The limitation of the simple thresholding method can be overcome by use of knowledge from other experiments. Spahn et al. (2000) introduced a method of RNA/protein separation based on thresholding, availability of generic histogram information, and region growing.
While a cryo-EM density map, at resolutions below 1/10 A_1 (i.e. worse than 10 A), of a structure composed of RNA and protein has a broad density histogram without trace of a separation of the underlying components, the histograms of regions solely containing RNA or protein (signature histograms) look distinctly different. Even though there is substantial overlap, there exists a lower-density range that must definitely originate from protein, and a higher-density range that must originate from RNA. Thus, knowledge of those signature histograms makes it possible to identify, by thresholding, those "core regions'' in the map that are protein or RNA with absolute certainty. Starting from the RNA core, Spahn et al. (2000) took advantage of the fact that, in each ribosomal subunit, RNA forms a single contiguous structure: 16S within the small subunit; 23S and—tightly associated with it—5S RNA. They used an image-processing method called region growing. First, the voxel with highest density is used as initial seed. The density threshold is then stepwise reduced. In each step, only those voxels are added in the growth process that are (i) above the threshold and (ii) directly connected with the already existing group of voxels. This process is stopped when the known volume of the RNA is reached.
As we have seen, the method uses signature histograms, derived from an analysis of any two regions of the density map exemplary for the two components. In the case shown, the knowledge of where to look for densities stemming from RNA and protein came from a published 5.5-A X-ray map in which the RNA and some of the published proteins were delineated in the small ribosomal subunit.
Was this article helpful?