From mayalama@cis.uab.edu  Thu Oct  8 17:49:24 1992
Date: Thu, 8 Oct 92 22:49:24 CDT
From: mayalama@cis.uab.edu (Yalamanchili V M Rao)
To: chemistry@ccl.net
Subject: Neural nets project in chemistry!!!





Dear friends in the computational chemistry list,

	My self and my friend from the chemistry
	are classmates in the neural networks 
	course in the computer science department
	of UAB, Birmingham.  We want to use the
	neural networks technology to any chemical
	problem as a part of our project in the
	course.  

	But we do not have any idea how to proceed.
	We do not know to what problems of chemistry
	this technology is applicable.  We came to
	know that there was a symposium held at 
	SanFransisco on the topic of applications
	of neural nets in chemistry.  We want to
	know some more details, including the 
	problems suitable for a graduate project
	work in the application of neural nets 
	to chemistry. 

	Any pointers in this regard are highly
	appreciated.

	Thanking you in anticipation,

	Please reply to : mayalama@cis.uab.edu


			Mallik

From rs0thp@RohmHaas.Com  Fri Oct  9 04:48:31 1992
From: rs0thp@RohmHaas.Com (Dr. Tom Pierce)
Subject: Re: water placement
To: ross@cgl.ucsf.edu (Bill Ross)
Date: Fri, 9 Oct 92 8:48:31 EDT


> Bill Ross writes:
> I am looking for comment & references on algorithmic water placement
> around a solute molecule. For example, the simplest method I know
> of (indeed the only one I have seen implemented) involves superposing
> a canonical volume of water (or an array of such volumes) on the
> solute and subtracting waters that overlap with solute atoms. This
> leads to vdw voids:  Some of this effort might be alleviated by a 
> better packing strategy, which would lower a hurdle for newcomers to
> the technology. Ideas/references?
>
Dear Folks,
During the development of modeling work at Rohm and Haas, Dr. Mario Blanco
developed a suite of methods and published them in an article titled,
"Molecular Silverware. I. General Solutions to Excluded Volume Constrained 
Problems", J. Computational Chemistry, Vol. 12, #2, 237-247 (1991).
One of the methods works well for solvation problems. Basically this method
packs a molecule around another with rotations and reorientations until 
the smallest radius of gyration is achieved. Then the method adds another 
molecule (solvent) and packs it, etc. The software is available from  Tripos.


-- 
Sincerely,  Thomas Pierce,
thpierce@rohmhaas.com or rs0thp@rohvm1
Official Disclaimer:"The opinions expressed are those of the writer 
and not the Rohm and Haas Company."


From ravishan@swan.wesleyan.edu  Fri Oct  9 05:20:16 1992
Date: Fri, 9 Oct 92 09:20:16 -0400
From: G. Ravishanker <ravishan@swan.wesleyan.edu>
To: ross@cgl.ucsf.edu
Subject: water placement



Bill

We encounter this problem a lot in all our simulations and the strategy we
have been using is the following, which is an overkill of sorts, but has
some advantages:

	1. Create a 3-dimensional grid out of the simulation box, whose
size is the diameter of water. 

	2. Start placing waters in random orientations in those grids
where you do not have a solute atom (or more correctly those grids which
have a solute atom within a threshold of .5 A or so are exculed).

	3. Do a long Monte Carlo simulation on water (and ions if you have
them). 

	4. Start the MD by slowly heating the system to 300K over 10 ps or
so. Equilibrate and then run production.

Most of our simulations are constant volume simulations and this strategy
works out great. When we don't do this and start from a random config and
do MD, there are all sorts of problems, that are obvious. Disadvantages of
this method are that the MC step takes a considerable amount of time to
"converge" (We use the solute potential energy as the convergence
criterion). The box filling algorithm cannot always guarantee exact
density, but you get pretty close to what you want. Advantages are that
the resulting MD looks stable, and that MC run can also be analyzed and
conclusions can be drawn.

What we planned to do from now on is basically take equilibrated boxes
from the strategy above from several molecules that we have studied and
replace the solute and pad any extra waters and remove the contact waters
and run a short MC. Since the water environment is not disturbed to a
great extent in the bulk of box, the MC length for convergence will be
much less than what you need by the strategy mentioned above. We strongly
believe that doing MC to get the waters in position pays off in the long
run in producing stable MD runs.

Hope this helps. Glad to be part of a technical discussion.....

Ravi


From shen@athena.cs.uga.edu  Fri Oct  9 06:01:40 1992
Date: Fri, 9 Oct 92 10:01:40 EDT
From: Mingzuo Shen <shen@athena.cs.uga.edu>
To: chemistry@ccl.net
Subject: Analytic energy second derivatives with ECP?


