Criteria for the comparison of electron microscopical 3D reconstructions of individual particles

Similarity criteria for electron microscopical 3D reconstructions can be defined in different ways. It is shown that the choice of the zero level in 3D reconstructions plays an important role, (a) for the discrimination of stained versus non-stained structural parts ("discrimination level"), and (b) for the physical definition of the signal ("zero potential level"). Of special interest are further structural features (artifacts) introduced by the reconstruction method chosen (single axis tilting in our papers). A crude estimate of the zero potential level is only possible if we neglect electron scattering from light atoms (concept of the "holey stain structure"). We calculate for the reconstructions of stained ribosome particles (W. Hoppe and H. Oettl and H. Tietz, J. Mol. Biol., in press) a signal-to-noise ratio of the order 3:1 (smaller by a factor of ca. 3 than the ratio calculated from electron noise). This explains why structural details can be recognized in single-particle reconstructions and why averages over only a few particles can already lead to consistent results. An analysis of the contrast decrease along the optical axis shows that this result is only true for strongly stained structural parts; averaging over large numbers of particles is required for the structural study of weakly stained parts. Possibilities of pattern recognition are discussed. The structural feature complementarity is explained by means of a model example; it can drastically reduce cross-correlation coefficients.