Discussions about release log entries

Message boards : Number crunching : Discussions about release log entries

To post messages, you must log in.

AuthorMessage
Profile Cureseekers~Kristof

Send message
Joined: 5 Nov 05
Posts: 80
Credit: 689,603
RAC: 0
Message 19031 - Posted: 21 Jun 2006, 5:59:00 UTC

Version 5.24 is up:
1. The symbol store that we put in 5.22 was not properly activated, so we're giving it another shot.
2. We can now use prior predictions for which parts of the chain are buried or exposed to guide the Rosetta search.
3. We can efficiently assemble predefined domains of the protein chain into a whole structure.

Rhiju gave this info in the Rosetta Application Version Release Log, but it doesn't sound very clear to me. Can someone give a little more info about the changes? Maybe David in a journal?
Member of Dutch Power Cows
ID: 19031 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Rhiju
Volunteer moderator

Send message
Joined: 8 Jan 06
Posts: 223
Credit: 3,546
RAC: 0
Message 19540 - Posted: 30 Jun 2006, 7:50:52 UTC - in response to Message 19031.  

Sorry, time has been a little short for us lately because of CASP -- we're still monitoring feedback. Really quickly:

The "symbol store" (a.k.a. the PDB file) is a file that translates the gobbledygook we get from the occasional Rosetta@home crash into plain English, so that we can track down where and why in the code an error occurs. Its super useful, but has been a little tricky to set up.

The "prior predictions" of burial or exposure come from other neural networks that are trained to look at protein sequences and guess which parts are on the inside and outside of the fold. This was just a quick first implementation; you can expect more progress in September.

As for domains... if a protein is more than ~150 residues long its just too difficult to sample every possible conformation extensively. Luckily most real proteins of large sizes (>200 residues) subdivide into domains of 70-150 residues that appear to fold into independent pieces that are not too intertwined with each other. So before we submit jobs to Rosetta@home, we often cut up a sequence into putative domains(based on algorithms like Ginzu and RosettaDom, developed by members of the Baker lab). Then we try to fold those pieces of the protein individually, and we can now use Rosetta@home (5.24 onwards) to take those individual sub-folds and pack them together into the final structure for the whole protein chain.

Thanks for asking the question -- I'll post this clarification in the version log, too.

Version 5.24 is up:
1. The symbol store that we put in 5.22 was not properly activated, so we're giving it another shot.
2. We can now use prior predictions for which parts of the chain are buried or exposed to guide the Rosetta search.
3. We can efficiently assemble predefined domains of the protein chain into a whole structure.

Rhiju gave this info in the Rosetta Application Version Release Log, but it doesn't sound very clear to me. Can someone give a little more info about the changes? Maybe David in a journal?


ID: 19540 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Rhiju
Volunteer moderator

Send message
Joined: 8 Jan 06
Posts: 223
Credit: 3,546
RAC: 0
Message 19541 - Posted: 30 Jun 2006, 7:52:24 UTC - in response to Message 19021.  
Last modified: 30 Jun 2006, 7:52:49 UTC

There was a question in another thread about these mysterious comments on 5.24, so here's some clarification (I hope):

The "symbol store" (a.k.a. the PDB file) is a file that translates the gobbledygook we get from the occasional Rosetta@home crash into plain English, so that we can track down where and why in the code an error occurs. It is super useful, but has been a little tricky to set up.

The "prior predictions" of burial or exposure come from other neural networks that are trained to look at protein sequences and guess which parts are on the inside and outside of the fold. This was just a quick first implementation; you can expect more progress in September.

As for "domains"... if a protein is more than ~150 residues long its just too difficult to sample every possible conformation extensively. Luckily most real proteins of large sizes (>200 residues) subdivide into domains of 70-150 residues that appear to fold into independent pieces that are not too intertwined with each other. So before we submit jobs to Rosetta@home, we often cut up a sequence into putative domains(based on algorithms like Ginzu and RosettaDom, developed by members of the Baker lab). Then we try to fold those pieces of the protein individually, and we can now use Rosetta@home (5.24 onwards) to take those individual sub-folds and pack them together into the final structure for the whole protein chain.


Version 5.24 is up:

1. The symbol store that we put in 5.22 was not properly activated, so we're giving it another shot.

2. We can now use prior predictions for which parts of the chain are buried or exposed to guide the Rosetta search.

3. We can efficiently assemble predefined domains of the protein chain into a whole structure.



ID: 19541 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Moderator9
Volunteer moderator

Send message
Joined: 22 Jan 06
Posts: 1014
Credit: 0
RAC: 0
Message 19571 - Posted: 30 Jun 2006, 14:18:17 UTC

This thread has been started to store posts relating to release log entries.
Moderator9
ROSETTA@home FAQ
Moderator Contact
ID: 19571 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Bob_Pombrio

Send message
Joined: 5 Jun 06
Posts: 1
Credit: 20,574
RAC: 0
Message 19629 - Posted: 1 Jul 2006, 13:03:54 UTC - in response to Message 19528.  

Rosetta 5.25 is up!
In 5.24 (and previously) there were occasional runs that were
producing funny behavior for some
clients that had stopped Rosetta and
restarted from a checkpoint. These clients would repeatedly
start from the same checkpoint repeatedly, even after
finishing each prediction. Keep us posted on bugs!

Quick note: some of the posts to this thread have been moved here
to keep the version log clean. Thanks!



How about some update instructions! I've also noted that if I stop running the graphics I do not get system lockups any longer on Win Xp.
ID: 19629 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile anders n

Send message
Joined: 19 Sep 05
Posts: 403
Credit: 537,991
RAC: 0
Message 19633 - Posted: 1 Jul 2006, 13:26:42 UTC - in response to Message 19629.  
Last modified: 1 Jul 2006, 14:06:15 UTC

How about some update instructions! I've also noted that if I stop running the graphics I do not get system lockups any longer on Win Xp.


Hi Bob

The 5.25 app. uploads automatic, no need to do anything.

About the grafics problem some grafic cards do have problems it´s not Win Xp.

Check your Boinc client so you have the latest release.

Happy cruching :)

Anders n

ID: 19633 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Astro
Avatar

Send message
Joined: 2 Oct 05
Posts: 987
Credit: 500,253
RAC: 0
Message 19667 - Posted: 2 Jul 2006, 3:46:30 UTC - in response to Message 19629.  

How about some update instructions! I've also noted that if I stop running the graphics I do not get system lockups any longer on Win Xp.

I too used to get the fatal windows errors with Rosetta, but since 5.24 they've been gone. you might be able to use the screensaver again.

tony
ID: 19667 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote

Message boards : Number crunching : Discussions about release log entries



©2024 University of Washington
https://www.bakerlab.org