For us, the most important date of this fall is the 80^th birthday of the founder of the QSAR discipline and the Honorary Chair of our Society, Corwin Hansch, on October 06, 1998. A short curriculum vitae and an appreciation of his scientific contributions is given below.

Another important event in late summer was the 12th European Symposium on Quantitative Structure-Activity Relationships. Molecular Modelling and Prediction of Bioactivity, in Copenhagen, Denmark, from August 23–28, 1998. The symposium was successfully organised by Klaus Gundertofte, H. Lundbeck A/S, Copenhagen, and his staff. In total, 36 lectures were presented in 9 sessions and more than 130 posters in 7 sessions.
About 340 scientists attended this symposium, which takes place every two years in different European countries, in between the biannual QSAR Gordon Research conferences in Tilton, NH, USA. It was the common opinion of the symposium participants that, from a scientific point of view as well as from the technical organisation, this meeting had been a real highlight. The focus of the conference was on QSAR, 3D-QSAR and other computational approaches, ligand-protein interactions, different quantitative aspects of molecular modelling and the modelling of membrane penetration. The very first scientific session of the meeting was dedicated to Corwin Hansch’s birthday.
More details on the conference, including the abstracts of all lectures and posters and several important e-mail and Web addresses can be found at the URL http://compchem.dfh.dk/qsar98/sci_prog.html. A short report on the conference will appear in issue #7 (November 1998) of the new journal IDrugs. The Symposium Proceedings ”Molecular Modelling and Prediction of Bioactivity” will be published by Plenum Press, in March 1999 (Editor: Klaus Gundertofte).

The 1999 Gordon Research Conference on Quantitative Structure-Activity Relationships will take place in Tilton, NH, U.S.A., from August 25-30, 1999. The Chair of this conference will be Gerry Maggiora, gmaggio@pwinet.upj.com, the Vice Chair (program chair) will be Kate Holloway, kate_holloway@merck.com. As both are interested to put together a highly stimulating and exciting scientific program, please send any proposals for lectures to Kate. If you are interested to participate in the conference, please apply in early 1999. For the complete schedule of the 1999 Gordon Research Conferences see http://:grc.uri.edu/99sched.htm.

The 13^th European QSAR Symposium will be held in Düsseldorf, Germany, in August 2000 (Chair: Prof. Dr. Hans-Dieter Höltje,e-mail hoeltje@uni-duesseldorf.de); the 14^th European QSAR Symposium will be chaired by Martyn G. Ford, Portsmouth University, England, in August/September 2002.

As you will see from the Report of the Chair, our Society is in good shape and we are looking forward to grow even further in the near future.

