Protein structure prediction: when is it useful?

Area of Science:

• Biomedical computational modeling
• Structural bioinformatics
• Drug discovery

Background:

• Computational protein structure prediction is vital for biomedicine, aiding tasks from family assignment to drug screening.
• Historically, prediction accuracy relied heavily on the availability of close structural templates.
• Significant advancements have been made in refining low-resolution models towards native structures.

Purpose of the Study:

• To review the progress and limitations in computational protein structure prediction techniques.
• To highlight recent improvements in model refinement using integrated knowledge and physics-based approaches.
• To identify persistent challenges in template detection and de novo structure prediction.

Main Methods:

• Combining knowledge-based information from diverse structural templates.
• Improving physics-based force fields to better model the energy landscape.
• Analyzing the state-of-the-art in homology detection and novel structure prediction.

Main Results:

• Recent progress enables refinement of low-resolution protein models closer to native states.
• Integration of multiple template sources and enhanced force fields contribute to improved model accuracy.
• Key challenges persist in identifying distantly related structural templates and predicting entirely new protein structures.

Conclusions:

• While model refinement has advanced, significant hurdles remain in homology detection and de novo structure prediction.
• Future research should focus on developing novel algorithms for remote template identification and ab initio structure prediction.
• Overcoming these limitations is crucial for expanding the applications of computational protein structure prediction.