With recent advances in structural genomics, there has been considerable interest in the rapid determination of protein structures in a high-throughput setting.  One bottleneck in structure determination arises in protein crystallography, and deals with interpretation of the electron-density map, the three-dimensional "picture" of the protein that crystallography produces.   This talk describes a novel probabilistic approach to the problem of poor-resolution density-map interpretation.  We formulate the problem of determining the position of each amino-acid residue as inference in a pairwise Markov field.  Belief propagation is used to infer the marginal probabilities of each amino acid's position given the density map.  A novel message approximation scheme allows inference on proteins with several hundred residues.  The approach is also amenable to production of multiple protein models that explain the observed density. Our method accurately interprets 3-4A density maps, outperforming existing methods, and further extending the resolution of density maps which can be automatically interpreted.