Stable de novo protein design via joint conformational landscape and sequence optimization

Area of Science:

• Computational biology
• Protein engineering
• Structural bioinformatics

Background:

• Generative protein modeling offers tools for designing protein sequences and structures.
• Accurate modeling of protein conformational landscapes and sequence design for stable folding remain challenges.

Purpose of the Study:

• To systematically analyze joint optimization of sequence-to-structure and structure-to-sequence mappings.
• To develop improved methods for protein conformational landscape modeling and sequence design.

Main Methods:

• Developed and analyzed a joint optimization approach for sequence and structure modeling.
• Validated the approach using large-scale protein stability measurements.
• Utilized joint models like TrRosetta and TrMRF, and combined ESMFold pLDDT with ESM2 Pseudo-likelihood.

Main Results:

• Joint optimization proved superior for designing stable proteins and predicting stability with high accuracy.
• Sequences generated by the joint model showed increased hydrophilic interactions.
• These interactions may aid in maintaining secondary structure and pairing features.

Conclusions:

• Joint optimization offers a superior strategy for generative protein modeling and design.
• The findings advance the ability to design stable proteins with predictable conformations.
• The study highlights the importance of considering sequence-structure interactions holistically.