6 ms·

Free Electron Laser Time

29 points by etiam 3y ago

perihelions 3y ago
- "The first X-ray structure of a protein took until 1958 (myoglobin), and it was an absolutely heroic effort with the technology of the time." Wow, that is remarkable [0]. I'm wondering how they calculated a 3-dimensional Fourier transform of x-ray data, on a 1958 machine [1]. Apparently they didn't trust their numerics and ran it twice? [0] https://sci-hub.se/10.1038/181662a0 https://sci-hub.se/10.1038/181662a0 [1] https://en.wikipedia.org/wiki/EDSAC https://en.wikipedia.org/wiki/EDSAC late edit: I'm awed by this data presentation format [2], derived—somehow—from a Fourier transform calculated by the mechanical computer described here [3] in 1945. [2] https://en.wikipedia.org/wiki/Dorothy_Hodgkin#/media/File:Model_of_the_Structure_of_Penicillin,_by_Dorothy_Hodgkin,_Oxford,_c.1945.jpeg https://en.wikipedia.org/wiki/Dorothy_Hodgkin#/media/File:Mo... [3] https://sci-hub.se/10.1088/0959-5309/54/5/307 https://sci-hub.se/10.1088/0959-5309/54/5/307
- thechao 3y ago
  The lab my wife worked in, 20+ years ago, still had a mahogany chest of "bones" to calculate Fourier transforms, by hand. Once you had phases, you'd plot density on transparent cels, and put them in a stack.
- COGlory 3y ago
  In this case, the x-ray data is already the constructive interference of the Fourier transform of the object shot by the X-ray beam. The raw diffraction data is actually just the intensities of the Fourier transform of whatever is providing the constructive interference (i.e. diffracting). What they actually had to calculate were the phases of the data, which is arguably much more challenging. Once those phases are calculated, you can do a reverse Fourier transform (also significant, also called a Fourier synthesis in that paper and elsewhere) and get your real-space map. This is an iterative process, so you don't usually just do it once, although I'm not sure what Kendrew, Perutz and Rossman did in the early days. https://en.wikipedia.org/wiki/Phase_problem https://en.wikipedia.org/wiki/Phase_problem Structural biology and cutting edge computing are inextricably linked, and probably always will be, which is actually how I fell into the field. My (not yet then) P.I. was trying to get Nvidia 3d vision working on Linux for refining the atomic structure into an X-ray map and had exhausted all the IT knowledge on campus before I stumbled in.
Something1234 3y ago
This would be enhanced with some pretty pictures for the layman to better understand what I'm looking at.
- COGlory 3y ago
  What are you wondering about? I'm a structural biologist and just got home for the day from running some electron diffraction samples. I could probably provide pretty pictures and accompanying explanations of anything up to the free-electron laser portion.

29 points by etiam 3y ago

- "The first X-ray structure of a protein took until 1958 (myoglobin), and it was an absolutely heroic effort with the technology of the time." Wow, that is remarkable [0]. I'm wondering how they calculated a 3-dimensional Fourier transform of x-ray data, on a 1958 machine [1]. Apparently they didn't trust their numerics and ran it twice? [0] https://sci-hub.se/10.1038/181662a0 https://sci-hub.se/10.1038/181662a0 [1] https://en.wikipedia.org/wiki/EDSAC https://en.wikipedia.org/wiki/EDSAC late edit: I'm awed by this data presentation format [2], derived—somehow—from a Fourier transform calculated by the mechanical computer described here [3] in 1945. [2] https://en.wikipedia.org/wiki/Dorothy_Hodgkin#/media/File:Model_of_the_Structure_of_Penicillin,_by_Dorothy_Hodgkin,_Oxford,_c.1945.jpeg https://en.wikipedia.org/wiki/Dorothy_Hodgkin#/media/File:Mo... [3] https://sci-hub.se/10.1088/0959-5309/54/5/307 https://sci-hub.se/10.1088/0959-5309/54/5/307

thechao 3y ago
The lab my wife worked in, 20+ years ago, still had a mahogany chest of "bones" to calculate Fourier transforms, by hand. Once you had phases, you'd plot density on transparent cels, and put them in a stack.
COGlory 3y ago
In this case, the x-ray data is already the constructive interference of the Fourier transform of the object shot by the X-ray beam. The raw diffraction data is actually just the intensities of the Fourier transform of whatever is providing the constructive interference (i.e. diffracting). What they actually had to calculate were the phases of the data, which is arguably much more challenging. Once those phases are calculated, you can do a reverse Fourier transform (also significant, also called a Fourier synthesis in that paper and elsewhere) and get your real-space map. This is an iterative process, so you don't usually just do it once, although I'm not sure what Kendrew, Perutz and Rossman did in the early days. https://en.wikipedia.org/wiki/Phase_problem https://en.wikipedia.org/wiki/Phase_problem Structural biology and cutting edge computing are inextricably linked, and probably always will be, which is actually how I fell into the field. My (not yet then) P.I. was trying to get Nvidia 3d vision working on Linux for refining the atomic structure into an X-ray map and had exhausted all the IT knowledge on campus before I stumbled in.

Something1234 3y ago

This would be enhanced with some pretty pictures for the layman to better understand what I'm looking at.

COGlory 3y ago
What are you wondering about? I'm a structural biologist and just got home for the day from running some electron diffraction samples. I could probably provide pretty pictures and accompanying explanations of anything up to the free-electron laser portion.