Corwin Hansch, the founder of the QSAR discipline was born on October 6, 1918, in Kenmare, North Dakota. Although, to most of us, he needs no introduction, a short CV shall be given here. He started his studies at Lincoln College, in 1936-37, worked on his B.S. at the University of Illinois, from 1937 - June 1940 and his Ph.D. at New York University, from 1940 - January 1944. In January 1944, he married Gloria Tomasulo. They have been married for 54 years.  Together, they have a son Clifford (born 1954), and a daughter Carol (born 1958).
From January 1944 - July 1944, he was a postdoc at the University of Illinois, Chicago. From July 1944 - October 1944, he worked on the Manhattan Project at the University of Chicago. From November 1944 - October 1945, he was active as a group leader on the Manhattan Project, at du Pont de Nemours, in Richland. From October 1945 - February 1946, he was a Research Chemist at du Pont de Nemours, in Wilmington, Delaware.
In February 1946, he joined Pomona College, Claremont, where he is still active today. During this time he spent a sabbatical year at the ETH in Zurich, in 1952, and another sabbatical year at Huisgen's lab, in Munich, in 1959. He spent his other sabbaticals at Pomona College.
Whereas his first field of research was high temperature dehydrogenation, early on he changed to correlations of biological activity with chemical structure, as can be seen from his publication with Robert M. Muir, on The Relationship Between Structure and Activity in the Substituted Benzoic and Phenoxyacetic Acids, Plant Physiol. 26, 369-374 (1951). Another publication with the same coauthor on the Electronic Effect of Substituents on the Activity of Phenoxyacetic Acids, in "Plant Growth Regulation", Iowa State University Press, (1961), p 431, followed. This was the year before (!) O. R. Hansen published his relationships between insecticidal activities of benzoic acids and electronic sigma constants.
Corwin Hansch initiated the field of quantitative structure-activity relationships in the years 1962 and 1963 with these seminal publications:  C. Hansch, P. P. Maloney, T. Fujita and R. M. Muir, Correlation of Biological Activity of Phenoxyacetic Acids with Hammett Substituent Constants and Partition Coefficients, Nature 194, 178-180 (1962);  C. Hansch, R. M. Muir, T. Fujita, P. P. Maloney, C. F. Geiger and M.Streich, The Correlation of Biological Activity of Plant Growth-Regulators and Chloromycetin Derivatives with Hammett Constants and Partition Coefficients, J. Amer. Chem. Soc. 85, 2817-2824 (1963). In 1964, he published the papers: C. Hansch and T. Fujita, r-s-p analysis - A method for the correlation of biological activity and chemical structure, J. Amer. Chem. Soc. 86, 1616-1626 (1964), and T. Fujita, J. Iwasa, and C. Hansch, A new substituent constant, p, derived from partition coefficients, J. Amer. Chem. Soc. 86, 5175 - 5180 (1964). These two contributions are considered to be the key references for the development of the Hansch Analysis, as it is now called, and for the calculation of partition coefficients from lipophilicity constants.
He has earned numerous honours and awards, including the Smissman-Bristol Award in Medicinal Chemistry from the ACS in 1976, and an honorary  professorship from the University of Beijing. His papers are citation classics and for several years he was one of the 300 most cited authors, out of 1 million scientists, active in all different fields of research.
He has more than 300 publications as well as the following books:
C. Hansch and G. Helmkamp, "Organic Chemistry", McGraw-Hill Book Co., New York, (1959), 2nd Ed., McGraw-Hill Book Co., New York, (1963); Substituent Constants for Correlation Analysis in Chemistry and Biology (together with Al Leo, 1979); Comprehensive Medicinal Chemistry (6 Volumes), Editor (1990); Exploring QSAR, 2 Volumes (together with Al Leo and David Hoekman, 1994); Classical and 3D QSAR in Agrochemistry, (edited together with Toshio Fujita, as ACS Symposium Proceedings, 1995).
Several books by other authors have been dedicated to Corwin Hansch, in honour of his contributions to Science. Those who are interested to read about this in more detail, should take a look at http://clogp.pomona.edu/medchem/chem/papers/cv-hansch.html, where it is formulated: Dr. Hansch's complete C.V. has to be seen to be believed! (see also E. Coats, J. Seydel and A. Leo, Corwin Hansch. The Pioneer of QSAR, Quant. Struct.-Act. Relat. 7, 119-120 (1988)).

A Congratulation Address was formulated by the members of The QSAR and Modelling Society and signed by all participants of the 12th European Symposium on Quantitative Structure-Activity Relationships. Molecular Modelling and Prediction of Bioactivity, in Copenhagen, with the following text:

The QSAR and Modelling Society is still growing – after 346 members in August 1995, 463 members in August 1996, 549 members in July 1997, and 599 members in July 1998, we now have 616 members (effective date:  October 26, 1998). Only a few members left the Society because they changed their fields of activities. One member left the Society, because ”he had never realized that he was a member”.
Many thanks go to Han van de Waterbeemd for this strong support of the Society, but also to Yvonne Martin and to several board members. Also Didier Rognan, who maintains the Web page, Gerd Folkers, who provides the Server, and Pierre-Alain Carrupt, who maintains the e-mail list of the Society, deserve our sincere gratitude.

A meeting of the members of our Society took place in Copenhagen, on August 24. The following topics were discussed:

In the year 2000, a new chair has to be elected. Our proposal, to vote also for a new Board in the same year, was unanimously accepted by the members present in the meeting. As we do not (yet) have a set of rules to run our Society, we should continue the tradition that the elected chair will be free to select his officers, i.e. the Secretary and Newsletter Editor (this function might be distributed to two members, because a heavy load of work is on the shoulders of Han), the Advisor to the Chair, and the Treasurer.

The cash balance of the Society on August 21, 1998, was 7,337.21 US-$ and 250 GBP. In the future, a limited number of funds will be given to scientists on certain occasions (see below).
Due to repeated reminders and the dedicated effort of our Treasurer James King, some members finally decided to pay their dues more regularly. Possibly this behaviour was also stimulated by the Russian colleagues, who collected their dues now in the third year to 100% (see also below).

