Differential Visual Proteomics by Single-Particle Electron Microscopy

The levels of protein expression often constitute a marker for the physiological state of an organism. In addition to their abundance, the three-dimensional conformation of proteins play a crucial role in the determination of the cellular function. In order to understand cellular physiology, it is therefore crucial to determine both the structure and the abundance of the protein content of single-cells. Methods such as mass spectrometry and cryo-electron tomography have been employed in the past for structural proteomic analysis, however they cannot provide high-resolution structural information of the protein content of single-cells. This can be overcome by single-particle electron microscopy, which can provide high-resolution 3D reconstructions of isolated proteins. Recent advances in single-cell sample preparation for single-particle electron microscopy, have transformed this approach into a promising alternative to existing techniques for structural proteomics.
This thesis presents the development of a novel algorithm named “differential visual proteomics” (DVP) based on single-particle electron microscopy, for the quantitative and structural analysis of proteome samples from single-cells. The developed algorithm was packaged into a software named VisProt, which consists of a graphical user interface that brings together self-written image analysis and data managing algorithms and invokes various established single-particle processing programs. The DVP algorithm was used to investigate the structural changes of proteins between disturbed and undisturbed cells. The suggested algorithm identified differences in protein concentrations in simulated cryo-electron microscopy datasets and in experimental negative stain electron microscopy datasets of cell populations. In addition, by employing the in-house developed tool for the loss-less preparation of single-cell samples for single-particle electron microscopy (cryoWriter), structural differences in the proteome of single-cells prepared in different environments were also identified.
In addition to the development of the DVP algorithm and the VisProt interface, this thesis presents the development of image processing tools based on deep learning, which intend to increase the throughput of procedures for data collection and image analysis in electron microscopy.