Real-space refinement in PHENIX is a method for refining atomic models against maps, including cryo-EM and crystallographic data. The program uses a simplified refinement target function that allows for fast calculations and efficient identification of optimal data-restraint weights. This method incorporates additional information such as secondary-structure and rotamer-specific restraints, as well as internal molecular symmetry. Refinement of 385 cryo-EM-derived models at resolutions of 6 Å or better showed significant improvements in model quality and fit to target maps. The refinement process involves defining a model, a target function that links the model parameters to experimental data, and an optimization method to improve the model agreement with the data. In PHENIX, gradient methods are used for this purpose. Real-space refinement (RSR) is particularly useful for cryo-EM as it allows for local refinement and is less affected by model incompleteness compared to reciprocal-space refinement (FSR). RSR is also used in crystallographic refinement. The refinement target function in RSR is based on the fit of the model to a target map, while in FSR it is based on diffraction intensities or structure factors. The target function in RSR is calculated using a sum over grid points, while in FSR it is calculated using a sum over reflections. The choice of target function is crucial for the success of refinement, as it determines the efficiency of convergence to an improved model. The refinement process includes various tasks such as gradient-driven minimization, simulated annealing, morphing, and residue side-chain optimizations. The program also determines internal symmetry and applies restraints or constraints between related molecules. The refinement results are output as a file in PDB or mmCIF format. The refinement target function in RSR is calculated using a sum over atomic centers, while in FSR it is calculated using a sum over reflections. The target function in RSR is less accurate but much faster to calculate compared to FSR. The refinement process also includes restraints such as covalent bond lengths and angles, dihedral angles, planarity, and chirality. Additional restraints may include similarity of related copies, secondary structure restraints, and reference-model restraints. The relative weight of restraints is chosen to balance the model fit to the map with reasonable deviations from ideal covalent bond lengths and angles. The weight is determined by systematically trying a range of plausible values and performing a short refinement for each trial value. The optimal weight is defined as the one that results in a model with reasonable bond and angle root-mean-square deviations and the best model-to-map fit. The refinement process also includes evaluation of model quality and model-to-data fit. In crystallography, model-to-data fit is quantified by crystallographic R and R_free factors. In cryo-EM, model and data validation is performed by comparing complex Fourier coefficients in resolution shells. The refinement process also includes evaluationReal-space refinement in PHENIX is a method for refining atomic models against maps, including cryo-EM and crystallographic data. The program uses a simplified refinement target function that allows for fast calculations and efficient identification of optimal data-restraint weights. This method incorporates additional information such as secondary-structure and rotamer-specific restraints, as well as internal molecular symmetry. Refinement of 385 cryo-EM-derived models at resolutions of 6 Å or better showed significant improvements in model quality and fit to target maps. The refinement process involves defining a model, a target function that links the model parameters to experimental data, and an optimization method to improve the model agreement with the data. In PHENIX, gradient methods are used for this purpose. Real-space refinement (RSR) is particularly useful for cryo-EM as it allows for local refinement and is less affected by model incompleteness compared to reciprocal-space refinement (FSR). RSR is also used in crystallographic refinement. The refinement target function in RSR is based on the fit of the model to a target map, while in FSR it is based on diffraction intensities or structure factors. The target function in RSR is calculated using a sum over grid points, while in FSR it is calculated using a sum over reflections. The choice of target function is crucial for the success of refinement, as it determines the efficiency of convergence to an improved model. The refinement process includes various tasks such as gradient-driven minimization, simulated annealing, morphing, and residue side-chain optimizations. The program also determines internal symmetry and applies restraints or constraints between related molecules. The refinement results are output as a file in PDB or mmCIF format. The refinement target function in RSR is calculated using a sum over atomic centers, while in FSR it is calculated using a sum over reflections. The target function in RSR is less accurate but much faster to calculate compared to FSR. The refinement process also includes restraints such as covalent bond lengths and angles, dihedral angles, planarity, and chirality. Additional restraints may include similarity of related copies, secondary structure restraints, and reference-model restraints. The relative weight of restraints is chosen to balance the model fit to the map with reasonable deviations from ideal covalent bond lengths and angles. The weight is determined by systematically trying a range of plausible values and performing a short refinement for each trial value. The optimal weight is defined as the one that results in a model with reasonable bond and angle root-mean-square deviations and the best model-to-map fit. The refinement process also includes evaluation of model quality and model-to-data fit. In crystallography, model-to-data fit is quantified by crystallographic R and R_free factors. In cryo-EM, model and data validation is performed by comparing complex Fourier coefficients in resolution shells. The refinement process also includes evaluation