GalaxyRefine is a web server for protein structure refinement that improves the accuracy of model structures generated by template-based methods. It uses side-chain repacking and molecular dynamics simulations to refine models, achieving consistent improvements in local and global structure quality. The method was tested in the CASP10 experiment and showed the best performance in enhancing local structure quality. GalaxyRefine can improve both global and local structure accuracy, as well as physical correctness, in a majority of targets when measured by GDT-HA, GDC-SC, and MolProbity scores. It also generates additional models through relaxation simulations with larger perturbations, leading to more significant changes from the initial model. The method has been shown to improve models generated by state-of-the-art structure prediction servers such as I-TASSER and ROSETTA. GalaxyRefine uses a combination of physics-based energy functions and database-derived terms to refine structures, with a focus on side-chain conformation and overall structure relaxation. The server is available online and provides five refined models, along with structural information in terms of GDT-HA, RMSD, and MolProbity scores. It is particularly effective in improving local structure quality and is recommended for refining model structures obtained from available structure prediction methods.GalaxyRefine is a web server for protein structure refinement that improves the accuracy of model structures generated by template-based methods. It uses side-chain repacking and molecular dynamics simulations to refine models, achieving consistent improvements in local and global structure quality. The method was tested in the CASP10 experiment and showed the best performance in enhancing local structure quality. GalaxyRefine can improve both global and local structure accuracy, as well as physical correctness, in a majority of targets when measured by GDT-HA, GDC-SC, and MolProbity scores. It also generates additional models through relaxation simulations with larger perturbations, leading to more significant changes from the initial model. The method has been shown to improve models generated by state-of-the-art structure prediction servers such as I-TASSER and ROSETTA. GalaxyRefine uses a combination of physics-based energy functions and database-derived terms to refine structures, with a focus on side-chain conformation and overall structure relaxation. The server is available online and provides five refined models, along with structural information in terms of GDT-HA, RMSD, and MolProbity scores. It is particularly effective in improving local structure quality and is recommended for refining model structures obtained from available structure prediction methods.