|
[ Home ] [ About/Contacts ] [ Meetings ] [ Members ] [ News ] [ Positions ] [ Resources ]
--------------------------------------------------------------------------------------------------------------
The QSAR and Modelling Society Newsletter
Issue No.8, October 1997
The QSAR and Modelling Society
Chair: Hugo Kubinyi
Officers: Yvonne C. Martin (advisor to the chair), James King
(treasurer) Han van de Waterbeemd (secretary/editor)
Board: John Block, Sergio Clementi, John Dearden, Bill Dunn,
Marvin Charton, Ferenc Darvas, Rainer Franke, Toshio Fujita, Peter Goodford,
Phil Magee, Jim McFarland, Oleg Raevsky, Joachim Seydel, Bernard Testa, Milon
Tichy
Honorary chair: Corwin Hansch Past chair: Phil Magee
Editorial
A long time passed by since I wrote my last Editorial on the
'Quiet Society' and at the end of this article I expressed my hope that in my
next Editorial I would be able to report on a 'Lively Society'. However, not too
much happened since then and again we had problems to collect enough
contributions from our members, in order to publish the Newsletter on a more
regular basis. One of the most important news with respect to the Newsletter is
that our Secretary Han van de Waterbeemd, decided to change his position. From
October 1st, 1997, he is now a colleague of Dennis Smith, at Pfizer Central
Research, Sandwich, Kent, UK (see below). Thus, we should not only be grateful
to him that he continues in his duties but also to his new company for
supporting these important activities. On this occasion, I would like to thank
also Dr. Didier Rognan ETH Zurich, Switzerland, for his ongoing activities to
maintain our Web pages, and Prof. Gerd Folkers, same institution, for providing
the server.
At the 1997 Gordon Research Conference 'Quantitative
Structure-Activity Relationships' in Tilton, NH, U.S.A., a members meeting of
our Society took place, where I presented a short report on the Society
activities during the last year. Our Society is still growing, having now more
than 560 members (effective September 1997). Members seem to be happy with their
Society. At least we did not receive many complaints (two members who had sent
their money but no application forms complained - after they were registered,
they received the membership list and the last Newsletter).
In addition to the UK QSAR discussion group which exists since several years,
there is now a Russian local group (see Newsletter 7) and an Italian group,
which formed in May 1997 (Chair: Sergio Clementi). Some 'inactive' members
(people not paying their membership fees over years) should follow the shining
example of the Russian group who collected 300 US-$ from their 30 members and
sent it to our treasurer! At the members meeting in Tilton, Tudor Oprea informed
us about the fate of Prof. Zeno Simon, one of the best-known QSAR scientists all
over the world. Without reasons (or at least for reasons difficult to understand
and to accept), he has been fired from his position at the university. Tudor
Oprea also informed us that such unmotivated decisions hit other scientists too,
including himself. After a discussion on the pros and cons of an official
complaint a committee of three persons formed (Ferenc Darvas, John Block and
John Dearden) who will come up with proposals how to react. Letters to Nature
and Science seem to be appropriate but some more information should be collected
before. Some other issues (QSAR journal, mailbox, addresses in the Internet) are
discussed below in detail.
Hugo Kubinyi
Information
of the Secretary
We are a still growing Society and now count 550 members. I would like to
thank Dora Schnur for distribution of the Newsletter in the US and Pharmacopeia
for their financial and logistic support. Carol Manners of Astra Charnwood is
taking care of distribution in the UK. If other members would like to take the
responsability for local distribution, please contact me. It would be of great
help to set up a network. Warm thanks to Pfizer Central Research, UK for
printing of this Newsletter.
Please send your contributions to this Newsletter or to the HomePage to my
new address:
| Dr. Han van de Waterbeemd |
|
Pfizer Central Research Dept. Drug Metabolism
Sandwich, Kent CT13 9NJ, UK |
|
Mailbox
Mailbox of the QSAR and Modelling Society
Some members proposed that we should have a mailbox where we interchange
questions, information and comments within the members (compare also the
proposal to form an Internet discussion group, Newsletter 7). This proposal has
been accepted by the members meeting at the 1997 QSAR GRC and Jim McFarland
agreed to install this mailbox and to maintain it. Please, send whatever you
consider interestingly enough to be distributed among the Society members (also
Meetings and Symposia reports would be highly welcome), to
Dr. James W. McFarland
reckon.dat consulting
Experimental Design and Data Analysis for the Pharmaceutical, Agrochemical and
Chemical Processes Industries
24-2 Burr Road, Old Lyme, Connecticut 06371, U.S.A
Phone/Fax +1-(860)-434 5125
Phone +1-(860)-434 2271
e-mail reckon.dat@ibm.net (e-mail
address is Jim's, not yet the Society mailbox!)
Society's Homepage
The best source for current information is our Web HomePage. You are
encouraged to participate actively in improving and updating this site by
sending us information and suggestions.
 |
Addresses of Members in the Internet.
We already have the e-mail addresses of our members posted in the WEB
pages. As the Medicinal Chemistry Section of the American Chemical Society
lists full addresses of their members, we will do this also, by the end of
the year. If you don't like to have your address, phone and fax number
posted in the Web, please inform us and we will take care that only your
e-mail address is provided. |
Hugo Kubinyi & Han van de Waterbeemd
|
Data sets.
Send your data sets for distribution now!! Science proceeds only by
continually checking and refining (observations and) models. As the premier
society for QUANTITATIVE modeling of biological activity, we have an
obligation to provide our structure-activity data for others to use to test
their new and improved methods. At last count, there were only sixteen data
sets on our web page. I know that many of you freely send them to others,
why not make them available to others? It is easier for you to provide them
to the web page once than to continually hunt for the data set as people
ask. Others will be glad you did it if you move to another institution with
subsequent loss of easily traceable address. If you send a data set, please
copy the e-mail to me so that I can personally send you a thank-you. This is
as important to the field as publishing that original model!! |
One of the latest additions to our site is a place to make reference data
sets available to the community. When you would like to submit a data set,
here is the format (ASCII file):
Name Data Set: (eg. The Selwood data set)
What: n Properties, m Compounds Source: (person who sent the data set)
Original Reference: (paper)
Remarks: (modified data)
Column Labels :
The Data :
Milano Chemometrics
The address is http://www.disat.unimi.it/chm/
.This is a site about chemometrics and QSAR, with several links to other
sites (obviously, also to QSAR and Modelling Society). Into the section
"Thoughts on Chemometrics" you can find some new descriptors
proposed by our group: these are available for use. If you have information
about meetings, QSAR announcements and other QSAR sites, please send me an
email.
