Maximum likelihood single particle reconstruction of macromolecules from electron microscope

Abstract

Recent advance in cryo-electron microscopy allows structural determination of macromolecules without forming crystals. There has been extensive researches of reconstruction approach for electron microscope image. The widely used method for icosahedral virus reconstruction is the model based approach. However, the drawback is that similar low resolution model should be prepared for initial reference. Contrarily, common-line technique is model-free method based on projection-slice theorem. However, the performance is not good in noise image. In case of helical macromolecules, a classical method for structure determination is the Fourier-Bessel transform approach using diffraction pattern. This approach often fails in some instances such as flexible and disordered specimens. In order to overcome these drawbacks, a single particle reconstruction approach has been recently proposed. However, this approach also requires a good initial guess to guarantee the convergence. Therefore, we propose an accurate ab initio maximum likelihood reconstruction algorithm by exploiting the symmetry and sparsity. More specifically, since the macromolecule structure assumes only sparse supports in real space and the symmetry provides several symmetric views from a single micrograph, a reasonably quality 3-D reconstruction can be obtained from the limited views using recently signal processing technique called compressed sensing. Furthermore, the precise view angles and symmetry parameters can be estimated using maximum likelihood method. With the estimated parameters, the multiple projections can be combined for the optimal quality reconstruction. In chapter 1, we first describe the icosahedral virus reconstruction, then in chapter 2, we discuss about helical macromolecules reconstruction.