Skip to main content

iMOL: Interfacing Molecular and Computer Science

The ‘Interfacing image analysis and molecular life science’ (iMOL) Research Training Group (RTG) is committed to educating PhD students in the theory and practice of advanced microscopy methods and cutting-edge image analysis. 

  • Microscopy is instrumental to our understanding of the basic principles of life at multiple spatio-temporal scales and resolutions. 
  • Machine learning techniques are used to mine the information hidden in the images and recover the signal. 

Recent technological developments in light and electron microscopy have broadened and deepened our understanding of biological processes. This underlines the importance of and the need for dedicated educational training programs on image analysis methods that support the understanding and interpretation of microscope-derived data.

Central to Imol RTG are hypothesis-driven life science applications that challenge textbook knowledge at the molecular scale. We will emphasize on neuroscience and analyze cell-cell junctions between epithelial cells that affect the blood-brain barrier. We will also analyze how vascular and neuronal structures intermingle along the developmental process in order to generate functional organs. In the interspace of those we will also look at neurodegeneration and the effect of the formation of extracellular amyloid plaques.

In the iMOL RTG we will optimize the design, construction and automated usage of cutting-edge microscopy techniques using interdisciplinary efforts. By working with cryo-electron tomography (cryo-ET), super-resolution and light sheet-based fluorescence microscopy (e.g. SIM, PALM/dSTORM, SPIM/DSLM), Raman microscopy and multi-photon microscopy, our efforts will span the complete range of scales and resolutions, i.e. from tissues to cells and ultimately to individual macromolecules. To interpret the resulting images, we will investigate and further develop state- of-the-art computational methods and implement them on dedicated processing units (e.g. GPUs) and supercomputers for semi-autonomous image analyses and interpretations.

Figure legend: Image reconstruction techniques and machine learning. Current algorithms operate under the assumption that the sample (f) has a rigid-body movement (Castano et al. 2007). However, this is not valid. Three-dimensional image reconstruction algorithms based on simultaneous algebraic reconstruction techniques minimize the difference between the recorded micrograph (p{i}) and the projection of the reconstruction (w{in} f{n}) in the direction of the acquisition (Kunz et al. 2016). This allows for computationally removing the background in the recorded micrographs and highlighting individual macromolecules in the cellular context in an otherwise heavily superimposed and unrecognizable picture. In this way, individual macromolecules can be treated separately. This promises that local sample movements and imperfection can be compensated for.

Research | Frangakis-Group | Institute of Biophysics