Computerized Classification of Surface Spikes in Three-Dimensional Electron Microscopic Reconstructions of Viruses

The purpose of this research is to develop computer techniques for improved three-dimensional (3D) reconstruction of viruses from electron microscopic images of them and for the subsequent improved classification of the surface spikes in the resulting reconstruction. The broader impact of such work is the following. Influenza is an infectious disease caused by rapidly-changing viruses that appear seasonally in the human population. New strains of influenza viruses appear every year, with the potential to cause a serious global pandemic. Two kinds of spikes – hemagglutinin (HA) and neuraminidase (NA) – decorate the surface of the virus particles and these proteins are primarily responsible for the antigenic changes observed in influenza viruses. Identification of the locations of the surface spikes of both kinds in a new strain of influenza virus can be of critical importance for the development of a vaccine that protects against such a virus. Two major categories of reconstruction techniques are transform methods such as weighted backprojection (WBP) and series expansion methods such as the algebraic reconstruction techniques (ART) and the simultaneous iterative reconstruction technique (SIRT). Series expansion methods aim at estimating the object to be reconstructed by a linear combination of some fixed basis functions and they typically estimate the coefficients in such an expansion by an iterative algorithm. The choice of the set of basis functions greatly influences the efficacy of the output of a series expansion method. It has been demonstrated that using spherically symmetric basis functions (blobs), instead of the more traditional voxels, results in reconstructions of superior quality. Our own research shows that, with the recommended data-processing steps performed on the projection images prior to reconstruction, ART (with its free parameters appropriately tuned) provides 3D reconstructions of viruses from tomographic tilt series that allow reliable quantification of the surface proteins and that the same is not achieved using WBP or SIRT, which are the methods that have been routinely applied by practicing electron microscopists. Image segmentation is the process of recognizing different objects in an image. Segmenting an object from a background is not a trivial task, especially when the image is corrupted by noise and/or shading. One concept that has been successfully used to achieve segmentation in such corrupted images is fuzzy connectedness. This technique assigns to each element in an image a grade of membership in an object. Classifications methods use set of relevant features to identify the objects of each class. To distinguish between HA and NA spikes in this research, discussions with biologists suggest that there may be a single feature that can be used reliably for the classification process. The result of the fuzzy connectedness technique we conducted to segment spikes from the background confirms the correctness of the biologists’ assumption. The single feature we used is the ratio of the width of the spike’s head to the width of its stem in 3D space; the ratio appears to be greater for NA than it is for HA. The proposed classifier is tested on different types of 3D reconstructions derived from simulated data. A statistical hypothesis testing based methodology allowed us to evaluate the relative suitability of reconstruction methods for the given classification task