Balancing the Functionality and Biocompatibility of Materials with a Deep-Learning-Based Inverse Design Framework.

Area of Science:

• Molecular design
• Computational chemistry
• Drug discovery

Background:

• Rational molecular design for specific functions is challenging.
• Ensuring chemical safety and sustainability is vital for market viability.
• Balancing molecular functionality with biocompatibility requires advanced methods.

Purpose of the Study:

• To introduce a novel deep learning framework for the inverse design of molecules.
• To enable the design of molecules with both desired functionality and biocompatibility.
• To address the limitations of traditional molecular design approaches.

Main Methods:

• Developed a deep learning framework with two predictive and one generative model.
• Utilized a virtual chemical space for targeted screening of novel molecules.
• Implemented an inverse design process for generating molecules with specific properties.

Main Results:

• Successfully generated molecules with specified motifs and compositions.
• Discovered synthetically accessible molecules targeting functional and safe properties.
• Designed ionic liquids (ILs) with enhanced antibacterial activity and reduced cytotoxicity.

Conclusions:

• The deep learning framework offers a versatile approach for molecular inverse design.
• The method effectively balances molecular functionality with biocompatibility.
• This approach facilitates the development of safer and more effective chemical products.