Summary of answers to QSAR question
Dear CCL'ers,
Following is a summary of the replies I received to my question about
how
long it should take to perform a QSAR analysis of 120 compounds. First
is
the original question and then come the answers. Thanks very much to all
who replied.
Dear CCL'ers'
I am about to start doing some QSAR on a set of 120 compounds. How
long should I expect it to take to overlay all 120 compounds and adjust
their
conformations, to achieve maximal overlap? A range of times or an
approximate number of hours will make a fine answer. Thanks alot.
Susan
gee, wouldn't that depend on what software and methods
you're using?
[7m[5m
___ _ http://ellington.pharmacy.arizona.edu/~bear O-topo-O
(___ |.) bear "-at-" .pharmacy.arizona.edu 5' : : .*
___)OARING |_)EAR, UA Pharmacy 404, Tucson 85721 |'*. .*'| |
Computer Modeling for Medicinal Chemistry; SAR CAMD | | *.,* | | |
Protein/DNA Structural Biology; Cancer Biochemistry 3',DNA helix|.*
Pharmacognosy, Herbs, Nutrition, Natural Dentistry '***' '**
[0m
It depends on a lot of things. If these are near analogs, where a
common
scaffold has the same numbering, a Sybyl spl (there is one called align,
or at
least I have one) will do a rough atom-to-atom alignment in a matter of
a minute
or so. If the structures are dissimilar, a manual use (in Sybyl) of the
orient
command will give you a very rough alignment in probably 30 or 40
minutes,
depending on how good you are with the keyboard. If you want to do
something
fancy, like alignments based on similarity of the ESP. . .this gets to
be time
consuming. If you are also taking into account multiple conformations,
or
tweaking the alignment given a first-pass CoMFA-QSAR. . .you just have
to try
it.
I infer from your address (prius.jnj) that you are with Johnson and
Johnson.
Debbie Loughney (sp?) is well versed in Sybyl. Have you discussed this
with
her?
Viel Vergnuegung!
Steve Bowlus
Hi there,
It depends on the machine power, qsar software, compound size, parameter
selecting .....Generally, as I am doing now for about 100 compounds(MW.
300) using a PowerMac8100 with CAChe Project leader, one month or 500
hrs
around is the common expense. Maybe I am lazy.
Any QSAR topic I am interesting in.
Yours,
Canping Pan
---------------------------------
E-mail: pan_canp "-at-" agr.kyushu-u.ac.jp
Tel: 092-611-7122(H)
092-642-2858(O)
Fax: 092-642-2864
http://133.5.200.42/pan.html
Hi Susan,
I am afraid your question is impossible to answer considering the
background information you supplied. None.
First you need a very well planned working strategy.
Do you now anything about the target protein. That is, do you now
anything about the ligand-target protein interactions?
How flexible are your compounds? You need to perform conformational
analyses on most of them.
You must align low energy conformations from all compounds. This is
the most crucial step in order to obtain a reliable 3D QSAR model.
Having all compounds aligned the most time consuming part is over and
the real fun begins.
Cheers
/jonas
My company offers two products that you may find VERY useful. The first
is
AMPAC, a semiempirical quantum mechanical package along with an easy to
use
graphical user interface. The second is CODESSA, and advanced QSAR/QSPR
program that does much of what you are currently doing manually
automatically
when paired with AMPAC. There are a number of reference papers in the
literature describing the sorts of correlations that are possible, and
one of
them (#1 below) sounds somewhat like what you want to do. If you are
interested and will send me your regular mail address, I'll be happy to
send
some propaganda in the way of brochures along. Also, perhaps we can
arrange a
demo if you think that you might be further intrigued. Let me know.
Best regards, Andy Holder
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
DR. ANDREW HOLDER
President
Semichem, Inc. || Internet Addr: aholder "-at-" cctr.umkc.edu
7128 Summit || Phone Number: (913) 268-3271
Shawnee, KS, 66216 || FAX Number: (913) 268-3445
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Dear CCL'ers,
You have all rightly pointed out that I did not provide sufficient
background
information for an answer to my question. I will provide it now (I
hope):
I am doing everything manually, from the overlays to the
conformational
searches. The compounds I am aligning fall into about half a dozen
structural categories (by my reckoning), within which there is not great
structural variation but among which there is little similarity. There
are
large flat fused ring structures and spiro compounds, single phenyl
rings
with long aliphatic chains on either side (in some cases branched, in
others, not) and pairs and larger sets of phenyl and non-aromatic rings
connected by
aliphatic chains and other moieties.
I hope this clarifies the earlier question somewhat. Thanks for your
replies,
both future and those already sent.
