VMD-L Mailing List
From: Francesco Pietra (chiendarret_at_gmail.com)
Date: Sun Oct 04 2015 - 06:01:37 CDT
Unexpectedly, unlike in the past, I am unable to reproduce (Gaussian DFT
B3LYP opt or opt=tight in vacuo) the crystal structure of the di-Fe(II)
cluster. One Fe comes out correctly with a square bipyramid octahedral
coordination. The coordination around the other Fe, which should include a
water molecule, is disrupted: instead of a trigonal bipyramid coordination
including water, the water molecule becomes H-bonded to a carboxylate and
Fe is left in a awkward situation. That is of no interest to the audience.
What may be of interest, is that I do not remember whether I used before an
implicit environment for the quantum mechanical calculations. Otherwise,
reproducing the crystal structure with calculations in vacuo should be
viewed as a miracle for this complex cluster. A miracle that was not
repeated in these days.
Unfortunately, not all from previous calculations was saved (it was a
couple of years ago, abandoned as I was unable to extract the force
constants). The pdb, obtained at those time with newzmat, shows that the
crystal structure was reproduced correctly. What else I still have is a
.fchk file reporting the freq analysis only, and the related Hessian matrix
that I extracted from the .chk file. Can these be submitted to ffTK or the
complete Gaussian log file is needed?
Thanks for a comment on in vacuo vs implicit solvent calculations and, of
course, about the ffTK requirements.
On Sat, Oct 3, 2015 at 10:55 PM, JC Gumbart <gumbart_at_ks.uiuc.edu> wrote:
> You are fine then! qmtool already had the scaling factor for B3LYP
> frequencies built in.
> We are dealing with exactly the same issues with cobalamins with a Co(I),
> (II), or (III) at the center. But if it’s fully bonded, then what atom
> type you use is irrelevant. Just create a new one and then calculate all
> the bonded terms from scratch. We are struggling because we are trying to
> come up with rational LJ parameters for a Co atom that is not fully bonded.
> On Oct 2, 2015, at 2:03 AM, Francesco Pietra <chiendarret_at_gmail.com>
> Hi James:
> I am carrying out repeated opt-freq (just to be sure to have a clean log
> file) of the metal cluster at B3LYP/6-31G*. Therefore, if I understand, the
> Gaussian log file should be handled by ffTK at its present release. If not,
> I would be grateful for your additional code. I'll let you know.
> A cluster with two bridged Fe(II) is not the easiest way to approach this
> problem with ffTK. Just devising Fe atom types for Fe at the center of (a)
> trigonal bi-pyramid and (b) square bi-pyramid Fe(II), with equatorial O
> atoms and axial N, O atoms is - for me - no trivial task. If you have any
> suggestion about model atom types from which to begin, it would be great
> francesco pietra
> On Thu, Oct 1, 2015 at 11:38 PM, JC Gumbart <gumbart_at_ks.uiuc.edu> wrote:
>> Hi Francesco,
>> Dealing with transition metals is a huge problem for both QM and MD. I
>> recognized that the bond/angle part of FFTK (because of its reliance on the
>> QMTool plugin) can’t handle any DFT methods other than B3LYP. We’ve
>> recently used BP86, so I changed the source code to add the scaling factor
>> for this method. All force constants read from the hessian have to get
>> scaled slightly, with HF being the biggest (about (0.9)^2) and the DFT
>> methods being much lower (BP86 is (0.99)^2). If you want, I can share the
>> source-code with you off-list; it’s just a few small changes.
>> See this paper for more discussion:
>> http://pubs.acs.org/doi/abs/10.1021/jp960976r (wow, it’s been cited 6000
>> I’m really glad you asked this question, as I hadn’t looked at that part
>> of the code for a very long time!
>> On Sep 27, 2015, at 2:20 PM, Francesco Pietra <chiendarret_at_gmail.com>
>> There was some discussion about parameterizing transition metal proteins
>> at the onset of ffTK, in February 2012.
>> I have lost details on how ffTK has been further developed, however I
>> came across a recent statement that the QM part of ffTK was taken from
>> Paratool. As the latter was originally applied to parameterize a non-heme
>> iron protein active center (in a way that was focused on deriving bond,
>> angle etc forces, not atom partial charges, just to have a "dirty"
>> parameterization that keeps the structure of the active center in order
>> during MD
>> my question is:
>> Has anyone verified whether the QM part of current ffTK works with a
>> Gaussian log file from a DFT calculation with models of transition metal
>> centers of proteins? My aim is "merely" to have such forces rapidly
>> extracted from the QM matrix (the way that MCPB does that in Amber world
>> for Zn(II) complexes), so that many variations of the coordination
>> situation around the transition metal can be investigated.
>> Which is probably the upper limit for MD dealing with transition metal
>> proteins. However, such proteins make a large fraction of life.
>> francesco pietra