Evan Kania, “Extracting 3D Structures of Biological Molecules from X-Ray Data”
Mentor: Ahmad Hosseinizedah
Poster # 176
Viruses and biomolecules are so small that they are unseen by the naked eye. Under a microscope, it may still be hard to see and or differentiate any details. Using an X-ray free-electron laser allows us to “see” biomolecules by methods such as X-ray diffraction. X-ray diffraction allows us to obtain information about the structure of biomolecules. X-ray wavelengths are short enough to penetrate between atomic nuclei allowing for interference and penetration. This interference results in a 2D pattern which can be thought of as shining light on an object and retrieving an image of its shadow. These images provide part of the information of the 3D structure; however, taking many images from all directions will provide almost complete information of its 3D structure. These images called snapshots correlate to a data matrix with values based on intensity. Snapshots are plotted in a high-dimensional space and reduced down to 13 dimensions via diffusion mapping. With knowledge of the orientation of each snapshot, a diffraction volume can be constructed. The diffraction volume obtained is constructed from 2d snapshots. An algorithm called interpolation is used, where the space between the planes is filled by generated data based on neighboring points. With a higher number of snapshots, the filled-in space during interpolation is more accurate. Reconstruction of the 3D object is achieved by machine learning techniques based on the symmetry of the object known as phasing. Obtaining the knowledge of biomolecule structures is important for understanding its biological processes, drug research, and contribution to scientific fields.