Hi all,

    I wonder if any ab initio MO based program packages can calculate
analytic energy second derivatives (SCF or MP2?) in the presence of
effective core potentials (ECP). More specifically, I have access to
Gaussian92, CADPAC 4.0, and MOLCAS 2.0, all running under the IBM
RS/6000 if that matters. Any pointers and if available, example
inputs (including custom ECP and accompanying valence basis sets),
are greatly appreciated.

    Thanks in advance,

Mingzuo Shen

From tudor@wucmd.wustl.edu  Fri Oct  9 04:39:36 1992
Date: Fri, 9 Oct 92 09:39:36 -0500
From: tudor@wucmd.wustl.edu (Tudor Oprea)
To: CHEMISTRY@ccl.net
Subject: RE: Biol. Alchem. & objectivity



Dear netters,

Chris Van Dyke is right.
Mea culpa.
I publicly apologize to Robert McElwaine for attacking his personal integrity.
Of course, I sustain my point of view in respect to the scientific "facts"
that he put forth.
As far as those pharmaceutical treatments are concerned, you could check them
in any pharmacy you like - I know what the answer will be.

I guess the disatvantage of the Email is that it goes so quick! Once you have
a first reaction (like being angry) it shows on the list. So I know what hate
mail means. I realize that my way of arguing was wrong. I also realize that I
probably went 'on the rough', far from the 'fairway' of science.
For all these, I apologize. You won't see any hate mail from me.


================================================================================
=  Tudor Ionel OPREA 				      = Tel. (1-314) 935 4672  =
=  Research Associate				      = Fax. (1-314) 935 4979  =
=  Washington University Center for Molecular Design  = Email:                 =
=  Lopata Hall, Box 1099, One Brookings Drive	      = tudor@wucmd.wustl.edu  =
=  St.Louis MO 63130				      =                        =
================================================================================



From UDIM018@FRORS31.bitnet  Tue Oct  9 08:10:44 1992
Date: 09 Oct 92  12:10:44 EDT
From: <UDIM018%FRORS31.BITNET@OHSTVMA.ACS.OHIO-STATE.EDU>
Subject: Enough on Barbie
To: chemistry@ccl.net


E. M. EVLETH
Dynamique des Interactions Moleculaires
Universite Pierre et Marie Curie
4 Place Jussieu, Tour 22, Paris 75005
33-1-44-27-42-08 (work), 33 = France; 1 = Paris
33-1-45-48-67-20 (home)
FAX 33-1-27-42-17
e-mail UDIM018 at FRORS31

Most of American kids and their parents are brain dead TV zombies and
having or not having a Barbie won't make any difference. Readers of this
e-mail are mostly working out of US academic ghettos and probably have
more in common with non-Americans who share the same transnational
intellectual cultural values than with their neighbors on the block.
Basically, Barbie is more American than those in the American academic
community.  Fortunately for the rest of the world that community is
first class and we like having you around.  Assez d'Angst et courage.

From FACCIONE@BRFUEL.bitnet  Fri Oct  9 12:07:10 1992
Date:         Thu, 08 Oct 92 13:43:06 BS3
From: Milton Faccione <FACCIONE%BRFUEL.BITNET@OHSTVMA.ACS.OHIO-STATE.EDU>
Organization: Fundacao Universidade Estadual de Londrina
Subject:      software
To: CHEMISTRY@ccl.net


DEAR NETTERS

I WOULD LIKE TO KNOW IF THERE IS ANY BODY OUT THERE WHO HAS
BIOSYM OR MPGS OR UNICHEM OR MOLECULA SIMULATION (QUANTA,
CHARMm AND X-PLOR) SOFTWARE IN PUBLIC DOMAIN AND HOW CAN I
GET THEM.
IF THERE IS ANY ONE WHO HAS A PROGRAM WHICH IS PUBLIC DOMAIN
WITH THE ABOVE FACILITIES MENTIONED. SO I'D LIKE TO HEAR ABOUT.
ANY HELP WILL BE VERY MUCH WELCOME.
MY COMPUTER IS A IBM 3090
TANKS IN ADVANCE

Milton Faccione
Faccione at brfuel.bitnet

From richard@iris26.biosym.com  Fri Oct  9 01:27:51 1992
To: CHEMISTRY@ccl.net
Subject: Neural nets in chemistry
Date: Fri, 09 Oct 92 08:27:51 -0700
From: Richard Macdonald <richard@iris26.biosym.com>



By coincidence, I just read 	Mallik's posting about
neural nets in chemistry  and I had planned to send my
own related message today.

