Quantum refinement is an improvement upon the molecular mechanics (MM)-based crystallographic refinement. In the latter, X-ray data are supplemented with additional chemical information through MM force fields, whereas quantum refinement describes crucial regions of interest in the macromolecule by quantum mechanics (QM) instead of MM. In this paper, we report the implementation of quantum refinement in the ChemShell QM/MM framework and its application in an investigation of the chromophore structure of the red fluorescent protein DsRed.M1. Both mechanical and electrostatic QM/MM embedding schemes are implemented and tested. In the quantum refinement of DsRed.M1, the anionic red acylimine chromophore adopts a nearly orthogonal arrangement (rather than a cis or trans form), and the bond lengths in the acylimine moiety are more consistent with a phenolate (rather than a quinoid) structure. These findings are in contrast to the structure deduced from a standard crystallographic refinement (PDB: 2VAD), but in agreement with our earlier results from a purely theoretical QM/MM study. On the other hand, the quantum refinement of the anionic acylimine form of DsRed.M1 yields a hydrogen bonding network around the chromophore, especially with regard to the arrangement of the water molecules and the Glu148 residue, that is closer to the 2VAD structure than to the previously optimized QM/MM structure. In our earlier study the initial classical molecular dynamics (MD) simulations during QM/MM setup apparently exaggerated the mobility of the water molecules around the chromophore. On the basis of the present results, it seems likely that the Glu148 residue is protonated in the DsRed.M1 protein. The calculation of electronic excitation energies allows for further assessment of the proposed structures, especially in the chromophore region. Using a combination of density functional theory and multireference configuration interaction (DFT/MRCI), we find excellent agreement between experiment and theory only for the structures obtained from quantum refinement and from QM/MM optimization, but not for the 2VAD structure. The present case study on DsRed.M1 thus demonstrates the merits of combining reliable theoretical and experimental information in the determination of protein structures.