Improved packing of protein side chains with parallel ant colonies

Area of Science:

• Computational Biology
• Biophysics
• Bioinformatics

Background:

• Accurate protein side-chain packing is vital for computational biology tasks like protein structure prediction and design.
• Existing methods often struggle with inaccurate energy functions, limiting the reliability of predicted protein structures.
• There is a need for improved computational methods to accurately model protein side-chain conformations.

Purpose of the Study:

• To develop and evaluate pacoPacker, a novel side-chain modeling method.
• To enhance the accuracy of protein side-chain conformation prediction using a parallel optimization strategy.
• To integrate diverse energy functions for more reliable side-chain packing.

Main Methods:

• pacoPacker employs a parallel ant colony optimization strategy with a shared pheromone matrix.
• The method combines multiple energy functions to determine optimal side-chain conformations.
• Rotamers were further optimized using rotamer minimization to improve library discreteness.

Main Results:

• pacoPacker achieved high accuracy in predicting side-chain angles, with 87.19% of X1 and 77.11% of X12 angles correctly predicted within 40°.
• Compared to state-of-the-art methods (CIS-RR, SCWRL4), pacoPacker showed superior performance in 51.5% of proteins tested (length ≤ 400 amino acids).
• The subrotamer strategy demonstrated a significant advantage in side-chain packing accuracy.

Conclusions:

• The parallel approach effectively integrates diverse search strategies and energy functions for protein side-chain packing.
• pacoPacker offers a competitive and accurate solution for side-chain conformation prediction.
• This framework facilitates the combination of various objective functions through parallel heuristic search algorithms.