After the UK and Russian groups, also an Italian group formed in May 1997 and a Romanian group in December 1997. Due to the enormous financial problems of the Russians and the Romanians, the proposal of the Chair, not to collect dues from these two groups, was unanimously accepted in the members meeting. The proposal is to spend the money, which is collected within the groups, for local meetings, books, etc.

A list server for our Society was installed by Pierre-Alain Carrupt, University of Lausanne, in November 1997. Many thanks do not only go to him, but also to Stefan Balaz, Johann Gasteiger, Osman Guner, and Roger Lahana, who had offered their help in the same respect. To send a message to the list, please use the address qsar_society@unil.ch. To subscribe to the list, send an e-mail to majordomo@unil.ch, with the following text: subscribe qsar_society [your e-mail address]; to unsubscribe, use the text: unsubscribe qsar_society [your e-mail address]; to change your e-mail address, unsubscribe the old e-mail address and subscribe the new one.
Several colleagues proposed to open the list to the scientific community. However, this proposal was not accepted by the members in the meeting; there is a (understandable) fear that too many foolish and embarassing messages will be received if the list is open to non-members.

The participation fees of two scientists that attended the 12th European Symposium on Quantitative Structure Activity Relationships, Copenhagen, one from Poland and one from Bulgaria, were paid by the Society. We will continue this new tradition also at the next QSAR Gordon Conferences and the European QSAR Symposia. Two complimentary sets of ”3D QSAR in Drug Design”, Volumes 2 and 3, Editors Hugo Kubinyi, Yvonne C. Martin and Gerd Folkers, Kluwer Academic Publishers, Dordrecht, 1998, were given to the Russian group (to Oleg Raevsky and Vladimir Poroikov, via Igor Baskin) and one set to the Romanian group (to Zeno Simon, via Tudor Oprea). This was made possible by the generosity of the publisher and the editors, who together decided that a certain number of personal copies should be given to the Society. Further books will be distributed soon.

As already mentioned several times, this journal is considered to be the ”home” journal of our Society. All members and especially the poster presenters at the last European QSAR Symposium are encouraged to send their full-length manuscripts to Michael Wiese, the new editor (together with Gerd Folkers). His address is:

 Prof. Dr. Michael Wiese
 Martin-Luther University
 Institute of Pharmaceutical Chemistry
 Wolfgang-Langenbeck-Str. 4
 D-06120 Halle
 Germany
 FAX  +49-345-552 7018
 e-mail  wiese@pharmazie.uni-halle.de

Of course, the publisher Wiley-VCH would also like to encourage you to order a personal copy of this important journal. First of all, it has a relatively high impact factor, as compared to many other journals, and second, an incredibly low price for personal subscriptions is offered to our members:
 

Please support our work, especially for the distribution of messages, by regularly updating e-mail addresses. If you detect a wrong address in the members list in our Web page, please inform
Han van de Waterbeemd (han_waterbeemd@sandwich.pfizer.com), and Pierre-Alain Carrupt (carrupt@ict.unil.ch).
An attempt to update the e-mail addresses at the Copenhagen meeting was unsuccessful because a hidden admirer of our Society took away the list of all addresses, on the very first day after it had been deposited.

We are a still growing Society and now count 619 members.  I would like to thank the following members for local distribution of the Newsletter:

If other members would like to take the responsibility for local distribution, please contact me. It would be of great help to set up a network. Many thanks to Pfizer Central Research, UK for printing of this Newsletter.
 

this is our first mail, distributed by a listserver to all Society members where we know the e-mail addresses. My e-mail, being manually sent to about 380 members on November 10, produced an enormous feedback. First of all, I got about 100 error messages, resulting from some 70 wrong e-mail addresses! In addition, I got several dozens of personal replies from members, most of them offering help in the distribution of the e-mails and appreciating this new service of the Society.

Fortunately, we do not any longer need these 'local mail distributors'. We were very pleased to receive also four offers for free listservers. These came from Pierre-Alain Carrupt, University of Lausanne, Switzerland, Roger Lahana, Synt:em, Nimes, France, Stefan Balaz, North Dakota State University, USA, and Osman Guner, MSI, San Diego, USA. Of course, only one site could be selected. We decided for Pierre-Alain Carrupt and we would like to thank him and his University very much but also all other colleagues who offered their help.