I am interested in finding references to the use of
neural networks to analyze NMR spectra.  I am new to
NMR so even the obvious refences would be welcome.

If other people have worked with neural networks and
would like to give a reflection on their experiences,
I would most appreciate it.  I expect there may be
may unpublished experiences (failures?) with neural
nets.

Please reply to the list, since I know I enjoy reading
the answers to other peoples questions...


***********************************************************
Richard D. Macdonald, Ph.D.  voice: (619) 546-5527
Biosym Technologies     internet: richard@biosym.com
9685 Scranton Road
San Diego, CA 92121-2777
***********************************************************



From jas@medinah.atc.ucarb.com  Fri Oct  9 08:51:05 1992
Date: Fri, 9 Oct 1992 12:51:05 -0400
From: jas@medinah.atc.ucarb.com (Jack Smith)
To: ross@cgl.ucsf.edu, rs0thp@RohmHaas.Com
Subject: Re: water placement



CompChem Netters:

   I believe that Mario Blanco now works at Molecular Simulations
(Formerly Polygen, Molecular Simulations, Cambridge Molecular
Design,...) and the Silverware software is also, or soon will be, 
part of their PolyGraf program.  The amorphous growth methods
used in some of the commercial polymer modeling packages
seem like a reasonable way to approach the solvent packing issues,
especially for complex solvents or binary mixtures in general.

-- Jack Smith

From jle@world.std.com  Fri Oct  9 10:23:23 1992
Date: Fri, 9 Oct 1992 14:23:23 -0400
From: jle@world.std.com (Joe M Leonard)
To: chemistry@ccl.net
Subject: MM3 request


Me again...

1) Does anybody have the references for the description AND
functional form of the MM3 H-Bond term.  I know it's been
changed in a recent JACS article, but I'm looking for the
original definition.  All I can find so far is a very brief
textual description.

2) I cannot reproduce the published results for formamide if the
7...28 H-Bond is included.  Should the term be included (i.e. is
1-4 interactions valid H-Bonds)?

Joe Leonard
jle@world.std.com

From wsonnen@alnitak.usc.edu  Fri Oct  9 04:49:50 1992
Date: Fri, 9 Oct 92 11:49:50 PDT
From: wsonnen@alnitak.usc.edu (Wayne Sonnen)
To: chemistry@ccl.net
Subject: simple correct formula


Dear netters,

With regards to the probability of selecting a barbie doll
that speaks the phrase "math class is though" as one of
four phrases out of 270 possibilities. The probability is,
assuming equal probability of selecting one of the remaining
phrases, is:

               P1 + P2 + P3 + P4

where P1, P2, P3 and P4 are the probabilities of selecting the
phrase on the first, second, third and fourth try.
Since the order is irrelevant, the probability of selecting
one phrase from 270 is 1/270. 
 therefore, P1 = 1/270

For the second phrase, there
will be 269 possibilities, but we have to figure in that
the first selection was a "miss", therefore we have:
   
     P2 =  (1/269)*(1 - 1/270)

For the third, we have:
 
     P3 = (1/268)*(1 - P1 - P2)

 For the fourth we have,
     
     P4 = (1/267)*(1 - P1 - P2 - P3)

With this formula, i calculate the probability of selecting a barbie
doll, with the above assumptions, to be 0.014856 (1.4856%).

Of course, one could perform a simple calculation, say MC, to arrive
at this same number. But i prefer to think the solution through and save
time.

Wayne Sonnen


P.S. 
 
 I did call Mattel to complain. You should be happy to know that they
are not programming the dolls with the above phrase any more. However,
they are not recalling the dolls either. Therefore the actual probability
of buying a doll with the above phrase is less. Anyone want to try to
figure out how much less?




From d3f012@gator.pnl.gov  Thu Oct  8 02:17:01 1992
Date: Thu, 8 Oct 92 09:17:01 PDT
From: d3f012@gator.pnl.gov
Subject: Re: Rotation Invariance, more thoughts
To: chemistry@ccl.net, jstewart@fai.com



From Jimmy Stewart:...


  Thanks for the response.  I had already caught both of these problems.

Mark Thompson

From jk@iris68.biosym.com  Fri Oct  9 06:43:49 1992
Date: Fri, 9 Oct 92 13:43:49 -0700
From: jk@iris68.biosym.com (Kottalam)
To: chemistry@ccl.net
Subject: Re:  probability of math-phobic Barbie



    My apology to Martin Feyereisen for saying
    his probability calculation also "seems wrong to me".

    His formula in terms of N and r is

                                1
    P =  1 - product of  ( 1 - --- )   for s from 0 to r-1
                               N-s


    and it is ABSOLUTELY CORRECT; No approximations!

