How does the folderol architecture compare to Cosm/foldingathome? The Pande group at Stanford seems to be doing a similar calculation, with the Cosm framework providing job assignment and security.
Cosm isn't GPL, and the proposed Cosm license seems to have some onerous restrictions.
The Folding@home architecture performs time-simulation of the way proteins fold. That means you only get to simulate a very small part of the folding of a protein before your work gets sent back to the central servers and you get another chunk to work on. Folderol tries a genetic algorithm aproach to folding, where it "breeds" possible protein shapes together depending upon how good they are. The best breed on your system after 1 million generations gets sent back to the central server.
With Folderol you start getting fully folded proteins back after a day (whether they be correct or not), where Folding@home you may have to spend many months of team work before you get a folded protein.
With Folderol you only get the end protein shape, where Folding@home are able to build a movie over time showing how the protein folded.
Hope this helps,
Malcolm
Malcolm.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
It would appear that there's a hole in the Folding@home security because there's an unconfirmed report that someone cracked it (See the Folding@home news group).
Lets just hope that no-one does it to Folderol!
Malcolm.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I'm a little bit unclear as to exactly what folding@home is actually doing. If they treat folding of a protein as a sequential process, then distributing that process across multiple machines is just going to make multiple copies of the same movie.
If instead, they're giving you pieces of a predestined simulation, then that implies that the results are predetermined (which means it's not prediction) and that they're just filling in the gaps. If this is the case, then it's not really even a simulation of folding but rather a form of infotainment.
I have been unable to find an unambiguous technical description of their work. They've made a pretty strong scientific claim in saying they've removed the barriers to simulating protein folding. Do you know of anywhere that they back this claim up?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello. This is my first post.
How does the folderol architecture compare to Cosm/foldingathome? The Pande group at Stanford seems to be doing a similar calculation, with the Cosm framework providing job assignment and security.
Cosm isn't GPL, and the proposed Cosm license seems to have some onerous restrictions.
Some relevant links are:
http://www.stanford.edu/group/pandegroup/Cosm/
http://www.mithral.com/projects/cosm/
Is folderol architected to have a Framework/calculation structure?
Any thoughts?
Chris
Hi Chris,
The Folding@home architecture performs time-simulation of the way proteins fold. That means you only get to simulate a very small part of the folding of a protein before your work gets sent back to the central servers and you get another chunk to work on. Folderol tries a genetic algorithm aproach to folding, where it "breeds" possible protein shapes together depending upon how good they are. The best breed on your system after 1 million generations gets sent back to the central server.
With Folderol you start getting fully folded proteins back after a day (whether they be correct or not), where Folding@home you may have to spend many months of team work before you get a folded protein.
With Folderol you only get the end protein shape, where Folding@home are able to build a movie over time showing how the protein folded.
Hope this helps,
Malcolm
Malcolm.
One other small point,
It would appear that there's a hole in the Folding@home security because there's an unconfirmed report that someone cracked it (See the Folding@home news group).
Lets just hope that no-one does it to Folderol!
Malcolm.
I'm a little bit unclear as to exactly what folding@home is actually doing. If they treat folding of a protein as a sequential process, then distributing that process across multiple machines is just going to make multiple copies of the same movie.
If instead, they're giving you pieces of a predestined simulation, then that implies that the results are predetermined (which means it's not prediction) and that they're just filling in the gaps. If this is the case, then it's not really even a simulation of folding but rather a form of infotainment.
I have been unable to find an unambiguous technical description of their work. They've made a pretty strong scientific claim in saying they've removed the barriers to simulating protein folding. Do you know of anywhere that they back this claim up?