BUNCH - forcing domains apart

Interactive modelling (MASSHA, SASpy) and global minimization programs (SASREF, BUNCH, CORAL, GLOBSYMM)
Post Reply
Message
Author
grondinjm
Active member
Posts: 36
Joined: 2012.06.27 16:27

BUNCH - forcing domains apart

#1 Post by grondinjm » 2014.07.25 17:27

Hello,

BUNCH modelling of a 2-domain protein connected by a linker (13aa) is giving models in which the 2 domains are pushed together and the linker is shoved up inside one of the domains.

How would I write the .cnd file such that a minimum distance is enforced between the domains? I'm not sure what a reasonable distance is to subject on this, nor do I fully understand how the .cnd file works. The domains are chains A and B, and there is a TER card between the two to denote the linker region. Chain A = 1-620, Chain B = 27-160. I want to separate A_571 from B_80 (which is residue number 53 in the chain). As a test I have tried:

dist 7.0
1 571 571 2 53 53

BUNCH does not accept this file. Please help! I am so confused. I am on a really tight time crunch on this, so any help is greatly appreciated.

Thanks,
JG

Alex
Active member
Posts: 692
Joined: 2007.08.09 21:10
Location: Planet Earth

Re: BUNCH - forcing domains apart

#2 Post by Alex » 2014.07.26 21:20

BUNCH modelling of a 2-domain protein connected by a linker (13aa) is giving models in which the 2 domains are pushed together and the linker is shoved up inside one of the domains.
What does the MW estimation suggest? Does it correspond to the mass you expect? I observed smth similar when my model was wrong, i.e. data corresponds to
smaller protein than expected.
It is hard to pinpoint your problem withuot seeing data/fit and models. so, you should:
1. check MW estimation - is that what you expect?
2. is it flexible? Kratky/d-Kratky plot?
3. before CORAL modeling place both domain in origin of coordinates and don't fix them.
4. or based on #2 consider using EOM.

HTH,
Alex

grondinjm
Active member
Posts: 36
Joined: 2012.06.27 16:27

Re: BUNCH - forcing domains apart

#3 Post by grondinjm » 2014.07.27 03:19

1. Actually you might be able to help me with this since we aren't sure how to go about this: we used online HPLC-SAXS to collect this data on an instrument that uses water as a standard, so I am not sure how to calculate MW based on this. Emails to beamline operators have not been returned, and my co-workers also don't know how to do this. DLS and its elution from the column suggest that it is monomeric.
2. Not flexible by Kratky or Porod.
3. I've never used CORAL, however I did not fix the domains in this BUNCH trial.
4. See #2

Since asking this questions I sent the data to a friend who ran BUNCH independently using the same input files, and he is getting normal-looking models where this linker is not bunched up. To my knowledge he has used default BUNCH settings with no domains fixed, however in comparing out .log files it seems that some of our BUNCH parameters are slightly different. I tried to re-run BUNCH on my end using his values however it seems these are values fixed to the program. The biggest difference I can see is between the 'cross penalty' values.

Do you see anything in these .log files that might identify the problem? Is there a way to change fixed settings like 'cross penalty'? "B1.log" is my friend's run, and "cat32-1_nf_b4.log" is from mine.

Thanks,
JG
Attachments
B1.log
(14.52 KiB) Downloaded 156 times
cat321_nf_b4.log
(14.18 KiB) Downloaded 145 times

Alex
Active member
Posts: 692
Joined: 2007.08.09 21:10
Location: Planet Earth

Re: BUNCH - forcing domains apart

#4 Post by Alex » 2014.07.28 12:25

1. Actually you might be able to help me with this since we aren't sure how to go about this: we used online HPLC-SAXS to collect this data on an instrument that uses water as a standard, so I am not sure how to calculate MW based on this. Emails to beamline operators have not been returned, and my co-workers also don't know how to do this. DLS and its elution from the column suggest that it is monomeric.
i had a good experience calculating MW using Porod volume for each recorded frame in the peak and calculating average. For my samples, it gave very close values to what I expect. I wrote a program to do this, however, it relies on certain file naming, in particular, used at Soleil HPLC-SAXS setup (foxtrot subtracted files). Where did you collect your data?
Since asking this questions I sent the data to a friend who ran BUNCH independently using the same input files, and he is getting normal-looking models where this linker is not bunched up. To my knowledge he has used default BUNCH settings with no domains fixed, however in comparing out .log files it seems that some of our BUNCH parameters are slightly different. I tried to re-run BUNCH on my end using his values however it seems these are values fixed to the program. The biggest difference I can see is between the 'cross penalty' values.
Do you see anything in these .log files that might identify the problem? Is there a way to change fixed settings like 'cross penalty'? "B1.log" is my friend's run, and "cat32-1_nf_b4.log" is from mine.
i see very high penalties, which suggest the model is not very reliable. Again, if i were you i would focused first on MW estimation and validation. Btw, do you see a single peak off
the column?
why don't u do simple tests - validate individual subunits; put subunits together in a manner you expect them to be located - does it fit better than individual subunits?

This is anyway like pointing fingers in the sky as I don't see your data/models.

HTH, Alex

grondinjm
Active member
Posts: 36
Joined: 2012.06.27 16:27

Re: BUNCH - forcing domains apart

#5 Post by grondinjm » 2014.07.28 20:35

1. Conveniently, I collected on SWING at SOLEIL!

Also, I think we have found a solution to the problem. To follow up on the test I did with my friend, we figured out that there is something strange going on with the prebunch.pdb I was using. Using the same .pdbs to create .alm in the same manner with crysol and using all the default settings for prebunch, both on the same versions of PREBUNCH, CRYSOL and BUNCH, we both get different prebunch.pdb files. Using his, I generate normal-looking models now. Not sure what the issue is, but the problem has been solved.

Post Reply