Roberto Todeschini (tode@alpha.disat.unimi.it)
Dipartimento di Scienze Ambientali
Via Emanueli 15 - I-20126 Milano, Italy
QSAR Web Server
I am pleased to announce the availability of the QSAR Web server developed
in our laboratory. Since all applications are written in JAVA, you will need
to use an adequate browser. The server has the following URL: http://mmlin1.pha.unc.edu/~jin/QSAR/
Since we continue to develop this server, we would greatly appreciate your
comments.
Alex Tropsha (tropsha@gibbs.oit.unc.edu)
Laboratory for Molecular Modeling, School of Pharmacy, University of North
Carolina, Chapel Hill, NC 27599-7360, USA. FAX +1-919-966 6919
Meeting
Reports
Comparing Liposomes and Octanol as Model Membrane for log P (QSAR
meeting in Osaka, November, 1996)
The use of liposomes rather than octanol as a model membrane is a subject
of increasing interest. At November's 24th Japanese QSAR meeting in Osaka,
Japan, John Comer of Sirius Analyrical Instruments Ltd. (Sussex, UK) described
how his team has recently measured log Pmem values for a range of monoprotic
acids and bases in both octanol and large unilamellar liposomes made from
dioleylphosphatidylcholine (DOPC). While logP values for the uncharged species
were similar in octanol and liposome, values for the ionized species were
significantly higher in liposome, reflecting the interaction between the
charged molecule and the membrane head-group. The pH-metric method used in
this study does not require the use of radio-labelled molecules.
John Comer
More information available from Sirius at http://www.sirius-analytical.com
The Gordon Research Conference ' QUANTITATIVE STRUCTURE-ACTIVITY
RELATIONSHIPS' (Tilton, NH, USA, August 1997)
In the first week of August, the 11th Gordon Research Conference on 'QUANTI-TATIVE
STRUCTURE-ACTIVITY RELATIONSHIPS' took place in its traditional location in
Tilton, New Hampshire. About 130 scientists, mostly from the US, but also from
different European countries, as well as from Japan, Brazil, etc., met to
discuss new developments in QSAR and QSAR-related modelling approaches. As for
all the conferences before, the number of applicants was much larger than the
number of people who could be accepted. Hershel Weintraub was the chair and
Gerry Maggiora was responsible for the scientific program. Overall the
lectures can be rated as good to excellent. There were 9 sessions (from Sunday
evening till Thursday evening) with a total of 21 lectures (some key
references are given in addition to the titles):
Session 'Inverse QSAR and QSPR' (Chair: Hugo Kubinyi)
Venkat Venkatasubramanian, Purdue University, West Lafayette, IN 'GENESYS:
An Evolutionary Hybrid Framework for Computer-Aided Products Design'
Dimitris K. Agrafiotis, 3-Dimensional Pharmaceuticals, Exton, PA
'Systems and Methods for Designing Combinatorial Chemistry Experiments'
Session 'Pharmacophoric Patterns' (Chair: Vincent van Geerestein)
Jordi Mestres, University of Girona, Catalonia (Spain)
'Molecular-Field-Based Similarity in Drug Design' (Program MIMIC; J.
Comput. Chem. 18, 934 (1997))
Jon Mason, Rhone Poulenc Rorer, Collegeville, PA '3D Pharmacophores for
Molecular Similarity and Diversity Applications'
Hugo Villar, Terrapin Technologies, San Francisco, CA 'Affinity
Fingerprints: Applications and Implications for Lead Generation'
Session 'Property Prediction' (Chair: Kate Holloway)
Peter Jurs, Penn State University, University Park, PA 'Prediction of
Chemical Properties of Organic Compounds from Molecular Structure'
Philip Howard, Syracuse Research Corporation, Syracuse, NY
'Comprehensive QSARs for Environmental Physical Properties, Transport, and
Degradation'
Session 'Structure-Activity Relationships' (Chair: John Dearden)
Christopher Lipinski, Pfizer, Groton, CT 'Drug Oral Absorption
Problems: Computational Alerts to Potential Poor Solubility and
Permeability'
William Dunn, University of Illinois, Chicago, IL 'A Solution to the 3D
QSAR-Problem of Treating Flexible Molecules which Can Assume Multiple
Receptor Alignments'
Alexander Tropsha, University of North Carolina, Chapel Hill, NC 'Unity
in Diversity: From QSAR to Combinatorial Chemistry'
Session 'New Trends I' (Chair: Jeff Blaney)
Andrew Rusinko, Glaxo Wellcome, Research Triangle Park, NC 'SCAM:
Statistical Classification of Molecules Using Recursive Partitioning'
John N. Weinstein, National Cancer Institute, Bethesda, MD
'Information-Intensive Drug Discovery for Cancer: QSAR Analysis of
p53-Inverse Agents'
Session 'Membrane-Associated Phenomena' (Chair: Peter Goodford)
Paul H. J. Nederkoorn, Vrije Universiteit Amsterdam (now Shell,
Amsterdam, The Netherlands) 'Signal Tranduction via G Protein-Coupled
Receptors: Ternary Complexes as Secondary Proton Pumps and GTP Synthases'
(compare Trends Pharmacol. Sci. 16, 156-61 (1995) and in M. G. Ford et
al., Bioactive Compound Design: Possibilities for Industrial Use, 1996,
pp. 77-87 and in Computer-Assisted Lead Finding and Optimization, pp.
513-526).
Terry R. Stouch, Bristol-Myers Squibb, Princeton, NJ 'Toward an
Atomic-Level Understanding of Membrane Affinity: Molecular Dynamics of
Solute/Membrane Interactions'
As Leo Herbette was unable to come, the audience got the chance to hear
an excellent lecture on Lipophilicity, Acidity Constants and Membrane
Transport, presented by Alex Avdeef, Pion Inc., Cambridge, MA.
