A multiresolution approach to the 3D reconstruction of a 50S ribosome from an EM-tilt series solving the alignment problem without gold particles [electron microscopy]

A multiresolution approach to 3D reconstruction from an electron microscope (EM) tilt series is presented. It is designed for reconstruction from a tilt series with a limited number of projections where the tilt angles are assumed to be known. Its main feature is that it solves the alignment problem very accurately from the image data without additional gold particles, including the determination of translational, orientational, and scaling parameters. The approach relies on a coarse-to-fine bootstrap control strategy based on the observation that with an appropriate matching procedure good alignment estimates can be obtained at a rather coarse level of resolution. This serves for a better reconstruction of the next finer level, from which a further refinement of the alignment estimate is obtained, etc. The actual reconstruction at each level of resolution is accomplished with a synthesis of the filtered back-projection method and iterative algebraic techniques, together with a set of convex constraints. The complete algorithm is applied to an EM tilt series of a 50S ribosome prepared in ice, consisting of 11 micrographs