Protein structure refinement is a crucial step for more accurate protein structures given initial models. Based on our previous 1.0 version of Artificial Intelligence-based Refinement (AIR), the new version of AIR 2.0 employs a decomposition strategy that decomposes a multi-objective optimization problem into a number of subproblems and optimizes them simultaneously using particle swarm optimization algorithm. The solutions yielded by AIR2.0 show better convergence and diversity than the 1.0 version, which are expected to be useful for future structure refinement.
Figure 1. Flowchart of AIR 2.0 algorithm.
Figure 2. Two examples of refinement for the proteins in CASP14. The starting model is shown in wheat, refined model in magenta and native structure in green while the main improved regions are highlighted by red dotted circle. For more information about the test targets, please refer to (a) R1029 of CASP14 and (b) R1042v2 of CASP14.
1. Cheng-Peng Zhou, Di Wang, Xiaoyong Pan, and Hong-Bin Shen. Protein structure refinement using multi-objective particle swarm optimization with decomposition strategy. Submitted. 2021.
2. Wang D, Geng L, Zhao Y J, et al. Artificial intelligence-based multi-objective optimization protocol for protein structure refinement[J]. Bioinformatics, 2020, 36(2): 437-448.