Session 'Toxicology Prediction' (Chair: Mark Johnson)
Douglas Bristol, Laboratory of Environmental Carcinogenesis, NIH/NIEHS,
Research Triangle Park, NC 'Approaches to Predictive Toxicology and
Toxicoinformatics'
Gilda Loew, Molecular Research Institute, Palo Alto, CA 'Computer-Aided
Assessment of Drug Toxicity Caused by Cytochrmome P450 Metabolism and
Inhibition'
Session 'New Trends II' (Chair: Lisa Balbes)
Paul Mezey, University of Saskatchewan, Saskatoon, Canada 'Accurate
Electron Densities of Large Molecules: Applications to QShAR' (Sh in QShAR
= Shape, compare P. Mezey, Shape in Chemistry. An Introduction to
Molecular Shape and Topology, VCH, 1993)
Tim Clark, Computer-Chemie-Centrum, University of Erlangen, Germany
'Quantum QSAR - Docking, Pharmacophores, and Similarity'
Don Abraham, Virginia Commonwealth University ' Intrinsic Activity at
the Molecular Level'
Evening 'Wrap Up' Session (Chairs: Hershel Weintraub and Gerry
Maggiora)
Robert S. Pearlman, Texas University, Austin, TX (after dinner speach)
'Challenges in Computer-Aided Molecular Design'
In addition, there were about 50 posters, organized in 4 sessions (Chairs
Phil Magee, Paul Seybold, Tudor Oprea and Carol Venanzi). It was an exciting
conference with a very stimulating atmosphere. The lively exchange of
scientific information was further supported by very active and engaged
discussions after all lectures. Having participated in all QSAR Gordon
Conferences (with the exception of the very first one, in 1975), I have
never experienced such an active discussion before.
Kate Holloway, Merck Research Laboratories, was elected by the audience
to be the program chair for the next QSAR Gordon Conference, in 1999. If you
would want to help her in this duty by giving advice or making proposals for
topics which should be included, etc., send a mail to
Dr. M. Katherine Holloway
Departments of Molecular Systems Building 42-2 Merck Research Laboratories
West Point, PA 19486 U.S.A.
FAX +1-(215)-652 6913
e-mail kate_holloway@merck.com
Hugo Kubinyi
Opinions
When will we change from semi-empirical to ab initio quantum methods
in QSAR analysis?
Due to the increased availability and capacity of PC's and the sinking
costs of computer workstations with Alpha, Sun or Risk type of processors,
calculations on large biologically active molecules by ab initio methods
have become feasible. Recent results obtained with the GAUSSIAN_94 program
on a large DNA fragment demonstrate that we can answer the title question
with "Yes, time has come". Nevertheless, there is also an increase
in the use of semi-empirical schemes. Reasons include:
- The interpretation of ab initio calculations is very difficult,
particularly for use in QSAR studies. We have called this "crisis of
interdisciplinary competence". - Errors in biological experiments,
particularly in vivo, are in the range of 15-20%, which is more than in the
ab initio calculations. Here such calculations would be coined by the
expression "hammer a nail with microscope help". - For serious
QSAR studies of many compounds consisting of hundreds of atoms, ab initio
calculations even on Alpha/128MRAM workstations would require 1-3 days of
computing time. This begins to exceed time to perform biological
investigations. - Since semi-empirical methods have been parametrized for
the reproduction of experimental data, the often give more correct results
compared to ab initio studies.
Since prices of hard- and software are falling rapidly, we think that we
do not have to wait another 10-15 years before semi-empirical methods will
be replaced by reliable ab initio techniques.
Andrew Pogrebniak, January 1997
Contributions
| QSAR Tips from BIOSAR Research, by Philip S. Magee (philm1@ix.netcom.com)
Factoring of logP for greater Insight.
LogP(octanol/water) is a composite descriptor describing the
difference in free energy (-2.303*RT*LogP) of developing a minimum
energy cavity in each phase. It is an extrathermodynamic property as
neither phase is pure even at infinite dilution. In addition, it is more
constitutive than additive. The early assumptions of additivity were the
result of measuring rather simple structures. The growth of the CLOGP
program over the years derives from the application of LogP measurements
to drugs and chemicals of increasing complexity. As in single phase
solvation, the intermolecular forces involved in LogP (differential
solvation) are polarizability (MR), hydrogen-bonding (HBA, HBD),
inductive and steric effects as in Charton's IMF model. Overlaid on
these familiar forces is a non-computational factor that prevents LogP
from being accurately calculable for complex structures, namely, the
difference in conformational energy in each phase including entropic
effects. There are now thousands of LogP correlations in the literature,
some good, some not so good. In each case, the assumption is made that
LogP(octanol/water) is an adequate model for whatever partitioning is
under study. The essence of this tip is that this is not a good
assumption when the partioning phase is very different from octanol. As
LogP is calculable, it is easy to factor the lipophilic and hydrophilic
sub-structures such that PL + PH = LogP. This is basically no different
from factoring Hammett's sigma into inductive and resonance components.
The factoring is statistically harmless and when octanol/water is a good
model, the loadings of PL and PH will be nearly identical and no
additional information or improvement in correlation will be obtained.
In many cases, however, solvent selection will favor either the
lipophilic (PL) or hydrophilic (PH) factors and it will be necessary to
add MR, HBA and HBD to correct for the imbalance of intermolecular
forces. These factored cases can be dramatically improved statistically
and lead to conclusions that are not accessible with LogP alone.
Philip S. Magee, In: QSAR in Environmental Toxicology - IV (J.L.M.
Hermens and A. Opperhuizen, eds.). Elsevier, Amsterdam, pp. 155-178. (or
request a copy).
Atom-by-Atom (positional) Analysis of Hypermolecule Sets.
The development of hypermolecules (lowest common structure) for sets
of compounds is an old story. When one studies the hypermolecule in
conjunction with the biodata, it becomes possible to develope an atomic
level picture for each position in terms of "favorable",
"unfavorable" or "indifferent". These qualitative
SAR's have provided valuable information for both physical organic and
synthetic chemists. Some years ago, I quantified this process by
redefining "favorable" as binding, "unfavorable" as
sterically repulsive, and "indifferent" as not in contact for
active-site binding events. Prior to this, studies on the energetics of
binding in adsorption chromatography led to the concept that most
binding events were several-point landings rather than molecular events.
