Controlled Protein Design via Statistical Energy Functions: A Rossmann Fold Case Study

Area of Science:

• Computational Biology
• Protein Engineering
• Structural Biology

Background:

• Protein design has advanced with computational methods, including statistical models and deep learning.
• Statistical models are vital for understanding protein physical properties, complementing deep learning approaches.

Purpose of the Study:

• To investigate the continued relevance and controllability of statistical models in protein design.
• To design a novel Rossmann fold protein using established statistical energy functions.

Main Methods:

• Utilized SCUBA (Side chain Unknown Backbone Arrangement) for scaffold design and ABACUS2 (A Backbone-based Amino Acid Usage Survey) for sequence design.
• Applied tailored restraints to incorporate specific structural features, generating 300 low-energy sequence candidates.
• Filtered sequences using AlphaFold2 predictions and performed experimental validation on nine selected designs.

Main Results:

• 69% of designed sequences showed high-confidence fold similarity via AlphaFold2.
• One designed protein's crystal structure (1.8 Å resolution) had a main-chain deviation of 2.602 Å from the computational model.
• The experimental structure closely matched the designed model, validating the imposed constraints.

Conclusions:

• Statistical models offer significant controllability in protein design, enabling accurate prediction and creation of specific protein folds.
• This study validates the effectiveness of statistical energy functions in de novo protein design.
• Findings provide valuable insights for future protein engineering applications using statistical modeling.