I know the source of the differences now. They are relatively minor
in terms of effect, but of course it is good to know for certain.
As I mentioned in another post, we are working on updating the
parameter sets to make parameter set mixing easier. Some of the
traditional ways we've done things don't work so well now, so we're
changing them.
The main difference in energies you saw was due to the fact that some
glycans contain n-acetyl groups (e.g., GlcNAc). An equivalent
sequence of atoms also occurs in each peptide backbone plane. To
maintain planarity around the NAc nitrogen in a carbohydrate
environment, however, a slightly higher force constant is needed than
what is provided in the protein force fields.
Traditionally, improper dihedrals like this have been written with
wild-cards. That is, one can be pretty sure that anything around atom
type "N" needs to be planar. For this reason, the improper is written
as "X -X -N -H ". So, when the glycam force field was loaded second,
the higher glycam dihedral force constant was applied to all
situations containing atom type N bound to type H. When 99SB was
loaded first, its constant was applied everywhere.
I'm working on an update to the Glycam parm set that does away with
wild cards. Instead, all impropers will be stated explicitly. I'm
still testing it, but it seems to work. This will make ordering far
less of an issue.
The other difference you saw was in the force constant and equilibrium
value for the C-N-H angle, also associated with atom sequences similar
to the peptide plane. For whatever reason, Glycam is still using the
parm94 values for that. This probably would not have caused major
problems -- that angle would merely have bent a little more easily.
The energies would differ a little, but the basic behavior should be
the same. However, I am testing the current glycam force field with
those values changed to the parm99 values. I don't expect significant
differences.
In all of these simulations, the total energy isn't necessarily
physically meaningful. The energy differences, however, should give
physically meaningful values. In those cases, it is most important to
be consistent. That is, if you load 99SB then GLYCAM for some
simulations, load them the same way for all simulations that will be
compared to each other.
In the file that you sent me the glycans are simple -- just mannose.
So, it doesn't matter a lot. But, to be as true as possible to the
99SB force field, in *this* case, it would be better to load Glycam
first then 99SB. If you have a different system, for example
containing a GlcNAc or a Neu5Ac, etc., things change. Like I said,
I'm working on a parameter set that will make those easier. In this
new set, the top file that would be generated for your system (not
necessarily another) will be identical to the one generated if you
load Glycam first then 99SB.
You are probably also concerned about the differing number of dihedral
parameters in the top file. That is ok, too. When leap generates a
top file, it tries to conserve space. So, if it finds two sets of
parameters that are identical, it only records those once, and just
has all relevant sets of atoms reference the one set. When Glycam was
loaded second, the saved parameters had one more unique set. So,
that's fine.
I'll send a separate announcement out in a day or so once I get the
new param set in shape.
On Thu, Dec 8, 2011 at 2:59 PM, Sai Kumar Ramadugu <sramadugu.gmail.com> wrote:
> Hi Lachele,
> I have downloaded the latest glycam parameter files. We have AmberTools
> 1.5 installed with bugfixes upto 7. I used the latest glycam parameters to
> generate the topology and coordinates files with loading glycam first and
> ff99SB second and vice versa.
>
> If I look at the differences in two topology files, the following are
> different.
> 1. One of the angle force constant is different ( 50 kcal/mol*rad^2 vs 30
> kcal/mol*rad^2) along with equilibrium angle value (2.094396 rad
> vs 2.129302 rad)
> 2. Number of dihedral types (NPTRA) are different 60 vs 61. Since the
> number of dihedral types are one more than other, there are additional
> periodicity and phase values added.
>
> Now I have downloaded the latest bugfixes that have upto 22 on
> ambermd.orgwebsite. I'll patch the latest bugfixes and will update you
> on the list if
> there are any differences. If I find the differences after the latest
> bugfixes, then I'll send you the pdb, parm7 and rst7 files for my system.
>
>
> Thank you for your time.