Susan
-------This is added Automatically by the Software--------
-- Original Sender Envelope Address: SMILLER "-at-" RARUSRAEXS1.prius.jnj.com
-- Original Sender From: Address: SMILLER "-at-" RARUSRAEXS1.prius.jnj.com
CHEMISTRY "-at-" www.ccl.net: Everybody | CHEMISTRY-REQUEST "-at-" www.ccl.net:
Coordinator
MAILSERV "-at-" www.ccl.net: HELP CHEMISTRY or HELP SEARCH | Gopher:
www.ccl.net 73
Anon. ftp: www.ccl.net | CHEMISTRY-SEARCH "-at-" www.ccl.net -- archive
search
Web: http://www.ccl.net/chemistry.html
Dear Susan,
Please, post a summary with your answers. Which software are you using
to do the QSAR study?
Thanks,
Ezequiel
If you know the corresponding points then this is pretty easy. The
cheapest way is to do ensemble distance geometry. I would use the most
constrained analogue of each class. Generate 100 ensembles with that
contain the chosen member of each class. Minimize the structures. Then
see which ensembles (1) still match & how well and (2) contain mainly
low energy conformations. You will discover either that you have only
a few solutions (good) or 50 or so different solutions which will tell
you that you will never have a unique choice of conformations so you
might as well just pick one set and move on. Ensemble distance geometry
is available in DGEOM from QCPE.
Sheridan, R. P.; Nilakantan, R.; Dixon, J. S.; Venkataraghavan, R., J.
Med. Chem. 1986, 29, 899-906. "The Ensemble Approach to Distance
Geometry: Application to the Nicotinic Pharmacophore".
If you don't know what atoms correspond in the different structures,
you can run a bunch of DGEOM runs to see which give solutions. Or,
you can use DISCO, an option with Sybyl.
Martin, Y. C.; Bures, M. G.; Danaher, E. A.; DeLazzer, J.; Lico, I.;
Pavlik, P. A., J. Comput.-Aided Mol. Design 1993, 7, 83-102. "A Fast
New Approach to Pharmacophore Mapping and its Application to
Dopaminergic and Benzodiazepine Agonists".
It shouldn't take you a month to do what you want, except that you are
apparently learning in the process with no good tutor. How do you plan
to do the QSAR?
Yvonne Martin, Senior Project Leader
Computer Assisted Molecular Design Project
D-47E, AP10 2fl
Abbott Laboratories
100 Abbott Park Road
Abbott Park, IL 60064-3500
Phone: 847 937-5362
FAX: 847 937-2625
yvonne.c.martin "-at-" abbott.com
Susan,
For the situation you have described, I highly recommend SEAL, an
automatic
alignment program, available from QCPE for a small fee. Forget manual
alignments - it will take forever! With SEAL, you can align all 120 of
your
molecules (even if you have ~20 conformations of each) in <= a few
hours.
Good luck.
Mary
/***************************
* *
* Mary Bradley *
* Senior Research Scientist*
* Rhone-Poulenc Ag Company *
* *
* bradleym "-at-" mindspring.com *
* *
***************************/
***********************************
Mary Bradley *
email: mbradley "-at-" mindspring.com *
fone & fax: (919) 572-6599 *
***********************************
Dear Suzan
I don't know what you mean by hand... Could you tell me what are the
modelling softwares available in you lab ?
I see two problems to face : Conformer generation (it seems that you
have very flexible compounds) and conformer fitting.
I would suggest you to start with the more rigid compounds, do a
conformational search (systematic, random, high temperature dynamic as
you want or as you can). Keep structures in a range of 10-20 kCal above
the minima (don't forget to investigate ring conformation).
Then make hypothesis on what would be the pharmacophore (charged atom,
hydrogen donnors or acceptors, aromatics, lipophilic groups etc...) or
any common points available in your series or reported in literature.
Then measure some key distances between these elements and try to
cluster the molecules according to these parameters (with SAS) or make
some sorts with excel to identify the common pharmacophore if you don't
have any specific softwares to do that.
Once you have identified the best clusters then you can superimpose the
best conformations.
Concerning the more flexible molecules, don't spend time to generate
conformers but simply check that you can simply superimpose one
reasonable conformation to the previously identified pharmacophore.
To my opinion, and without any additional information, you have to plan
some months to do the job.
Hope this helps
Don't hesitate to contact me for help
Francois
------------------------------------------------------------------------
--
Dr. Francois CROIZET - Riom Laboratoires CERM - Chemistry Dept. (CMC) -
ZI de la Varenne, Rue H. Goudier, BP 140 - 63203 RIOM Cedex - FRANCE -
Phone: (33) 473 33 49 70 - Fax: (33) 473 33 49 97
E-mail: f.croizet "-at-" organon.akzonobel.nl
========================================================================
==
For automated overlay of two molecule with conformational flexibility on
PC, PowerFit may be useful. You can go to our web site to download a
fully functional program to evaluate. Currently only Windows 95 and NT
version is available for download.
