Toward better refinement of comparative models: predicting loops in inexact environments

Area of Science:

• Computational biology
• Structural bioinformatics
• Protein modeling

Background:

• Comparative protein modeling aims for atomic accuracy but is limited by refinement challenges.
• Existing methods struggle to refine homology models due to complex sampling problems involving loops and surrounding residues.

Purpose of the Study:

• To investigate the impact of side-chain inaccuracies on loop refinement in homology models.
• To develop an improved loop refinement method that addresses these surrounding structural errors.

Main Methods:

• Perturbed native protein structures by altering loop and surrounding side-chain conformations.
• Applied a previously published loop prediction method and an augmented method incorporating simultaneous side-chain optimization.
• Evaluated loop refinement accuracy using backbone root-mean-square deviation (RMSD) on native and perturbed structures.

Main Results:

• Side-chain perturbations significantly increased the difficulty of loop prediction, with median RMSDs increasing substantially.
• The augmented loop prediction method, which optimizes surrounding side chains, demonstrated significant improvement in loop refinement accuracy on perturbed cases.
• The new method successfully predicted loops closer to native conformations in comparative models from blind tests.

Conclusions:

• Simultaneous optimization of surrounding side chains is crucial for accurate loop refinement in homology models.
• This work presents a significant step towards achieving high-accuracy comparative protein models.
• Further development is needed for full comparative model refinement, but this method addresses a key challenge.