Accurate modeling of protein conformation by automatic segment matching

Area of Science:

• Structural biology
• Computational biology
• Biophysics

Background:

• Protein structure determination is crucial for understanding biological function.
• Existing methods for de novo protein structure modeling can be computationally intensive and may require significant human intervention.
• Accurate protein models are essential for drug discovery and functional analysis.

Purpose of the Study:

• To develop and validate an automated method for de novo protein structure modeling using segment matching.
• To assess the accuracy and efficiency of the segment match modeling approach.
• To investigate the method's robustness to errors and missing data in atomic coordinates.

Main Methods:

• The segment match modeling approach utilizes a database of known protein X-ray structures.
• Target structures are built from amino acid sequences by segmenting and matching database fragments based on sequence, conformation, and van der Waals' interactions.
• Multiple models are generated and averaged to improve accuracy, a novel concept for modeling tasks.

Main Results:

• The method successfully modeled eight test proteins (46-323 residues) with an average all-atom root-mean-square deviation of 1.26 Å (range 0.93–1.73 Å).
• Model accuracy is comparable to experimental refinement and variations in protein crystal packing.
• The approach is robust to C-alpha positional errors (up to 1 Å) and missing C-alpha atoms (up to 50% missing).

Conclusions:

• Segment match modeling provides an accurate, automated, and efficient solution for de novo protein structure prediction.
• The method's insensitivity to input data errors and its ability to improve results through model averaging have broad implications for structural modeling.
• This technique offers a powerful tool for generating complete atomic coordinates without human intervention, accelerating structural biology research.