Efficient Sampling of Protein Loop Regions Using Conformational Hashing Complemented with Random Coordinate Descent.

Area of Science:

• Computational structural biology
• Biophysics
• Protein structure prediction

Background:

• Protein loops exhibit conformational heterogeneity, leading to structural ambiguity in experimental data.
• Generating diverse, realistic loop conformations is computationally intensive.
• Accurate protein structure determination requires modeling these flexible regions.

Purpose of the Study:

• To develop a novel algorithm for de novo loop construction in proteins.
• To improve the efficiency and accuracy of sampling protein loop conformations.
• To address the challenge of conformational heterogeneity in loop regions.

Main Methods:

• Developed the Hash/RCD algorithm, combining knowledge-based conformational hashing with random coordinate descent (RCD).
• Utilized templates to guide the sampling process.
• Tested the algorithm on a benchmark set of 195 loops across 29 proteins.

Main Results:

• Achieved a 100% closure rate on the benchmark loop set.
• Reduced loop sampling time significantly, from over 400 ms to 141 ms.
• Maintained the accuracy of state-of-the-art methods while improving speed.

Conclusions:

• The integration of conformational hashing and RCD overcomes limitations of individual methods.
• Hash/RCD facilitates rapid sampling of numerous native-like loop conformations.
• This method can aid in identifying protein structures using experimental data and scoring functions.