    Homework:  Show that the above expression evaluates
               to r/N which I got earlier.

    I am the anti-Barbie doll who says "Math is beautiful".


       Kottalam      jk@biosym.com     (619) 546 5366

       \ | / 
       - O - Scientists are objective;
       / | \ The Earth sucks and they call it gravity.


From ornitz@Kodak.COM  Fri Oct  9 14:41:36 1992
Date: Fri, 9 Oct 92 18:41:36 EDT
From: ornitz@Kodak.COM (Barry Ornitz)
To: CHEMISTRY@ccl.net
Subject: Neural Net Applications


Richard Macdonald and Mallik recently asked about neural network applications
in chemistry.  I am not personally aware of anyone doing NMR spectral analysis
with neural nets but it certainly seems like a good application.

We (Eastman Chemical Company) have been a member of the University of
Tennessee's Measurement and Control Engineering Center for a number of years.
UT has been using neural nets for quite a while in high-speeed image analysis 
and more recently in the analysis of spectroscopic data.  The on-line process
Raman spectrometer, built by UT M&CEC and currently installed in our cellulose
esters acid concentration department, utilizes a neural net for its data
analysis.

Neural nets appear to do an excellent job of analyzing near-infrared spectra,
although the training problem is formidable.  Based on work by UT and other
researchers, it has been found that neural nets often do a far better job on
modeling nonlinear processes than do most conventional chemometric techniques
like PLS.  I am not personally familiar with the work but our Texas Eastman
facilities have done a considerable amount of work in chemical plant process
models with neural nets.

Recently Dr. Paul Gemperline of East Carolina University has joined the UT
Measurement & Control Center as part of a NSF tie-grant.  He is very active in
neural net applications to chemometrics.  His presentations and explanations
are excellent.  Even I can understand them!

For those interested, I suggest you contact Paul directly or Dr. Arlene
Garrison at the University of Tennessee.  She may be reached at 615/974-2375
(or garrison@utkvx.utk.edu).

As a side note, there was some earlier discussion on the network decrying
university researchers having to do industrial related research.  We have
found that working with the University of Tennessee has been quite beneficial
to both us in industry and to the academics too.  The real winners, though,
are the students!
						Barry
 -----------------
|  ___  ________  |       Dr. Barry L. Ornitz          WA4VZQ
| |  / /        | |       Eastman Kodak Company
| | / /         | |       Eastman Chemical Company Research Laboratories
| |< < K O D A K| |       Process Instrumentation Research Laboratory
| | \ \         | |       P. O. Box 1972, Building 167B
| |__\ \________| |       Kingsport, TN  37662       615/229-4904
|                 |       INTERNET:   ornitz@kodak.com
 -----------------

From harry@wpcs01.phys-chemie.uni-wuerzburg.de Fri Oct  9 04:03:29 1992
Date: Fri, 9 Oct 92 09:02:35 +0100
From: Harald Lanig <harry@phys-chemie.uni-wuerzburg.dbp.de>
To: chemistry-request@ccl.net
Subject: Potential Surfaces with MOPAC

Dear Netters,

coming back to the origin intension of this list:
>> DISCUSSING COMP_CHEM_PROBLEMS !!! <<
I will tell you my problem:

I am working on 1,1' coupled isoquionlines. As you can imagine, these
molecules can rotate about the central bond from one minima (the two
rings are almost perpendicular) over a transttion state to the 2nd.
minimun, wich seems to be a mirror image of the 1st. one.

I have calculated all three conformations (using MOPAC 6.0) and my idea
is to show this isomerisation by a 3-dim potential surface. This is
possiple using the GRID function of MOPAC, here I can define two angles etc.
to scan. If I am using a dihedral angle describing the position of the
two rings to each other and another angle (a-b-c) between the rings, I always
will get a jump in the energy, because the rings are hindering themselve
and, if the strain is great enough, the will snap!

Also a one-dim reaction-path using only the dihedral between the rings will
show this snap.

Can anybody tell me how to choose the right parameters to show this 
reaction without jumps. Or is this the wrong way visualizing this motion?

I hope a few people will answer me, perhaps on the list, helping me and
the list to come back to its real intention. I will be very sad (and angry)
if the list is closed, because this is a chance for people new to this
research starting contacts and scientific discussions !!

Thanks for reading (and answering) and forgive my terrible english.

-Harald   

  
---------------------------------------------------------------
Harald Lanig                 Institut fuer Physikalische Chemie
                             Marcusstr. 9-11
phone: +49(0)931-31-510      8700 Wuerzburg
  fax: +49(0)931-57-849      Germany
email: harry@phys-chemie.uni-wuerzburg.dbp.de
---------------------------------------------------------------