Model-based, 3D reconstruction techniques depend on reliable starting models. We present an extension of the random-model method (RMM) that allows the ab initio generation of suitable starting models directly from un-averaged, experimental images of asymmetric or symmetric particles. Therefore, the asymmetric RMM can also be used to determine point-group symmetry. The procedure is facilitated by the use of (a) variable angular step-sizes during iterative origin and orientation searches, (b) high numbers of particle images, and (c) highly defocused images. The method is inhibited by mixed-handedness orientation assignments and by particles with inconspicuous features. For symmetric particles, symmetric RMMs can overcome these deficiencies.

Copyright 2010 Elsevier Inc. All rights reserved.