Scotty : lattice coincidences for macromolecular crystallographic phasing

Abstract

When a macromolecular crystal lattice targeted for study is closely related to a previously studied crystal lattice, phasing the target crystal by difference (fast-)Fourier transform (DFFT) methods is preferable to performing molecular replacement. The application Scotty , within the Phasertng codebase, is software for the identification of coincident lattices and downstream processing. All crystallographic PDB entries are organized into a scikit-learn `BallTree' index under the Niggli cell distance metric `NCDist'. Nearest neighbours under the metric are progressed to test structure-factor intensity correlation. The structure from the lattice with the highest correlation with the target is used to phase the target lattice, and the coordinates are taken forward to coordinate refinement using a wide convergence radius protocol. The method can identify lattice coincidences accounting for very significant non-isomorphism. The nearest-neighbour search is space-group agnostic, so that coincident lattices are identified even when the nominal space groups are different, the symmetry of one lattice being described as a subgroup of the other or a higher metric symmetry.