Electron crystallographers who have been working on determination of protein structure have set a goal of obtaining image information to a resolution of about 3.5 Å, from specimens tilted up to 60 degrees. This information would allow the construction of a three-dimensional density map within which the path of the peptide chain could be followed and locations of side chains defined. The recent determination of an atomic model of the membrane protein bacteriorhodopsin (bR) from EM data (1) which was not as complete as we would like, used a good deal of other biochemical and biophysical data to constrain the model. In cases where this type of information is not as extensive as with bR, isotropic high-resolution data would be required. Significant advances in several different areas have brought us tantalizingly close to reaching our goal, but there are still improvements to be made.The essential limitations in obtaining high resolution data from proteins arise from the radiation sensitivity of the specimen, which severely limits the electron exposure that can be used in recording an image and thus limits the signal-to-noise ratio (SNR). Increasing both the electron dose, which is possible with cold specimens, and the area processed, which required implementation of significant computer software, have each given about a factor of three improvement in SNR. Still, with conventional imaging, a study by Henderson and Glaeser (2) revealed that the best images contained only a small fraction of the signal that would be present in a perfect image. Factors such as the envelope of the contrast transfer function and the modulation transfer function of the photographic film account for some loss of contrast, but the factor causing the most loss was found to be beam-induced specimen motion. This motion results from the stress which is produced by changes in bond structure during the course of radiation damage.
Ab initiophase determination of proteins with high-resolution data by direct methods