Hekstra Lab
80 posts
Hekstra Lab
@HekstraLab
Research group of Prof. Doeke Hekstra | Harvard Department of Molecular & Cellular Biology
Cambridge, MA Beigetreten Ocak 2021
164 Folgt493 Follower
Can AlphaFold refine structures for X-ray & Cryo-EM/ET data? Now it can! Great collab with @MoAlQuraishi and Randy Read, led by Alisia @FadiniAli and @mhli41. Preprint: biorxiv.org/content/10.110… Summary 🧵below. Code to follow...
Alisia Fadini@FadiniAli
Structural biology is in an era of dynamics & assemblies but turning raw experimental data into atomic models at scale remains challenging. @mhli41 and I present ROCKET🚀: an AlphaFold augmentation that integrates crystallographic and cryoEM/ET data with room for more! 1/14.
English
@the_kittylord David, it's not *that* bad. There is just more progress to be made!
English
Dang if this approximation from crystal structures are that far off then these companies training AI on them are in for a rude awakening!
Hekstra Lab@HekstraLab
Crystal structures are *not* God-given truth. They approximate, w/ flaws & errors, X-ray diffraction data. AlphaFold etc. have been trained on structures, not data. SFCalculator now differentiably connects structures to diffraction data. What does this enable? 🧵 1/4
English
@YazdizadehAli We'll (re)tweet about that soon. See biorxiv.org/content/10.110…!
English
Crystal structures are *not* God-given truth. They approximate, w/ flaws & errors, X-ray diffraction data. AlphaFold etc. have been trained on structures, not data. SFCalculator now differentiably connects structures to diffraction data. What does this enable? 🧵 1/4
English
@YazdizadehAli @HekstraLab For atomic model building with maps, yes. But not yet to work on reconstructions from single-particle images.
Check out this work where using SFCalculator, we guide Alphafold predictions with crystallography and cryoEM data (half maps): biorxiv.org/content/10.110…
English
@HekstraLab Does anyone have the complete PDF? I don't have access to it 😭
English
Now in Cell: K+ ion channels enable electric currents consisting of potassium ions to generate action potentials in heart, brain, and muscle. For the first time, we can now see these currents atom-by-atom. cell.com/cell/abstract/… a thread🧵 1/8
English
Led by Rick Hewitt and Kevin Dalton, the result is Laue-DIALS, hosted by @rs-station.bsky.social at rs-station.github.io/laue-dials/. Wanna try this out but need advice? Please get in touch. Further upgrades coming soon! Read more at pubs.aip.org/aca/sdy/articl… 2/2
English
X-ray diffraction can access protein dynamics. Why so few datasets? One reason? Lack of powerful pulsed X-ray sources. XFELs are limited. Synchrotrons throw away >99% of photons. Working with Aaron Bewster (DIALS/LBNL) and BioCARS (biocars.uchicago.edu) to change this! 1/2
English
Enabling better crystal structures & better structure prediction. Can’t wait to see where folks take that! SFCalculator is implemented in PyTorch, TensorFlow, and JAX. Open Source. See biorxiv.org/content/10.110… for more info & github.com/Hekstra-Lab/SF… to get started. 4/4
English
With SFCalculator you can also *train* structure prediction / generative models directly on X-ray data. As @JHoltonMADSci has shown, the data are much more accurate than the models. 3/4
English
Fruit of a close collaboration with Rama Ranganathan's group, where Doeke, Ian (@kiwhite), and Rama first developed EFX, and BoRam worked tirelessly to grow crystals and analyze data. New hardware and software from @HekstraLab were critical for success! 8/8
English
In evolution, members of a protein family are often variations on a theme. The dynamic states resemble stable states of other channels: The sequence of steps of ion permeation is conserved from channel to channel with variations in relative energy. 7/8
English