This also proved true in hypermolecule studies of acetylcholinesterase
inhibition where only 3 of 11 hypermolecule sites were strongly involved
in the event. The overall protocol for setting up the matrix is
described in the reference as well as the statistical technique for
handling a matrix of this size. However, the positional descriptors can
be described here. Electronegativity was found to be inversely related
to sigma charge and was therefore used for the electronic effect. Rv is
used for steric interactions and atomic MR for polarizable volume. The
weakest descriptor is the LogP related f value of the atom. This needs
to be improved as adjacency effects are not addressed and because f,
like LogP, is a composite descriptor for the polar atoms. Other than
these reservations, the technique has performed very well and may be of
some general interest for describing active site bindings.
Philip S. Magee, Quant. Struct.-Act. Relat. 9, 202-215 (1990)
Philip S. Magee, In, QSAR: Rational Approaches to the Design of
Bioactive Compounds, C. Silipo and A. Vittoria, eds., Elsevier,
Amsterdam, 1991, 549-552.
Heat of Formation at a Sensitive Indicator of Molecular Mechanics
Modeling.
As a first experiment, take about 6 diverse structures, model by AM1
and record the heat of formation from the log file. Now select 2-3 of
your molecular mechanics programs, reminimize the structures and then
determine the resulting heat of formation by single-point AM1. I suspect
that you will be surprised, as I was, by how far from the electronic
minimum of AM1 some of these structures are. In my experiments, I found
deviations of more than 20 kcal/mol for some of the more flexible
compounds. This was important to me as, in a research contract from a
company, I was trying to model the head groups of some dermal drug
permeation enhancers (detergent-like). None of the more complex cases
would converge by AM1 and these were minimized with MM2 followed by AM1
single-point to determine the Hf for use as a descriptor. The work was
abandoned after realizing the sensitivity of Hf to "minimum"
conformation. However, there is some merit in the procedure. In a recent
study of a cyclic pentapeptide, the structure was modeled by Charmm,
OPLS, and MM3, followed by AM1 single point to assess the Hf. Although
the "true" electronic minimum was never determined by AM1
(failed to converge), it was very clear that MM3 (lowest Hf) strongly
outperformed the other two models for this particular case. This method
depends strongly on your point of view and the goals of your problem. I
dont consider it to be publishable, just interesting and of possible
use. I would be very pleased with any experimental feedback or comments.
Two-Value Regression as an Alternate to Linear Discriminant
Analysis
For non-embedded cases (separable) where the biodata are either
affirmative (=1) or negative (=0), linear discriminant analysis is
normally the method of choice. This type of problem turns up in medical
assessments where, for example, a compound is either an allergen or a
nonallergen. There is an alternate method of analysis, namely, two-value
regression and there are some interesting advantages over LDA such as
standard statistical measures (Student T values, rsq., s, F etc.) and
the ability to plot class (1/0) against Y estimate. The method was first
used for QSAR by Yvonne Martin et al. and much later by Magee et al. For
a balanced set (n[0] = n[1]), the descriptors and coefficients are
identical to those of LDA. LDA places an n-dimensional plane between the
sets, while two-value regression defines a correlation line between the
two centroids. The only drawbacks are the need for a balanced set and
the inability to separate more than two classes.
Y.C. Martin et al., J. Med. Chem. 17, 409-413 (1974). P.S. Magee
et al., Quant. Struct.-Act. Relat. 13, 22-33 (1994).
Estimation of Descriptors for LFE and QSAR
A comprehensive set of equations for estimating missing sigma, steric
and bulk values from related descriptors was published in 1992 by P.S.
Magee. Thus, sigma(+ or -) can be estimated from sigma(m&p), Upsilon
from Es and various bulk descriptors from each other. These relations
are useful for providing good "quantitative" estimates of
missing values. Back in the early 70's, I experimented with deliberate
descriptor errors using a random number generator and a strong QSAR
equation in Pi, Pisq and Sigma (n = 18). For each compound in the set, I
varied each descriptor by +0.2, -0.2 or no change for random numbers:
0-33, 34-66, 67-100. In over 100 runs, it was found that more than 40%
errors were necessary to significantly alter the equations. The reason
is simple. If each descriptor accounts for about 30% of the variance,
then a few "errors" in a few of the 18 compound descriptors
will have little effect on the relation. This statistical experiment
convinced me that it is OK to guestimate missing descriptors and, in
fact, better than eliminating compounds from the set. In the last 20
years, I have guestimated many descriptors and have never generated an
outlier. All that is required is to get the sign right and the magnitude
approximately right. The most difficult to guestimate are steric
descriptors and this generally requires some thought. First select a
known group you believe to be slightly larger than the unknown, then one
you believe to be slightly smaller. Assign the midpoint of these numbers
to the unknown group. I think you will usually find the equation to be
stronger than if the compound were deleted. As for the referees, well
that's another problem.
P.S. Magee, In "Rational Approaches to Structure, Activity
and Ecotoxicology of Agrochemicals", ed. by W. Draber and T.
Fujita, CRC Press, 1992, Chapter 3
Mechanistic Insights from Toxicity Data
Jim King and I tackled a very difficult problem under a U.S. Army
contract. The object was to explore published toxicity data from a
mechanistic point of view and evaluate the differences among different
animals and different routes of administration. On the plus side, there
was a lot of data for most of the chemical classes in RTECS and Sax's
compilation. However, it was of borderline precision and known to be
somewhat biased toward the most toxic value reported. In a single lab,
the strain, sex, weight and age of an animal was well-defined, but
between labs a mouse could have almost any origin and characteristics.
In addition, the techniques and vehicles of the different investigators
were not standardized. As one example, our substituted phenol set (n =
50) came from 17 different labs with many animal,vehicle and other
experimental variables. This might suggest a need for 16 indicator
variables to achieve a reasonable correlation. However, in this and many
other cases, the mechanistic factors dominate provided one allows about
10 compounds per descriptor. By using this ratio and not attempting to
push the data too far, we were able to distinguish mechanistic details
(transport, electronic, steric) among different animals and routes of
administration. In some cases, transport was rate determining and only
logP or Pi was significant. In others, only the binding and reactivity
factors, sigma and upsilon, were significant. There were substantial
differences between animals and, within a given animal, between the
routes of administration (ip, iv, oral, dermal). In general, we found
that interlab data with differences in the animals, vehicles and
techiques to be analyzable at about 10 compounds per descriptor.
P.S. Magee and J.W. King, In "Probing Bioactive
Mechanisms", ed. by P.S. Magee, D.R. Henry and J.H. Block, ACS
Symposium Series 413, 1989, Chapter 24.
 | Articles worth pointing out , by Yvonne Martin |
-
First, an article that was the basis of an interesting
presentation at the QSAR Gordon Conference. It really illustrates
the power of simple logP calculations.
Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J.,
"Experimental and computational approaches to estimate solubility
and permeability in drug discovery and development settings,"
Advanced Drug Delivery Reviews 1997, 23, 3-25.
-
Five of a number of papers related to multivariate statistical
analysis. These are interesting in that they are in a journal that
QSAR folks might not read.
(1) Goutis, C., "A fast method to compute orthogonal loadings
partial least-squares," J. Chemometrics 1997, 11, 33-38.
(2) Denham, M. C., "Prediction intervals in partial
least-squares," J. Chemometrics 1997, 11, 39-52.
(3) Dayal, B. S.; Macgregor, J. F., "Improved pls
algorithms," J. Chemometrics 1997, 11, 73-85.
(4) Faber, K.; Kowalski, B. R., "Propagation of measurement
errors for the validation of predictions obtained by principal
component regression and partial least-squares," J. Chemometrics
1997, 11, 181-238.
(5) Berglund, A.; Wold, S., "INLR, implicit nonlinear latent
variable regression," J. Chemometrics 1997, 11, 141-156.
-
Four new logP calculation papers appeared. Are we seeing
improvements in the methods or an asymptotic approach to the
measurement error?
(1) Wang, R. X.; Fu, Y.; Lai, L. H., "A new atom-additive
method for calculating partition-coefficients," J. Chem. Inf.
Computer Sci. 1997, 37, 615-621.
(2) Bodor, n.; Buchwald, P., "Molecular Size Based Approach to
Estimate Partition Properties for Organic Solutes," Journal of
Physical Chemistry B. 1997, 101, 3404-3412.
(3) Breindl, A.; Beck, B.; Clark, T.; Glen, R. C., "Prediction
of the n-octanol/water partition-coefficient, logP, using a
combination of semiempirical MO-calculations and a
neural-network," Journal Of Molecular Modeling 1997, 3, 142-155.
(4) Haeberlein, M.; Brinck, T., "Prediction of water-octanol
partition-coefficients using theoretical descriptors derived from the
molecular-surface area and the electrostatic potential," Journal
Of The Chemical Society Perkin Transactions 1997, 289-294.
Glossaries
|
A glossary of terms used in Computational
Drug Design appeared in Pure & Appl.Chem.69 (1997) 1137-1152
and was prepared for IUPAC by H. van de Waterbeemd, R. Carter, G.
Grassy, H. Kubinyi, Y. Martin, M. Tute and P. Willett. |
|
Another hyperglossary on computational chemistry was prepared by
Jan Tollenaere and Ed Moret at the University of Utrecht in The
Netherlands (http://cmcind.far.ruu.nl/webcmc/glossary.html). |
Software
PASS, a Program for the Prediction of Activity Spectra from
Molecular Structure.
Using the computer system PASS (Prediction of Activity Spectra for
Substance) we have already found new leads with antiulcer, antitumor and
antiamnestic activity, and discovered new mechanism of action for some
compounds with known effect [1-5]. Why the PASS might be useful for you?
PASS helps you in the following:
- To find the most probable new leads with the required activity
spectra among the compounds in the corporate and commercial data bases;
- To reveal new effects and mechanisms for the old substances in the
corporate and personal data bases; - To select the most prospective
compounds for high throughput screening from the set of available
samples;
- To determine which screens are more relevant for particular compounds.
Only structural formula is necessary to estimate which kinds of activity
are more probable for the organic compound. The files of structures can
be prepared by either the PASS chemical editor CHAR or the MDL
Information Systems, Inc. chemical editor ISIS/Draw. If you already have
the structural files under ISIS/Base or ChemBase (MDL), these files can
be used directly. The prediction can be done even for structures that
are only planned to be synthesized. PASS 4.20 training set includes
about 10'000 biologically active substances. PASS 4.20 predicts the
probabilities of presence/absence for 114 biological actions
simultaneously (main and side pharmacological effects, mechanisms,
specific toxicity). It is shown that the approach, used in PASS, can be
applied to the other biological activities [6-8]. The calculation of the
biological activity spectrum per compound in ordinary IBM PC 486/100 MHz
takes about 0.4 seconds (predicted activity spectra for ~10000 compounds
after ~1 hour of calculation).
What is the mean accuracy of the prediction? It is 70-80% both in
leave-one-out cross validation and in prediction for independent test
sets including about 5000 compounds diverse in both structure and
activity [1-4]. PASS prediction accuracy more than 3 times exceeds the
expert's guess-work [9]. In blind prediction for 100 new drug-candidates
from the PharmaProjects database it is more than random guess-work by a
factor 46 [10].
What is the economic viability of prediction? Experiment of PASS's
application in high throughput screening for independent diverse test
series shows that the economic viability is about 500-800% [11].
How reliable are the predictions? It depends on the researcher's
purpose. The ultimate decision on how many and which structures should
be selected for testing depends on probability for a compound to be
active, experimental possibilities, and the researcher's aspiration
concerning the results innovation.
The reliability of prediction is high when the Probability value is
more than 70%. However, the tested compound may turn out to be an analog
of well-known drug from the training set.
The reliability of prediction and the compound's similarity to
well-known drugs are less, if the Probability equals to 30-70%. It is
much less if the Probability is lower than 30%. However, the less is the
calculated probability for the activity, the more is the chance to
discover a new chemical entry (the compound from chemical series for
which this activity was never found).
It is shown that the original algorithm of structure-activity
relationships analysis used in PASS provides very robust results of
prediction [12].
Where PASS is used? PASS is used in 6 Institutions:
- Department of Chemistry of the Lomonosov Moscow State University;
- Institute of Biomedical Chemistry (Moscow);
- National Research Center for Biologically Active Compounds (Staraya
Kupavna, Moscow Region);
- Research Institute of Pharmacology (Moscow);
- Pyatigorsk State Pharmaceutical Academy (Pyatigorsk);
- State Academy of Chemistry and Technology (Ivanovo).
PASS value is proved by many examples in which the results of
computer-aided predictions are confirmed by experiments. Some of these
examples are given below.
The activity spectra have been predicted for 300 new chemical
compounds, synthesized in the Chemical-Pharmacetical Research Institute
(Novokuznetzk). Twenty compounds have been selected for the testing as
probable antiulcer agents. Nine compounds have been synthesized and
tested. The potent antiulzer activity is found for 5 of these compounds.