For automated overlay of two molecule with conformational flexibility on
PC, PowerFit may be useful. You can go to our web site to download a
fully functional program to evaluate. Currently only Windows 95 and NT
version is available for download.
Ms. (Dr.?) Miller:
For the past few days, I have been following with interest
your postings to the CCL concerning QSAR. (BTW, I haven't
seen (most of) the responses to you show up in my mail, so I'm
assuming they're going directly to you.) Since I do molecular
modeling and algorithm design, I haven't felt competent to
comment on your questions. After your last posting, however,
I feel compelled to register extreme surprise (and to stick
my nose in where it doesn't belong :-).
Why, pray tell, would your mentor even suggest doing _all_
of the overlays by hand??? There is certainly value to
doing a small, random set of overlays by hand, just to
confirm that the results make sense. However, if you're
needing to do anything more than a couple of handfuls of
compounds, noninteractive fitting seems the only way to
go. Otherwise, you end up wasting enormous amounts of
human time doing things a mindless computer was invented
for. On the other hand, if you're using Tripos software,
I guess noninteractive isn't really a choice anyways?
Ah well, good luck with your project. I hope you're able
to finish in something approximating a reasonable amount
of time.
Sorry for butting into the middle of your more substantive
discussion with people who actually do QSAR...
Marti Head
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Martha S. Head, Ph.D. (Marti)
Center for Advanced Research in Biotechnology
9600 Gudelsky Drive
Rockville MD 20850
voice: (301)738-6104
email: mhead "-at-" indigo18.carb.nist.gov
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Hi all,
Thanks again for all of your responses. I will post a summary in a
couple
of days when the discussion is over. All of the responses I've received
so
far have come to CCL and not to my personal mailbox, but if it is
desired,
I can still post a compilation of the responses.
I forgot to mention 2 very important facts:
1) I am using SYBYL and have no access to anything else, save the
free software on the net. I cannot use that however, for the
second
reason.
2) My mentor insists that I do the overlays by hand, that is, by
using
the fit module in SYBYL and picking points, with the mouse, to
fit.
I must also do torsion adjustments by hand, that is, by choosing
a torsion angle for each bond and applying it, using either the
twist
widget in the upper left corner of the screen or the modify
torsion
command.
Susan
-------This is added Automatically by the Software--------
-- Original Sender Envelope Address: SMILLER "-at-" RARUSRAEXS1.prius.jnj.com
-- Original Sender From: Address: SMILLER "-at-" RARUSRAEXS1.prius.jnj.com
CHEMISTRY "-at-" www.ccl.net: Everybody | CHEMISTRY-REQUEST "-at-" www.ccl.net:
Coordinator
MAILSERV "-at-" www.ccl.net: HELP CHEMISTRY or HELP SEARCH | Gopher:
www.ccl.net 73
Anon. ftp: www.ccl.net | CHEMISTRY-SEARCH "-at-" www.ccl.net -- archive
search
Web: http://www.ccl.net/chemistry.html
Susan,
The superimposition procedure is performed due to that the 3D
structure of the target protein is not known. You seek atoms or
mutual points on your ligands that are most likely to interact with
target protein. However, very often your fits will be much better if
you also allow extension vectors to act as a fit point. For example,
a basic nitrogen may interact through its lone pair of electrons or,
if it is protonated, through the proton. Therefore, if you apply
extension vectors in the direction of the lone pair of electrons
, and use this vector as a fit point, much more reliable fits will
occur. You use a UNIX machine and I suggest that you use Apollo or a
similar program for this. Even SYBYL has a module.
Your mentor does not allow you to perform conformational searches on
your ligands? Then why bother at all. This is, one of the most
crucial steps in 3D QSAR, especially if your ligands are flexible.
In SYBYL I suggest you to use the random search procedure!
The conformational search and the alignment proecedures are tedious
and cost a lot of time, but must be performed rationally and
thoroughly.
Succes
/jonas
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Jonas Nilsson e-mail: j.nilsson "-at-" farm.rug.nl
Univ. Centre for Pharmacy tel: +31(0)503633302
Dept. of Med. Chemistry fax: +31(0)503636908
Antonius Deusinglaan 1
NL-9713 AV Groningen
The Netherlands
http://pc131.farm.rug.nl/main.htm
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
(In reply to your message dated Tuesday 11, March 1997)
--
Something of interest for QSAR work.
Accord for Excel can be used for structure relationships and Activities.
Check out
WWW.synopsys.co.uk
Take care, all.
Susan