Pierre-Alain will take care that all incoming messages are checked whether their content fits some (unwritten) rules of our Society and of the University. This will guarantee that no bad or 'nonsense' messages will arrive at your site. Messages from companies should be restricted to scientific information on programs, approaches, or sources of such software; we appreciate such information but we will definitely not
distribute mere commercials. Long messages, data sets, attachments, etc., should not be distributed via the mailbox. Please, send such material also in the future to our Secretary, Han van de Waterbeemd, who will check whether the material should be included in our WebPage.

Dear colleagues, please use this new medium for the interchange of information and for a lively platform. Messages of general interest will also be published in the Newsletters and in the WebPage.

Some technical information and the e-mail address to which you can send your messages will be given by Pierre-Alain.

The best source for current information is our Web Home Page (http://www.pharma.ethz.ch/qsar). You are encouraged to participate actively in improving and updating this site by sending us information and suggestions.

We want to announce you, hereby, about the existence af the Rumanian QSAR-Group. Besides the QSAR and Quantum Chemistry Group of Timisoara, formed in 1975, there are members in Bucharest and Cluj. Hereby, a list of the members of the Rumanian QSAR-Group, with their addresses. We are glad that the Hypermolecule-principle, a based also of our MTD-method, is also nowadays of interest (see QSAR-Tips of Philip S.Magee in Newsletter No. 8).

The Chair of the Rumanian Group is Zeno Simon, the Secretary Ludovic Kurunczi (zsimon@cbg.uvt.ro  or  dock@icht.sorostm.ro)

Dr. A. Chiriac or Dr. Z. Simon, Catedra de Chimie, Facultatea de Chimie-Biologie-Geografie, Universitatea de Vest, Str. Pestalozzi 16, 1900-Timisoara, Romania. Tel +40-56-190377

Rumanian QSAR-Group Members:

a) Universitatea de Vest, Timisoara, Catedra de Chimie
A.Chiriac, C.Bologa, M.Mracec, Z.Simon
b)  Institutul de Chimie al Academiei, Timisoara (Bd. Mihai Viteazu, 24) Mioara Mracec,
Simona Timofei, F.Elenes, E.Seclaman
c) Univ. de Medicina si Farmacie, Timisoara, Fac. Farmacie (P-ta Eftimie Murgu, 2)
D.Ciubotariu, D.Dragos, A.Grozav, L.Kurunczi
d) Univ.Politehnica, Timisoara, Fac. Chimie Industriala
D.Hadaruga, M.Medeleanu
e) Univ. Politehnica, Bucuresti, Cat. Chimie Fizica
A.T. Balaban, O. Ivanciuc
f) Universitatea Bucuresti, Cat. Chimie Fizica
V.E.Sahini, Alexandrina Donescu, Mihaela Hillebrand, Elena
Volanschi, Josette Weinberg

Members from abroad:
T.I.Oprea, Astra Hassle, Molndal-Sweden,
S.Muresan, ATO-DLO, Wageningen, The Netherlands,
T.Sulea, Biotech.Res.Inst., NRC, Montreal, Canada