The economic viability is about (300/20)100 =3D 1500%.
The activity spectra have been predicted for 520 new chemical
compounds, synthesized in the Institute of Organic Chemistry of Russian
Academy of Science (Moscow). Fourteen compounds have been selected for a
testing as the most prospective. It is shown that the results of 22
experiments made on the 5 various kinds of activity, coincide with
predictions in 20 cases. The accuracy of prediction is about 90%.
The activity spectra have been predicted for 332 new chemical
compounds, synthesized in the Institute of Organic Synthesis and Coal's
Chemistry of Kazakhsan National Academy of Science (Karaganda). Eighteen
compounds have been selected as the most prospective, and tested in 24
experiments (some of the selected compounds have been tested on 2-3
various kinds of activity). The experimental results coincide to the
predictions in 18 cases. The accuracy of prediction is about 80%.
Recently PASS has been tested on the sample of 34 new chemical
compounds, that are under study by Procter & Gamble. It is shown for
8 already tested compounds from the sample, that 20 predicted activities
are confirmed in the experiments, and 2 predicted activities are not
confirmed in the experiments.
The example of a predicted activity spectrum for LEVAMISOLE (CAS No
16595-80-5) is given below. LEVAMISOLE has been launched as
antihelmintic in 1969, and as immunomodulator in 1980 (Prous J., The
Year's Drug News. Therapeutic Targets, Barcelona, 1995, p.493, 536).
| Activity |
Probability to be: |
|
|
Active, % |
Inactive, % |
| Antihelmintic |
93.6 |
0.5 |
| Immunomodulator |
85.4 |
0.8 |
| Immunosuppressant |
72.6 |
2.1 |
| M-Cholinergic Blocker |
65.4 |
2.8 |
| N-Cholinergic Blocker |
59.7 |
2.8 |
| Cholinergic Blocker |
57.7 |
3.5 |
| Muscle Relaxant |
54.3 |
4.0 |
Let us consider what we can suggest concerning the activity of
substance 16595-80-5 based on the PASS prediction (only activities with
the probability more than 50% are given in the table). If we try to
select for which actions the substance 16595-80-5 have to be tested, we
obviously should test it as Antihelmintic, Immunomodulator/Immunosuppressant,
Cholinergic Blocker, Muscle Relaxant. In the first two screens we shall
find that 16595-80-5 is a lead compound for Antihelmintic and
Immunomodulator action; other tests show us possible side effects for
the compound. Therefore, based on the PASS prediction, it is possible to
find the main and side effects of the substance.
| WHAT WE PROPOSE FOR YOU ? |
We are open for collaboration with any person and organization from
both Industry and Academy in the following ways:
- PASS will be tested free on a small sample of your compounds;
- PASS possibilities will be extended to cover your fields of interest
on the basis of appropriate Agreement;
- PASS predictions will be provided for your compounds on the basis of
Confidentiality Agreement;
- PASS will be purchased 'as is' and used for research and education
purposes on the basis of Licence Agreement.
| For more information contact |
Prof. Vladimir V. Poroikov, Sc.D. in Pharmacology, Head of Laboratory
for Structure-Function Based Drug Design Institute of Biomedical
Chemistry of Russian Academy of Medical Sciences, Pogodinskaya Street,
10, Moscow, 119832, Russia
Tel: +7-095-246 3380, Fax: +7-095-245 0857, E-mail: vvp@ibmh.msk.su
| References |
1. PASS: Computerized prediction of biological activity spectra for
chemical substances. D.A.Filimonov, V.V.Poroikov. Bioactive Compound
Design: Possibilities for Industrial Use, BIOS Scientific Publishers,
Oxford, p.47-56, 1996.
2. Computer-aided prediction of biological activity spectra of chemical
substances on the basis of their structural formulae: computerized
system PASS. D.A. Filimonov, V.V. Poroikov, E.I. Karaicheva et. al.
Experimental and Clinical Pharmacology (Rus), 1995, V.58, No 2, p.56-62.
3. Computer Aided Prediction in New Drugs Research and Development.
V.V.Poroikov. Sc. D. in Pharmacology Thesis (Rus). Staraya Kupavna
(Moscow Region): National Research Center for Biologically Active
Compounds, 1995. - 348 pp.
4. Opimization of Synthesis and Pharmacological Testing of New Compounds
Based on Computerized Prediction of Their Biological Activity Spectra.
V.V. Poroikov, D.A.Filimonov, A.V.Stepanchikova et.al. Chim.-Pharm. J. (Rus),
1996, v.30, N 9, p.20-23.
5. Revealing of new Mechanism for Barbiturates Activity Potentiation on
the Basis of Computer Aided Prediction of Biological Activity Spectra.
V.V. Poroikov, A.P. Boudunova, V.P. Shamshin, S.A. Suhanova, A.A.
Trapkova, Yu.V. Burov. Bull. Natl. Res. Center for Biologically Active
Compounds (Rus), 1994, No 1, p.39-45.
6. Computerized Prediction of Antiamnestic Activity for Chemical
Compounds: PASS Possibilities Extending. D.A. Filimonov, V.V. Poroikov,
A.P. Boudunova, A.V. Rudnitskih, Yu.V. Burov. Abstr. SCI Conference
"Design of Bioactive Compounds", 4-7 September, 1995, Potsdam,
Germany, p.26.
7. PASS: Prediction of Activity Spectra for Substance. Adrenergic
Compounds Set. E.V. Shilova, D.A. Filimonov, V.V. Poroikov. ibid, p.27.
8. Discovery of new chemical entry with antiulcer activity by using
computer aided prediction. Filimonov D.A., Trapkov V.A., Boudunova A.P.,
Burova O.A., Poroikov V.V. Abstr. XIVth International Symposium on
Medicinal Chemistry, Maastricht, The Netherlands, 8-12 September, 1996,
P-1.12.
9. Comparison of the Results of Prediction of the Spectra of Biological
Activity of Chemical Compounds by Experts and the PASS System.
V.V.Poroikov, D.A.Filimonov, A.P. Boudunova. Automatic Documentation and
Mathematical Linguistics. Allerton Press, Inc., 1993, 27, No 3, p.
40-43.
