PhD candidate Ilja Gubins started working on a research project on the classification of bio-molecular complexes using cryo-electron tomography (cryo-ET). Innovative computational methods are an integral part of the research approach.
As Ilja says: “Machine learning has revolutionized every industry, starting from automation of manual labour to new unexpected discoveries. Applying it to structural biology has already proven to be very fruitful, and we believe that deep learning in particular can bring noticeable improvements to cry-electron tomography.”
Cryo-electron tomography (cryo-ET) provides 3D images of macromolecular complexes in their physiological environment at molecular resolution. The main research question is how to improve the successive stages in cryo-ET: detection, localization, and classification of macromolecular complexes.
Innovation in computational methods remains key to derive biological information from the extremely low signal/noise ratio tomograms. The project aims to develop and apply new methods and techniques using pattern recognition and machine learning, such as shape classification and deep learning.
Since last March, PhD candidate Ilja Gubins has been working on this collaboration project supervised by professors Friedrich Förster (Chemistry) and Remco Veltkamp (Information and Computing Sciences).