Analysis and classification of volumes with missing data. This is a project  for analysing heterogeneity in three dimensions. While heterogeneity in the appearance of single particles can easily be observed in a micrograph (2D projection), in most cases, the cause of the difference in appearance can only be determined in  three dimensions. Exceptions can be differences based on present of missing ligands, when the general shape of the particle is not affected and the overall 3D structure of the molecule is known.

Apparent heterogeneity can be caused by conformational differences between the observed molecules or can be cause by different orientations of the particles on the specimen support. The techniques we are developing can determine the cause of the heterogeneity by analyzing multiple 3D structures obtained either by tomographic methods of by Random Conical Tilting.

The new methods for 3D pattern analysis of volumes with missing data, PPCAEM (Expectation Maximization Probabilistic Principle Component Analysis) can not only analyze these differences, but also estimates the missing data in each of the analyzed volumes. The reconstitution of each volume with the estimated missing data then makes it possible to visually examine the true differences between the reconstructions.

New image processing system, the “Environment for Modular Image Reconstruction Algorithms” (EMIRA). This is a new general and easily extendable image and data processing system, with the core written in Python 2.7. The syntax has been kept simple so that most users can learn using the system in less than one hour.  The system is distributed under the GPL license. For more details see “Resources”.
The system runs under LINUX (developed under CENTOS 5 and 6).

More coming soon