      Seventy scientists, from industry, government and university gathered in Dayton (Ohio, USA) for the Workshop “Computational methods in Toxicology” (April 20-22, 1998). The Workshop was organized under the aegis of the US Air Force Research Laboratory’s Operational Toxicology Branch and of the Aeronautical Systems Center (Dayton, Ohio); the organizing committee was composed of J. Frazier, K. Geiss, J. Labanovski, R. Pachter, S. Trohalaki, and T. Windus. The scientific programme was designed to provide an extensive overview of the approaches for modeling and predicting toxicity, as well as of the QSAR methods most relevant to toxicological applications. The general goal was to contribute to the task of the US Air Force Laboratory in developing systematic methods for assessing potential adverse effects of chemicals of interest to the Air Force.
       Oral presentations were made by 18 invited speakers; a poster session was also organized. Overviews of the methods and tools which can be used for the design of structure-toxicity models were presented by several speakers. P. Jurs (University Park) presented a general approach to QSAR modeling. M. Randic (Des Moines), S. Basak (Duluth) and D.J. Klein (Galveston) focused  on theoretically-derived descriptors. J. Devillers (Lyon) discussed the notion of inter-communicating hybrid systems, which combine different linear and non-linear modeling techniques, as well as statistical and graphical tools. A. Tropsha (Chapel Hill) discussed the variable selection in QSAR procedures.        An updated overview of the computational models developed for use in toxicology was presented. R. Benigni (Rome) reported on two important exercises for the prediction of chemical carcinogenicity held under the aegis of the US National Toxicology Program (NTP), and showed that the various approaches were able to point to the presence or absence of alerting chemical functionalities, but did not accomplish the task of making gradations within each potentially harmful class. D. Bristol (Research Triangle Park) further reported on the NTP carcinogenicity prediction exercises, and particularly focused on pattern-recognition models like the decision-tree induction and neural-network techniques. H.S. Rosenkranz (Pittsburgh) presented a number of analyses -performed with the CASE/MULTICASE system- aimed at assessing the effects of size, informational content, ratio of actives/inactives in the model on predictivity. N. Greene (Leeds) illustrated the characteristics and potential of DEREK, an expert system designed to identify potential toxicological hazards from chemical structure, as well new developments represented by the StAR and METEOR projects. M. Arnott (Boyertown) presented the OncoLogic expert system for the prediction of chemical carcinogenicity, and discussed its use in the second NTP prediction exercise. The important point of the construction and mining of toxicological databases -with the main goal of predicting the toxicity of chemicals not yet tested- was addressed by E. Thompson (Cincinnati) and E. Matthews (Rockville). A.M. Richard (Research Triangle Park) discussed the contribution of QSAR and molecular modeling  as means for inquiry into mechanism. R.L. Lipnick (Washington) gave an historical overview of the process by which correlative and mechanistic QSAR models were used and incorporated for review of industrial chemicals under TSCA. P. Seybold (Dayton), W.C. Herndon (El Paso), and K.L.E. Kaiser (Burlington) presented applications of QSAR models to a number of toxicity end points.
       A final round table discussion continued the discussions initiated from the oral presentations, and further focused on the strengths and weaknesses of the different approaches to toxicology modeling.
      Abstracts of the invited talks and posters can be found at: http//www.osc.edu/CCM/toxicology. The proceedings of the workshop will be published in the journal SAR and QSAR in Environmental Research.

In the last week of August, the 12th European Symposium on QSAR took place in Wonderful Copenhagen and was attended by 340 participants from 34 countries.
The Opening session included a talk by James P. Snyder with the interesting title Strategies for Molecular Design Beyond the Millennium.
The program was divided into 7 major sessions.
New Developments and Applications of Multivariate QSAR was dedicated to Corwin Hansch on the occasion of his 80th birthday. Svante Wold presented a plenary lecture entitled Multivariate Design and Modelling in QSAR, Combinatorial Chemistry, and Bioinformatics.
In the session The Future of 3D-QSAR, Gabriele Cruciani presented his lecture Handling Information from 3D Grid Maps for QSAR Studies.
Prediction of Ligand-Protein Binding was covered by Peter J. Goodford (Conformationally Flexible Molecules) and Gerhard Klebe (Structural and Energetic Aspects of Protein-Ligand Binding in Drug Design).
Computational Aspects of Molecular Diversity and Combinatorial Libraries aimed at covering the new challenges faced by QSAR and Modelling scientists. Plenary lectures were Analysis of Large, High-Throughput Screening Data Using Recursive Partitioning by Stan S. Young and 3D Structure Descriptors for Biological Activity by Johann Gasteiger.
Thue W. Schwartz presented the lecture Surprisingly elusive binding sites for non-peptide ligands in 7TM receptors in the session Affinity and Efficacy Models of G-Protein Coupled Receptors, and Terry R. Stouch gave his lecture Understanding Membrane Permeation through Simulation in the session Modelling of Membrane Penetration.
A total of 27 oral reports were also given during the symposium and more than 150 posters remained on show for the whole week.
A number of social events were included in the program. The participants were honoured by a very memorable reception at the Copenhagen City Hall. We had an excursion In the footsteps of the Vikings and enjoyed the atmosphere in the Tivoli Gardens during and after the gala dinner.
The organisers and the Plenum publisher expect to bring out the Proceedings in March 1999.
Organising a symposium like this is becoming increasingly more time demanding. It was therefore suggested in the group of former and present chairs during this meeting to appoint the next two chairs, i.e. chairs for the meeting in the years 2000 and 2002. These will be Professor Hans-Dieter Höltje (Düsseldorf, Germany) and Professor Martyn Ford (Portsmouth, England), respectively. Congratulations. We are looking forward to these important events.

