Reconstruction Puzzles

What are reconstruction puzzles? These are puzzles in which there's a known protein structure that's been solved by crystallography of cryo-EM before, but we have reason to believe that it could be improved. The factors we look at to decide whether this is the case are the statistics pertaining to quality of geometry and steric clashes as well as statistics on the fit of the protein to the electron density. Some of the reconstruction puzzles are older, when tools for solving structures were not as good, but some are newer. While the PDB has quality control metrics and flags structures for authors to work on, ultimately the system is designed to be permissive in what structures are deposited. As a result, Foldit Players can contribute by taking some of these structures that have issues and bringing them up to snuff.

Why is this important? There's a couple reasons, but it comes down to every structure in the PDB mattering. Sometimes, this has to do with our ability to use it as a database for things like machine learning, in which errors can cause problems. However, I like to think about it from the perspective that each structure tells us something that's unique in science. Science is meant to build off the discoveries of others, and this is the reason that the PDB is a public database; scientists download these structures to look for clues as to how the biology works, and they use those structures to design and interpret their own experiments. Mistakes in structures can then propagate to incorrect assumptions and conclusions down the road. Over time, these mistakes are found out, but potentially at the loss of a lot of time, money, and energy. It's better if we can nip it in the bud by improving the structures in the PDB themselves.

( Posted by  agcohn821 75 554  |  Mon, 05/30/2022 - 19:22  |  2 comments )
alcor29's picture
User offline. Last seen 1 day 10 hours ago. Offline
Joined: 11/16/2012
What about AF?

Are we better at this than Alpha-fold/DeepMind?

horowsah's picture
User offline. Last seen 2 weeks 3 days ago. Offline
Joined: 03/17/2016
Groups: DU Proteins S20
Short answer- alphafold isn't

Short answer- alphafold isn't really built for these particular problems, but looking at the best computational methods available, so far it seems like Foldit players definitely have something the computers don't as far as I an tell.

User login
Download links:
  Windows    OSX    Linux  
(10.12 or later)

Are you new to Foldit? Click here.

Are you a student? Click here.

Are you an educator? Click here.
Social Media

Only search
Other Games: Mozak
Recommend Foldit
Top New Users

Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Amazon, Microsoft, Adobe, Boehringer Ingelheim, RosettaCommons