3D Electron Cryomicroscopy
A group of researchers, led by Holger Stark (Director of the Max Planck Institute (MPI) for Biophysical Chemistry, has made the cell's protein factory - the ribosome - more visible than ever before. The scientists broke the resolution barrier of 0.3 nm for the first time and were able to directly observe the ribosome at an unprecedented level of detail which has important implications to our current understanding of how antibiotics act on the ribosome.
The aim of this group, headed by Professor Holger Stark, the Director of the Institute, is to determine three-dimensional structures of macromolecules with the single particle electron Cryomicroscopy technique (cryo-EM) for which Jacques Dubochet, Richard Henderson and Joachim Frank were awarded with the Nobel Prize for chemistry in 2017. Macromolecular complexes represent the biological “nanomachines” in our cells that contribute to the most important and essential biological functions. It is important to study their structure and dynamic behavior at the highest possible level to finally understand the molecular basis of potential malfunctions that can lead to severe diseases.
The current success of the cryo-EM technology has already contributed a wealth of information on the structural organization and the architecture of large macromolecular assemblies in the last six years and the number of solved structures is exponentially growing. Cryo-EM is a powerful technique to generate three-dimensional reconstructions of any object imaged in a transmission electron microscope while the sample is kept at either liquid helium or liquid nitrogen temperature. Usually hundreds of thousand molecular electron microscopic images are needed to calculate a three-dimensional structure at high resolution. The images are obtained at low temperature in high-end transmission electron microscopes operated at low temperatures. To obtain 3D structure information from these images, advanced image processing software plays an essential role and the significant amount of data processing involved in this structure determining process requires enormous computational power. The idea of calculating three-dimensional density maps from two-dimensional projection images is similar to computer tomography in medicine and can simply be understood as the computational reversal of the imaging process in the transmission electron microscope.
To minimize the beam damage that electrons can cause to molecules while passing through the sample, images are usually recorded at very low electron dose which creates a significant amount of noise in the images. This noise is the reason why so many images are required to calculate high-resolution structures and why extensive computer aided image processing is needed to finally calculate a three-dimensional structure of a macromolecule. Generally, the availability of GPU computing to the field of cryo-EM based structure determination of macromolecules had therefore an enormous impact on the way structures can be obtained today. Instead of long computational runs on huge CPU clusters it became sometimes possible to calculate high-resolution structures within hours on local computers equipped with several GPUs such as the DGX.
The previous computing infrastructure was considerably extended by the use of the new DGX-1 AI-/HPC compute server. The main application lies in image recognition and image processing) and has thus considerably accelerated the overall work processes in cryo-EM based structure determination of biological macromolecules.