combined with predictive capabilities enables inverse design.

Main Results:

• Demonstrated accurate prediction of molecular properties including octane number, flash point, and receptor binding affinity.
• Successfully performed virtual screening to identify drug candidates with specific binding profiles.
• Recommended molecules with targeted octane numbers for biofuel applications.

Conclusions:

• MACAW offers an efficient and effective approach for molecular property prediction and generation.
• The tool is well-suited for small- to medium-sized datasets common in biosciences.
• MACAW complements traditional retrosynthesis by enabling molecule recommendation based on desired specifications.