Page 1 of 3 1 2 3 >
Topic Options
#26357 - 01/28/11 07:28 PM Merge silicate with protein parameters
gianluca Offline
Forum Member

Registered: 02/05/04
Posts: 68
Loc: Seattle, WA and Zürich, Switze...
I would like to calculate the interaction energy between a protein and a quartz surface. For that I have created the quartz surface as in the example in the folder "toppar/silicates". Now, I need to merge the topology and parameters of silicates with those of amino acids. For this I do:

-----------------------------------------------
! { Read protein topology }
OPEN READ UNIT 2 CARD NAME top_all22_prot.inp
READ RTF CARD UNIT 2
CLOSE UNIT 2

! { Read silicates topology }
open read formatted unit 2 name top_silicates.inp
read rtf card unit 2 append
close unit 2

! { Read protein parameters }
OPEN READ UNIT 2 CARD NAME par_all22_prot.inp
READ PARAM CARD UNIT 2
CLOSE UNIT 2

! { Read silicates parameters }
open read formatted unit 2 name par_silicates.inp
read para card unit 2 append
close unit 2
-----------------------------------------------

I renumbered the masses in top_silicates.inp so that they do not overwrite the masses in the protein topology file and I also deleted RESI TIP3P and the corresponding atoms from par_silicates.inp in order to avoid a conflict with the TIP3P defined in par_all22_prot.inp. Is this procedure correct?

After this, I load the PSF and PDB of the quartz surface and then I try to generate the PSF of my protein:

! { Read sequence of the protein }
OPEN UNIT 1 READ CARD NAME ./protein.pdb
READ SEQUENCE PDB UNIT 1
CLOSE UNIT 1

! { Generate the psf structure }
GENERATE PROT FIRST ACE LAST CT3 WARN SETUP

Here, I receive the errors:

<CODES>: No atom parameters for atom 1 type 108
<CODES>: No atom parameters for atom 2 type 107
<CODES>: Total number of atoms with no paramter: 2592
<CODES>: No angle parameters for 1 ( CAL CAL )
.
.
<CODES>: And more missing angle parameters.
<CODES>: A TOTAL OF20067 MISSING PARAMETERS

I have no CAL in my protein! However, if I set "bomlev -2", it continues and produces a correct PSF/PDB of surface+protein. I am even able to perform a steepest descent minimization and all seems fine. However, I keep getting warnings about missing bond and angle parameters for CAL and MG. Is what I'm doing still correct? My final goal is to perform Monte Carlo simulations where I rotate the protein (stiff) on the surface in order to find the energetically most favorable orientation and distance with respect to the surface.

Any hint would be greatly appreciated!

Top
#26358 - 01/28/11 08:06 PM Re: Merge silicate with protein parameters [Re: gianluca]
rmv Online   content

Forum Member

Registered: 09/17/03
Posts: 7350
Loc: 39 03 48 N, 77 06 54 W
After changing the integer atom parameter numbers for the MASS statements, you must make a new PSF for the silicate, using the modified topology file. It's not clear that you've done that.

You should probably be using PSF+COOR instead of PSF+PDB
_________________________
Rick Venable
computational chemist


Top
#26359 - 01/28/11 09:06 PM Re: Merge silicate with protein parameters [Re: rmv]
gianluca Offline
Forum Member

Registered: 02/05/04
Posts: 68
Loc: Seattle, WA and Zürich, Switze...
Originally Posted By: rmv
After changing the integer atom parameter numbers for the MASS statements, you must make a new PSF for the silicate, using the modified topology file. It's not clear that you've done that.

I have not done it. I will do it. Thanks!

Top
#26360 - 01/29/11 02:21 AM Re: Merge silicate with protein parameters [Re: rmv]
gianluca Offline
Forum Member

Registered: 02/05/04
Posts: 68
Loc: Seattle, WA and Zürich, Switze...
Thank you! Everything works now. I don't get any more the error messages and I no longer need "BOMBlev -2".

I have another question though. Is it easy to modify crystal.img to take into account the presence of the protein? I am using the same crystal.img and setup as in the test.inp file in the "toppar/silicates" folder. Obviously, I get during minimization a lot of warnings that the protein atoms are too close to the quartz surface.

Do I need images to perform rigid body Monte Carlo simulations?

Thank you!