10. Russian computer program to cut screening costs? P. Charlish.
PharmaProjects Magazine, 1996, v.1, No 11, p.2-3.
11. Computer Assisted Prediction of Biological Activity Spectra:
Estimating the Effectivity of Use in High Throughput Screening. V.V
Poroikov, D.A Filimonov, A.P.Boudunova. Abstr: XIVth International
Symposium on Medicinal Chemistry, Maastricht, the Netherlands, 8-12
September, 1996, P-3.05.
12. Robust prediction of many biological activities. D.A Filimonov.,
V.V. Poroikov. Abstr. 11th European Symposium on Quantitative
Structure-Activity Relationships: Computer-Assisted Lead Finding and
Optimisation, Lausanne, Switzerland, September 1-6, 1996, P-59A.
The
Journal QUANTITATIVE STRUCTURE-ACTIVITY RELATIONSHIPS
This VCH journal is considered to be the "home" journal of
THE QSAR AND MODELLING SOCIETY.
New Editors of the Journal 'Quantitative Structure-Activity
Relationships'
After being responsible as the Editor of the journal 'Quantitative
Structure-Activity Relationships' for more than 15 years, Prof. Joachim
K. Seydel retired also from this activity. I would like to express, in
all our names, our sincere thanks to him. His ongoing enthusiasm for the
QSAR discipline contributed so much to the quality of this journal and
to our sciencific community. The new Editors are Prof. Michael Wiese,
University of Halle, and Prof. Gerd Folkers, ETH Zurich. Please
send your manuscripts to:
Prof. Dr. Michael Wiese:
Department of Pharmacy Martin-Luther-University Wolfgang-Langenbeck-Strasse
4 D-06120 Halle/Saale Germany
Phone: +49-345-552 5040, FAX: +49-345-552 7018, e-mail: wiese@medchem2.pharmazie.uni-halle.de
and consider that a publication in this journal will reach your
audience of QSAR and modelling colleagues much better than a publication
in JACS, JCICS, JMC, Biochemistry, etc. Of course, Ferenc Darvas remains
the Editor of the Abstracts Section. Please consider also to subscribe
personally to the QSAR journal. It's good and it's cheap, extremely
cheap for members of our Society (call VCH, phone +49-6201-6060, for the
current price).
Preview Quantitative Structure-Activity Relationships, 16, No. 5
(1997)
|
A. Martinez, C. Ochoa, J. Rodriguez, M. Rodriguez, A. Castro, M.
Gonzlez, M. M. Martinez :Comparative Molecular Field Analysis (CoMFA)