A number of abstract books are still available. Please drop an e-mail to kgu@lundbeck.com Our Russian colleagues should by now have received a copy. If not, please contact Professor Poroikov.

Scoring functions allow predictions of ligand binding affinities if 3D-structure data are available for the receptor site. The site atoms must be, hereby, characterises by structural parameters such as electric charge, hydrophobicity, etc. CoMFA yields only very incomplete informations of this kind, via stdev* coeff contour plots. On our opinion, more information could be obtained from CoMFA results by use of more realistic potentials and by listing, for each grid point atom, the "regression coefficients" indicating their contributions to predicted affinities.
In the receptor-ligand complex, interdistances between atoms of the two partners will not decrease sensibly below van der Waals contact distances, which should be used to calculate cut-off values for electrostatic interactions. For steric interactions, van der Waals grid point atom_ligand intersection volumes were recently proposed as alternative for the 6-12 Lenard-Jones potential. Also, steric misfit should always be repulsive! Hydrophobic potentials (via fragmental constants) should also be included. Receptor site atoms (regions) nearby grid points could thus be characterised by electric charge, stiffness towards deformation and lipophilicity.

1 Introduction
Recently, Wold and coworkers developed a novel filtering technique for spectral data called orthogonal signal correction (OSC) [1]. OSC is a partial least squares projections to latent structures (PLS) based solution that removes from the X-data (the matrix of predictors) variation that is unrelated to Y (the matrix of responses, or a single response variable y). This is possible because OSC uses Y to construct a filter of X. The result of OSC is a model based on one or more PLS components conveying information about the correction of X. Conventional PLS diagnostics, such as, scores and weights are then available to interpret exactly which kind of information that was peeled off from X.
We will now give an introductory account of the OSC algorithm. In this report OSC has been used to remove one component at a time from X using the conventional NIPALS algorithmic framework [2]. This has the advantage that the approach will cope also with moderate amounts of missing data, as do ordinary principal component analysis (PCA) and PLS. Prior to calculations, X and Y can be transformed, mean-centered, and scaled according to standard procedures.

1.1 Orthogonal signal correction, OSC
The first step in the calculation of each OSC component is the formation of the first principal component score vector of X. This score vector, t, can be seen as a “precursor correction vector”, which is orthogonalized with respect to Y to give t*, the actual correction vector. Subsequently, PLS weights (w) are computed such that Xw becomes as close as possible to t*. The correction vector t* is then processed in the NIPALS algorithm through X to give an updated score vector t, which is then again orthogonalized with respect to Y. This is iterated until convergence. This means that t* is directed towards the longest vector that is orthogonal to Y and which still provides a good modelling and prediction of X. After convergence, a loading vector p is computed and then t*p’ is subtracted from X. A second component may then be calculated, and so on. The resulting residual matrix E constitutes the filtered X-matrix.
With this short report we wish to disseminate the OSC concept to the QSAR community. By way of example it will be shown that OSC may enhance the predictive ability of a QSAR model.

2 Illustration
 We here use a series of polychlorinated biphenyls (PCBs) as an illustration. The entire series of 209 PCBs has been multivariately characterized by Tysklind and coworkers [3]. The X-variables comprise 52 predictors. Twenty-one (21) of these are of spectral origin (UV measurements in the 200-300 nm region) and the rest are either measured or quantum-chemically calculated physico-chemical descriptors. One biological response that has been measured for comparatively many PCB congeners by Ahlborg and coworkers is the in vitro inhibition of cell-cell communication in rat liver cells [4]. This response, expressed as the highest concentration (in *M) causing no inhibition, is available for 34 congeners and will be used as the log y-variable in this contribution. In summary, this data set is appropriate because (i) it contains a multitude of chemical descriptors of which many are spectroscopically determined, and (ii) the y-data are available for sufficiently many compounds. Throughout this contribution, the PCBs will be referred to by their standard IUPAC numbers (1-209).