Top
#26365 - 01/29/11 05:24 PM Re: Merge silicate with protein parameters [Re: gianluca]
lennart Online   content

Forum Member

Registered: 09/25/03
Posts: 4267
Loc: ~ 59N, 15E
"no longer need bomlev -2" - bomlev -2 is NEVER part of a correct procedure.
_________________________
Lennart Nilsson
Karolinska Institutet
Stockholm, Sweden

Top
#26367 - 01/29/11 06:02 PM Re: Merge silicate with protein parameters [Re: lennart]
gianluca Offline
Forum Member

Registered: 02/05/04
Posts: 68
Loc: Seattle, WA and Zürich, Switze...
Originally Posted By: lennart
"no longer need bomlev -2" - bomlev -2 is NEVER part of a correct procedure.


I have been wondering how come that "bomlev -2" is set at the beginning of test.inp in the folder "toppar/silicates". But I totally agree with you. Thanks for pointing this out.


Edited by gianluca (01/29/11 06:04 PM)

Top
#26368 - 01/29/11 07:16 PM Re: Merge silicate with protein parameters [Re: gianluca]
rmv Online   content

Forum Member

Registered: 09/17/03
Posts: 7350
Loc: 39 03 48 N, 77 06 54 W
I've never really looked at the silicates subdir, or even unpacked the .tar.gz archive until now. It looks like one could adjust the values in param.str to change the system size; changing the value for Z might help. You can also extend the slab in the xy plane by adjusting n and m values.

You need images to have an infinite quartz slab, and this approach uses image patches, so the slab edges may be ill-defined w/o images and the image patches. The protein is also subject to the image transformations, so you may need a bigger slab, such that protein is not interacting with image copies of itself


(Almost never, anyway. The exceptions are very rare, but I'm aware of 2 cases where one must use bomlev -2; normal modes of larger molecules in VIBRAN, and reading trajectory files out of sequence. Perhaps these two cases should be downgraded to -1.)
_________________________
Rick Venable
computational chemist


Top
#26372 - 01/30/11 12:13 AM Re: Merge silicate with protein parameters [Re: rmv]
gianluca Offline
Forum Member

Registered: 02/05/04
Posts: 68
Loc: Seattle, WA and Zürich, Switze...
Oh right! I missed the Z. I set it to 200 and now it's working!

I wonder whether if you make the slab large enough, the protein might not experience the ill defined edges.

Right, until now I have only used bomlev -2 to read trajectory files out of sequence. I was surprised to see it in a test file. They probably forgot to take it out. That's why I first thought that I have to use it, too.

I want to perform a Monte Carlo simulation with protein+quartz surface and I wonder whether the following parameters are correct for nonbonded interactions:

nbonds atom shift rdie vdw vswitch -
cutnb 16 ctofnb 12 ctonnb 10 wmin 1.5 eps 2.0

If I do a SD minimization with these values, the slab expands in the z direction. If I use:

nbonds atom fshift cdie vdw vshift -
cutnb 16 ctofnb 12 ctonnb 10 wmin 1.5 eps 1.0

the slab keeps its shape. However, I might want to use a distant dependent dielectric. What are the best values?

Thanks


Edited by gianluca (01/30/11 12:14 AM)

Top
#26377 - 01/30/11 01:15 PM Re: Merge silicate with protein parameters [Re: gianluca]
rmv Online   content

Forum Member

Registered: 09/17/03
Posts: 7350
Loc: 39 03 48 N, 77 06 54 W
Both SHIFT and RDIE are bad choices and should not be used.

If you ignore the parts about the IPS method, the RECOMMENDED: section in nbonds.doc is still essentially correct.

If you can't use Ewald, use FSWITCH or FSHIFT for electrostatics. For the VDW term, VSWITCH or VFSWITCH are recommended.

The ill-defined edges have unsatisfied connectivity from implied bonds; the protein is not relevant to that.
_________________________
Rick Venable
computational chemist


Top
#26381 - 01/30/11 03:56 PM Re: Merge silicate with protein parameters [Re: rmv]
gianluca Offline
Forum Member

Registered: 02/05/04
Posts: 68
Loc: Seattle, WA and Zürich, Switze...
Originally Posted By: rmv

The ill-defined edges have unsatisfied connectivity from implied bonds; the protein is not relevant to that.


For my Monte Carlo moves, I leave the surface stiff. I only need it for electrostatics and VDW evaluation between protein and surface. If I keep the center of mass of the protein above the center of the surface, the edges should not matter, should they?

Top
Page 1 of 3 1 2 3 >

Moderator:  alex, chmgr