>
>
> Regards
> Sai
>
>
>
>
>
> On Thu, Dec 8, 2011 at 11:22 AM, Lachele Foley (Lists) <lf.list.gmail.com>wrote:
>
>> First off, make sure you're using a recent, fully patched version of
>> AmberTools 1.5. Use tleap or xleap to generate the files (not sleap).
>> There have been similar issues in the past that were fixed.
>>
>> We also recommend using the most up-to-date version of GLYCAM (see
>> glycam.org/params).
>>
>> Ideally, the order doesn't matter, and we try to fix conflicts as soon
>> as we know about them, but occasionally they happen. In that case, it
>> is generally better to load ff99SB first and then Glycam. If you are
>> using the most recent tools and parameters, please give us more
>> details about the system you are building as well as the versions of
>> AT and glycam.
>>
>>
>>
>> On Thu, Dec 8, 2011 at 12:07 PM, Sai Kumar Ramadugu <sramadugu.gmail.com>
>> wrote:
>> > Dear Amber Users,
>> > I have a complex of protein with a disaccharide. When I load the ff99Sb
>> > first and GLYCAM second versus GLYCAM first and ff99SB second, the energy
>> > terms for a zero step minimization in gas phase are different for angle,
>> > dihedral, 1-4LJ, 1-4Coulomb, and the total potential energy.
>> > Below are the output values from the respective minimizations:
>> > 1. ff99SB first, GLYCAM second
>> >
>> > NSTEP ENERGY RMS GMAX NAME
>> > NUMBER
>> > 1 -1.3061E+03 1.3969E+02 3.4776E+03 SG 1025
>> >
>> > BOND = 85.4389 ANGLE = 175.7590 DIHED =
>> > 231.3310
>> > VDWAALS = 855.0251 EEL = -7406.8357 HBOND =
>> > 0.0000
>> > 1-4 VDW = 428.9510 1-4 EEL = 4324.2069 RESTRAINT =
>> > 0.0000
>> >
>> > 2. GLYCAM first and ff99SB second
>> >
>> > NSTEP ENERGY RMS GMAX NAME
>> > NUMBER
>> > 1 -2.3382E+02 1.3969E+02 3.4776E+03 SG 1025
>> >
>> > BOND = 85.4389 ANGLE = 170.9682 DIHED =
>> > 881.3310
>> > VDWAALS = 855.0251 EEL = -7406.8357 HBOND =
>> > 0.0000
>> > 1-4 VDW = 430.1394 1-4 EEL = 4750.1157 RESTRAINT =
>> > 0.0000
>> >
>> >
>> > The differences appear due to the different values of equilibrium angle
>> > value and force constant value for one atom,
>> > the dihedral angle force constants and phase are different, the scee and
>> > scnb values are different after the mixed scaling is included. Therefore
>> > the respective energy terms are different in the output files.
>> >
>> > Is it due to the fact that some of the atomtypes are same in both ff99SB
>> > and GLYCAM and the parameters get overwritten based on the order of
>> > inclusion of leaprc files in xleap?
>> > When I was going through the manual (AmberTools) for carbohydrates, the
>> > order in the manual is load GLYCAM first and ff99SB second.
>> > Does this mean that whenever you have a protein and sugar complexed, the
>> > order of leaprc files is always 1st GLYCAM and 2nd ff99SB?
>> >
>> > Any suggestions will be helpful.
>> >
>> >
>> > Regards
>> > Sai
>> > _______________________________________________
>> > AMBER mailing list
>> > AMBER.ambermd.org
>> > http://lists.ambermd.org/mailman/listinfo/amber
>>
>>
>>
>> --
>> :-) Lachele
>> Lachele Foley
>> CCRC/UGA
>> Athens, GA USA
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
--
:-) Lachele
Lachele Foley
CCRC/UGA
Athens, GA USA
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Dec 15 2011 - 10:30:02 PST