3 Data Analysis
 The 34 compounds were sorted according log y. Every third compound was withdrawn from this sorted list of compounds, thus creating a prediction set with 11 observations. Consequently, the training set contained the remaining 23 congeners.
 PLS modelling was carried out as implemented in the SIMCA-P 7.01 software [5] using cross-validation [6].

4.3 Interpretation of OSC-model (what was removed from X ?)
 By interpreting the PLS weights w of the OSC model (Figure 5) it is possible to gain an understanding of what was removed from X. We understand from Figure 5 that OSC has removed variation from a predictor entitled RRF-4, which is a chromatographic response factor. Also, much has been removed from the hybridization in the x-direction of the molecular structure and the ionization potential. Regarding the spectral data (the UV-data), which are found in the right-hand part of the graph of Figure 5, OSC has peeled off most variation from the variables reflecting 225, 230 and 300 nm. Thus it appears that OSC has removed variation both among the measured and calculated physico-chemical descriptors, as well as among the spectral predictors.

4.4 Scaling of data prior to OSC
 In the above OSC-model, OSC was applied to a mean-centered and unit-variance scaled X-matrix. This is less commonly done with spectroscopic data, where usually mean-centered but unscaled data are “OSCed” [1]. In our case there is a substantial spread among the X-variable’s initial standard deviation, and invariably the 21 spectroscopic variables have much lower standard deviations than the other variables. The ratio of the largest to the smallest initial variable standard deviation exceeds 6.7*107.  This means that when OSC is performed on mean-centered X-data the spectroscopic variables will be masked by the physico-chemical parameters, and the correction efforts will affect only the latter.
Figure 6 displays PLS weights of the OSC analysis of mean-centered data. As seen, in the mean-centered case, OSC only filters three predictors, which is unsatisfactory. With this correction, the predictive power of the resulting QSAR model was Q²ext = 0.72, i.e., reminiscent of the predictive ability of the reference model established in section 4.1. We also tested to compute two OSC components instead of one, but no further improvements were accomplished.

Figure 7: PLS weights w of the QSAR model resulting after OSC-pretreatment (of scaled and centered data).
 
Among the UV-descriptors we may see that the predictively most relevant wavelengths for the modelled response are the regions 205-215 nm and 255-275 nm. The UV-absorption characteristics of the PCBs display two major absorption bands, viz. a main band in the lower 200nm region and a so called k-band in the 250 nm region, the latter being found mainly for non-ortho substituted congeners. Among the non-spectroscopic variables, we find that ionization potential and other energy indices, such as, core-core repulsion, and electronic and binding energies, dominate the model. Thus, the PLS weights suggest that the chemical properties of the PCBs that seem to be influential for the cell-cell communication inhibition are associated with the steric hindrance of the two phenyl rings. This is, for instance, inferred from the importance of the 255-275 nm region in the modelling. Hence, the QSAR model indicates that non-ortho and para-substituted PCBs are the least potent congeners, and that multiple-ortho and meta-substituted congeners are the most active ones.

5 Discussion
 Orthogonal signal correction was originally developed for filtering of noisy spectral data, and has indeed proven useful in such circumstances [1]. Using OSC has the advantage that models of better predictive ability often results, and that these models have a limited number of components compared to models based on non-filtered data.
 The multivariate characterization of the PCBs consists of 21descriptors reflecting their UV-absorption characteristics. The measured UV-spectra display considerable variation across the series of PCBs, a variation which may or may not be of relevance for QSAR modelling. It occurred to us that it would be of interest to try to pre-process these data with OSC prior to QSAR modelling. But because the multivariate characterization also embraced another set of 31 measured and quantum-chemically calculated descriptors, which also may account for variation that is not of predictive relevance in QSAR, we decided to include all 52 X-variables as part of the OSC pre-processing scheme.
 With the acquired 8% increase in external predictivity, this small study indicates the usefulness of OSC in QSAR. However, more rigorous and elaborate testing schemes are necessary in order to uncover if and when OSC may be of utility in QSAR. We envision OSC to be of interest in for instance 3D-QSAR where applications usually deal with thousands of variables,  and where most variables carry limited information about a given endpoint for a given problem.
 In summary, we find OSC to be a useful pre-processing tool which deserves a place in the chemometrics tool-box. It can be a useful supplement to other, more established, correction techniques like multiplicative signal correction [7] or standard normal variate correction [8]. It has to be tested from application to application whether pre-processing is needed, and, if so, which technique or combination of techniques to deploy. Our aim has been to highlight the existence of OSC to the QSAR community, and to point out that this emerging technique may be used as a fruitful supplement to other filtering approaches.
 

