Increasing the efficiency and accuracy of the ABACUS protein sequence design method

Area of Science:

• Computational biology
• Protein engineering
• Bioinformatics

Background:

• The ABACUS method uses statistical energy functions for protein sequence design, but its accuracy and efficiency require improvement.
• Key components were not optimized for sequence design or integrated effectively, limiting performance.

Purpose of the Study:

• To enhance the accuracy and computational efficiency of the ABACUS method for protein sequence design.
• To optimize components for better performance in sequence design and computational applications.

Main Methods:

• Developed a solvent accessibility model with improved mutual information with residue types.
• Optimized rotamers for accurate sidechain atomic position approximation.
• Devised an empirical function for inter-atomic packing, fitted to native structures.
• Accelerated energy calculations using interpolation in high-dimensional feature spaces.

Main Results:

• ABACUS2 accurately reproduced native sidechain conformations in repacking tests.
• Achieved a 37.7% native residue type recovery rate, surpassing ABACUS1's 32.7%.
• Experimentally verified that proteins designed using ABACUS2 on native backbones are well-folded.

Conclusions:

• ABACUS2 offers improved accuracy and efficiency for protein sequence design.
• The optimized components and accelerated calculations enable broader applications.
• Experimental validation confirms the utility of ABACUS2 for designing functional proteins.