on [6] +[6] Fused Pyrazines with Nematocide Properties. |
|
P. Ertl: Simple Quantum Chemical Parameters as an Alternative
to the Hammett Sigma Constants in QSAR Studies. |
|
P. Chiba, M. Hitzler, E. Richter, M. Huber, C. Tmej, E. Giovagnoni,
G. Ecker: Studies on Propafenone-type Modulators of Multidrug
Resistance III: Variations on the Nitrogen. |
|
A. Paleti, S. P. Gupta: Quantitative Structure-Activity
Relationship Studies on Some Nonbenzodiazepines Binding to
Benzodiazepine Receptor. |
|
L. Eriksson, E. Johansson, M. Moeller, S. Wold: Cluster-based
Design in Environmental QSAR. |
New Books
H. Kubinyi, G. Folkers and Y. C. Martin, Eds., 3D QSAR in Drug
Design. Vol. II. Ligand-Protein Interactions and Molecular Similarity
SECTION 1. Ligand-Protein Interactions
T. Liljefors: Progress in Force Field Calculations of Molecular
Interaction Fields and Intermolecular Interactions
R. C. Wade, A. R. Ortiz and F. Gago: Comparative Binding Energy
Analysis
T. I. Oprea and G. R. Marshall: Receptor-Based Prediction of
Binding Affinities
M. K. Holloway: A Priori Prediction of Ligand Affinity by Energy
Minimization
M. R. Reddy, V. N. Viswanadhan and M. D. Erion: Rapid Estimation
of Relative Binding Affinities of Enzyme Inhibitors
R. M. A. Knegtel and P. D. J. Grootenhuis: Binding Affinities and
Non-Bonded Interaction Energies
I. T. Weber and R. W. Harrison: Molecular Mechanics Calculations
on Protein-Ligand Complexes
SECTION 2. Quantum Chemical Models and Molecular Dynamics
Simulations
B. Beck and T. Clark: Some Biological Applications of Semiempirical
MO-Theory
W. Andreoni: Density-Functional Theory and Molecular Dynamics: A
New Perspective for Simulations of Biological Systems
P. Carloni and F. Alber: Density-Functional Theory Investigations
of Enzyme-Substrate Interactions
D. Rognan: Molecular Dynamics Simulations: A Tool for Drug Design
SECTION 3. Pharmacophore Modelling and Molecular Similarity
R. D. Clark, A. M. Ferguson and R. D. Cramer: Bioisosterism and
Molecular Diversity
H. Kubinyi: Similarity and Dissimilarity - A Medicinal Chemist's
View
A. K. Ghose and J. J. Wendoloski: Pharmacophore Modeling:
Methods, Experimental Verification and Applications
S. Anzali, J. Gasteiger, U. Holzgrabe, J. Polanski, J. Sadowski, A.
Teckentrup and M. Wagner: The Use of Self-Organizing Neural Networks
in Drug Design
D. A. Thorner, D. J. Wild, P. Willett and P. M. Wright: Calculation
of Structural Similarity by the Alignment of Molecular Electrostatic
Potentials
A. C. Good and W. G. Richards: Explicit Calculation of 3D
Molecular Similarity
R. S. Pearlman and K. M. Smith: Novel Software Tools for Chemical
Similarity
R. Todeschini and P. Gramatica: New 3D Molecular Descriptors -
The WHIM Theory and QSAR Applications
T. W. Heritage, A. M. Ferguson, D. B. Turner and P. Willet:t EVA
- A Novel Theoretical Descriptor for QSAR Studies
H. Kubinyi, G. Folkers and Y. C. Martin, Eds., 3D QSAR in Drug
Design. Volume III. Recent Advances
SECTION 1. 3D QSAR Methodology. CoMFA and Related Approaches
Y. C. Martin: 3D QSAR: Current State, Scope, and Limitations
U. Norinder Recent: Progress in CoMFA Methodology and Related
Techniques
R. T. Kroemer, P. Hecht, S. Guessregen and K. R. Liedl: Improving
the Predictive Quality of CoMFA Models
A. Tropsha and S. J. Cho: Cross-Validated R2 Guided Region
Selection for CoMFA Studies
G. Cruciani, S. Clementi and M. Pastor: GOLPE-Guided Region
Selection
G. Klebe: Comparative Molecular Similarity Indices Analysis -
CoMSIA
F. Lindgren and S. Raennar: Alternative Partial Least Squares (PLS)
Algorithms
SECTION 2. Receptor Models and Other 3D QSAR Approaches
M. Hahn and D. Rogers: Receptor Surface Models
M. Gurrath, G. Mueller and H.-D. Hoeltje: Pseudoreceptor Modeling
in Drug Design: Applications of Yak and PrGen
D. E. Walters: Genetically Evolved Receptor Models (GERM) as a 3D
QSAR Tool
W. J. Dunn III and A. J. Hopfinger: 3D QSAR of Flexible Molecules
Using Tensor Representation
B. D. Silverman, D. E. Platt, M. Pitman and I. Rigoutsos: Comparative
Molecular Moment Analysis (CoMMA)
SECTION 3. 3D QSAR Applications
E. A. Coats: The CoMFA Steroids as a Benchmark Data Set for
Development of 3D QSAR Methods
T. Langer: Molecular Similarity Characterization Using CoMFA
K. H. Kim: Building a Bridge Between G-Protein-Coupled Receptor
Modelling, Protein Crystallography, and 3D-QSAR Studies for Ligand
Design
K. H. Kim, G. Greco and E. Novellino: A Critical Review on Recent
CoMFA Applications
K. H. Kim: List of CoMFA References, 1993-1996
Publisher
US residents: Kluwer Academic Publishers, Order Dept. P.O.Box 358
Accord Station, Hingham MA 02018-0358, U.S.A., FAX (617)-871 6528,
e-mail kluwer@wkap.com
Europe and other Countries: Kluwer Academic Publishers, Spuiboulevard
50 P.O. Box 17, NL-3300 AA Dordrecht The Netherlands, FAX +11-78-639
2254, e-mail kluwer@wkap.com
Both volumes are announced for December 1997
Prices:
Volume II, ISBN 0-7923-4790-0 Hardbound, NLG 265.00 / GBP 90.00 / USD
151.00 (prepublication offer, valid from August 03 - December 01, 1997)
Volume III, ISBN 0-7923-4791-9 Hardbound, NLG 235.00 / GBP 80.00 /
USD 134.00 (prepublication offer, valid from August 03 - December 01,
1997)
Set of Volumes II and III, ISBN 0-7923-4792-7 Hardbound, NLG 400.00 /
GBP 126.00 / USD 228.00 (prepublication offer, valid from August 03 -
December 01, 1997)
J. Sangster (Sangster Research Laboratories, Montreal, Canada)
Octanol-Water Partition Coefficients: Fundamentals and Physical
Chemistry. ISBN:0-471-97397-1. Price: $115.00 Pages: 178. Wiley, New
York.
Contents: Thermodynamics and Extrathermodynamics of Partitioning.
Experimental Methods of Measurement. Discussion of Measurement Methods.
Methods of Calculating Partitioning Coefficients. This book describes
methods used for obtaining accurate LogP values as well as outlining the
principles behind the use of such data. It gives a complete introduction
to partition coefficients and describes all current methods of measuring
LogP explaining the strengths and weaknesses. It is the most complete
survey of measurement methods in any publication.
Martin, Y.C. and Willett, P., Editors Designing Bioactive
Molecules: Three-Dimensional Techniques and Applications
ACS Professional Reference Books;; ISBN: 0-8412-3490-6; Catalog No.
3490-6-419; Due November 1997; 352 pages; Clothbound; $139.95.
Contents:
The development of 3D database systems: Cynthia Sellassie, Corwin Hansch
& ID Kuntz
Databases of 3D coordinate information: Robin Taylor
Generation of 3D coordinates: Daren Green Current
3D database searching systems: Wendy Warr & Peter Willett
Integration of molecular modeling and database searching: Mark Bures
Pharmacophore mapping: Yvonne Martin
Structure generation: Valerie Gillet and Peter Johnson
Docking: J. Scott Dixon & Jeffrey M. Blaney
Calculation of 3D similarity: Philip Dean
Approaches to 3D QSAR: Giovanni Greco, Ettore Novellino, and Yvonne
Martin
Summary and future developments: Peter Willett & Yvonne Martin
Meetings
and Courses
1998
|
12th European Symposium on Quantitative Structure-Activity
Relationships: Molecular Modelling and Prediction of
Bioactivity, August 23-28, 1998, Copenhagen, Denmark.
Contact: Dr. Klaus Gundertofte (fax: +45-36-301385). http://compchem.dfh.dk/qsar98/. |
|
3rd Swiss Course in Medicinal Chemistry. October 11-16,
1998, Leysin, Switzerland.
Contact: Prof. Bernard Testa (fax: +41-21-692 4505).http://www.pharma.ethz.ch/leysin |
1999
Contributions
to the Newsletter
All members are invited to contribute to the content of our
Newsletter. This Newsletter shall not be a one-man show, it gains from
your experience. Our publishing policy will not allow us to accept
scientific contributions which better should be sent to a reviewed
journal. However, tips and tricks, key references, conferences, books,
shareware, even the announcement of new commercial software, are
welcome. We depend on your active participation!
Please send your comments and contributions to
Han van de Waterbeemd: c/o Pfizer Central Research Dept. Drug
Metabolism Sandwich, Kent CT13 9NJ, UK
FAX :+44-1304-616433 E-MAIL: han_waterbeemd@sandwich.pfizer.com
In the future we would like to send the Newsletter only to members
who don't have access to the World-Wide Web. Please notify us if you do
not need the printed version. Your positive reply will help us to spend
the mailing expenses for better purposes.
!!!!!!!!! Thusfar only 36 members joined this action !!!!!!!! THANK
YOU FOR BEING THE NEXT
- The QSAR and Modelling Society - The QSAR and Modelling Society -
The QSAR and Modelling Society -
This page is free for publicity. Interested? Contact Han van de
Waterbeemd at fax +44-1304-616433.
|
Last Updated: June 27, 2001
|
