Probabilistic Principal Component Analysis using Expectation Maximization (PPCA-EM) for Analyzing 3D Volumes with Missing Data

L Yu, RR Snapp, T Ruiz, M Radermacher
2010 Microscopy and Microanalysis  
There are different techniques to obtain 3D structures of macromolecular assemblies from electron micrographs: electron tomography with either single-axis, dual-axes or conical tilting [1,2], random conical reconstruction [3, 4] , angular reconstitution [5] , and orthogonal tilting reconstruction [6] . The techniques including tilting are the most suitable for investigating structures of heterogeneous samples. However, most tilting techniques leave out areas of missing data in the 3D
more » ... ions because the tilt angle range is limited in the electron microscope. If the volumes are combined correctly, a volume with no missing data or much less missing data can be achieved. This requires correct identification of the volumes representing the same structure. However, missing data present an obstacle to analyzing variations between 3D volumes since they cause artifacts in the 3D reconstructions. Especially when the data is missing in different orientations, artifacts can be easily misinterpreted as structural differences. 836
doi:10.1017/s143192761005734x fatcat:c5i4q7cswzdqjjty7lxr3go7iq