SASREF06 vs SASREF07

Interactive modelling (MASSHA, SASpy) and global minimization programs (SASREF, BUNCH, CORAL, GLOBSYMM)
Post Reply
Message
Author
biox
Active member
Posts: 17
Joined: 2010.06.04 09:11

SASREF06 vs SASREF07

#1 Post by biox » 2012.08.24 11:13

Hello,
I have good quality saxs data for a multi-domain elongated glycosylated protein complex whose crystal structure is not known yet. The position and possible type of glycosylation is known. I am using the homology based models for individual domains and some homology based contact restrictions. I have optimized the SASREF06 chi sqr from ~4 to 2.8 so far while crysol [version 2.8 (r965)] chi sqr reduced from 9 to 2.6 . All these runs were on local supercomputer.

Recently, for a run, I submitted the input files on local supercomputer running SASREF06 and on the ATSAS online server running SASREF07. I got a very different SASREF chi sqr values for these 2 different servers.

>>>>>On local cluster, 5 different runs with identical input files produced models with chi sqr from 2.8 to 3.0 as reported by SASREF06 while crysol 2.8 determined chi sqr to be between 2.9 to 4.1 (average=3.5) .

>>>>>For identical input files, single SASREF07 run on ATSAS online server produced incrdeible chi sqr of 1.08. But when I ran crysol 2.8 on the model produced by this SASREF07, the chi sqr value was 3.6 .

I checked and found SASREF06 vs SASREF07 uses different penalties and so on, but what is the advantage of using SASREF 07 then, if the crysol chi sqr is again the similar? Also what produces such a large discrepancy between SASREF07 chi sqr vs Crysol 2.8 chi sqr?

Thanks for the reply in advance.

SaxsMax
Administrator
Posts: 348
Joined: 2007.08.03 19:16

Re: SASREF06 vs SASREF07

#2 Post by SaxsMax » 2013.01.25 14:54

Sasref7 additionally adjusts a background constant while Sasref6 does not. That is why the former typically yields better fits.
Difference between Sasref and Crysol is in shell evaluation:
in Sasref(s) each individual subunit is surrounded by the shell,
in Crysol the shell is evaluated for the molecule as a whole.

biox
Active member
Posts: 17
Joined: 2010.06.04 09:11

Re: SASREF06 vs SASREF07

#3 Post by biox » 2013.01.25 15:27

Thanks for the reply.
Is it not true that the crysol chi square value is more meaningful for a model than chi square reported by sasref?
The chi square value which is included in publication is also the crysol chi square.
I have this question in my mind since a long time. Why the SASREF algorithm does not refine a model based on the shell of the whole protein rather than based on the chi square calculated for individual domain? Especially for a multi-domain proteins (domains ~ 8 or 9), when I run multiple runs (lets say 20), and plot the sasref chi vs. crysol chi square, i found an inverse relationship. Models with lower sasref chi had higher crysol chi square and vice versa. Don't you think, inclusion of crysol chi square calculation after each iteration during sasref refinement and evaluation of models from one step-to-next step based on this whole-protein hydration shell chi square will lead sasref refinement to a more meaningful model?

SaxsMax
Administrator
Posts: 348
Joined: 2007.08.03 19:16

Re: SASREF06 vs SASREF07

#4 Post by SaxsMax » 2013.01.25 17:29

Crysol in its present form also has its limitations.
Particularly, for elongated/skewed particles as well
as for the shapes with the cavities, envelope function
evaluated on Fibonacci grid would not present the hydration shell very accurate.

A new Crysol version is actually planned for the next Atsas release which shall overcome these issues.

sch
Member
Posts: 8
Joined: 2014.02.14 17:27

Re: SASREF06 vs SASREF07

#5 Post by sch » 2014.02.14 17:31

SaxsMax wrote:A new Crysol version is actually planned for the next Atsas release which shall overcome these issues.
So has this change been made? I saw that in SASREF included in ATSAS 2.5.2 now has a random seed input option but did not see any other reported changes. Are these issues still a problem? (Was this change made in 2.5.1?)

franke
Administrator
Posts: 408
Joined: 2007.08.10 11:09
Contact:

Re: SASREF06 vs SASREF07

#6 Post by franke » 2014.02.14 22:31

Such changes are never in any bugfix release (or shouldn't be). The new crysol will come with 2.6.0 at the earliest.
For a rough guideline, see: Semantic Versioning.

Post Reply