[ghemical-devel] Re: Ghemical

[Geoffrey Hutchison]

Ciao Francesco,

I'm forwarding your message to the Ghemical mailing list and will  
make a few comments. I certainly appreciate your feedback -- it's  
good to hear from users. (By the way, what do you describe as "large"  
molecules? 40-50 atoms? More?)

I agree with the difficulty in running DFT jobs under MPQC in  
Ghemical right now. One workaround, obviously is to use the "export"  
feature to write an MPQC input file directly. This has the advantage  
of running a batch of MPQC jobs on another computer (which is what I  
usually do with GAMESS or MPQC).

You mention "lack of a modern molecular mechanics program" for  
"reasonably minimized conformers." I'm not sure what you mean -- I've  
done Monte Carlo conformer searching using Ghemical and found pretty  
reasonable conformers. In fact, I think the MC searching algorithm in  
Ghemical is pretty good -- often finding better candidates than other  
programs. While Ghemical doesn't offer tons of MM force fields, I  
think its MM parameters are pretty decent.

I'm not sure when the Ghemical-2.0 release will happen, but I suspect  
that Michael Banck will try to get a Debian package fairly soon --  
he's one of the developers of Ghemical too.

Best regards,
-Geoff

> version installed on my debian testing/unstable is 1.91-2.
>
> making clear that i am a natural product chemist, therefore  
> interested in
> computation for "large" organic molecules, the problems i noticed,  
> and which
> let me set aside ghemical for the time being, are:
>
> --lack of input to mqcp for DFT calculations (many body  
> calculations cannot be
> carried out on my molecules; choosing a restrited basis set, they are
> useless);
>
> --lack of a modern  molecular mechanics program to input to mqcp  
> reasonably
> minimized conformers.
>
> although i am quite new to mqcp i am investing on it (to escape  
> institutional
> restrctions for cpu time, i am building at home a workstation based  
> on a
> couple of dual core amd64 opterons on a tyan k8we mainboard and a  
> couple of
> hd raid1 300GB each and i plan to feed it debian testing). i was  
> attracted by
> the native conception of mqcp for parallel calculations, so that i  
> am even
> planning to change to it on mainframe. i have experimented a few  
> calculations
> on a amd athlon 1ghz and found them absolutely faster than on  
> gamess (i do
> not know gaussian which is used by my colleagues: i believe that  
> carrying out
> molecular calculations blindly on proprietary software is a waste  
> of time)
>
> that said, it will take time before i am ready to carry out  
> calculations on
> the new workstation. however, i am extremely interested in the new  
> version of
> ghemical. certainly an input to ab initio from a graphical  
> interface relieves
> fro the hardest problems if the molecule is large. it would be  
> great if you
> can provide me a debian package in advance (please the package  
> because these
> trials are during spare time: the burden of normal duties is heavy  
> indeed and
> compilations normally take time).
>
> finally, thank you very much for contacting me. it was great pleasure
>
> francesco
>
> On Friday 31 March 2006 02:28, you wrote:
>> Dear Francesco,
>>
>> While there are other open QM programs, I can certainly understand
>> your choice of MPQC. Have you checked out the free program Ghemical?
>> The current development version for v2.0 is quite good -- I use it
>> for molecular mechanics and conformational analysis daily. I'd be
>> glad to send you a "snapshot," or you can wait a few weeks for v2.0
>> to be released.
>>
>> http://www.uku.fi/~thassine/ghemical/
>>
>> (Incidentally, Ghemical uses Open Babel for file import/export,
>> although pieces exist to run MPQC and MOPAC directly, and support for
>> GAMESS-US has been contributed recently.)
>>
>> Cheers,
>> -Geoff
>>
>> On Mar 29, 2006, at 2:36 PM, Francesco Pietra wrote:
>>> preoccupied because i am investing in mpqc for my computational
>>> future,
>>> abandoning other software.
>>>
>>> the generation of unambiguous coordinates for all atoms of a
>>> complex organic
>>> molecules (from 20 to 40 carbons, plus all other atoms, IDROGENS
>>> INCLUDED) is
>>> what i need to build an input for ab initio computation. and the
>>> generation
>>> can only be from graphics, because it is only the graphics that  
>>> allows
>>> debugging in case of complex molecules (unless one aims at doing
>>> that by
>>> inspection of dreiding model and with the aid of mechanicall metric
>>> tools,
>>> which, however, will never work). and the input to mpqc must be
>>> from well
>>> minimized conformers through a modern force field, otherwise there
>>> is no hope
>>> for any ab initio method.
>>>
>>> which talk is not to disregard the work that is being done around.
>>> difficult,
>>> heavy work. but the competion of proprietary software is severe.
>>> most of this
>>> creates only illusion, but the atomic coordinates are generated
>>> accurately.