References
[1] S. Wold, H. Antti, F. Lindgren and J. Öhman, Orthogonal Signal Correction of Near-Infrared Spectra, Chemometrics and Intelligent Laboratory Systems, In press.
[2] S. Wold, L. Eriksson and M. Sjöström, PLS in Chemistry, Encyclopedia of Computational Chemistry, Wiley, In press.
[3] a) P. Andersson, P. Haglund, and M. Tysklind, The internal barriers of rotation for the 209 polychlorinated biphenyls, Environmental Science and Pollution Research 4:75 (1997). b)  P. Andersson, P. Haglund, and M. Tysklind, Ultraviolet absorption spectra of all 209 polychlorinated biphenyls evaluated by principal component analysis, Fresenius Journal of Analytical Chemistry 357:1088 (1997).
[4] Helena Hämming, Lars Wärngård and Ulf G. Ahlborg, Inhibition of Dye Transfer in Rat Liver WB Cell Culture by Polychlorinated Biphenyls, Pharmacology & Toxicology 69:416 (1991).
[5] SIMCA-P 7.01, Umetri AB, www.umetri.se
[6] S. Wold, Cross-validatory Estimation of the number of Components in Principal Components and Factor Analysis Models, Technometrics 20: 397 (1978).
[7] P. Geladi, D. MacDougall, H. Martens, Linearization and Scatter-correction for Near-infrared Reflectance Spectra of Meat, Applied Spectroscopy 3:491 (1985).
[8] R.J. Barnes, M.S. Dhanoa, S.J. Lister, Standard Normal Variate Transformation and De-trending of Near-infrared Diffuse Reflectance Spectra, Applied Spectroscopy 43:772 (1989).

PARM, a new genetic algorithm for 3D QSAR studies

Basing on the Walters’ s GERM (Genetic Evolved Receptor Model), PARM (PseudoAtomic Receptor Model) uses a combination of genetic algorithms and a cross-validation technique to produce atomic-level pseudoreceptor models starting from a set of known ligands.
In the PARM computation, 15 kinds of pseudo receptor atoms are defined first. Then the molecules in the training set are superimposed on a specific pharmacophore model, and a set of grid points is generated around the common surface of the superimposed ligands. Receptor models are obtained by placing atoms at these points in 3D space to simulate a receptor active site. These atoms interact with the ligands and the interaction energy between the each ligand and the receptor model is computed. By using a genetic algorithm and cross-validation technique, a number of atomic-level pseudoreceptor models which have a high correlation between intermolecular energy and bioactivity can be built. A QSAR equation is constructed for each model in the linear form of Bioactivity = A + B*E_inter. Energetic computation in PARM makes use of the TRIPOS 5.0 force field.
PARM generates the receptor models in the MOL2 file, so that we can check the characteristics of the receptor model within the SYBYL software.

The PARM predictive capability has recently been compared with two traditional  3D QSAR techniques such as CoMFA (Comparative Molecular Field Analysis) and HASL (Hypothetical Active Site Lattice).
Further testing are currently carried out to fully determine the power of  the PARM capability to solve problems associated with drug discovery, using different sets of molecules also showing high degrees of freedom  .

PARM does represent a new methodology that can be used in conjunction with existing techniques, and may yield valuable insights and solutions.

[1] Chen H. M., Zhou J. J., Xie G. R., PARM: A genetic evolved algorithm to predict bioactivity, J. Chem. Inf. Comput. Sci., 38: 243- 250 (1998).
[2] Santagati M., Doweyko A., Santagati A., Modica M., Guccione S.*, Chen H.M., Uccello Barretta G., Balzano F., 5-HT1A Receptors Mapping by Conformational Analysis (2D NOESY/MM)  and “THREE WAY MODELLING”  (HASL, CoMFA, PARM), Proceedings of the 12th European Symposium on Quantitative Structure-Acticity Relationships. Copenagen (Denmark), Aug. 23-28 1998.
 "Molecular Modelling and Prediction of Bioactivity", Gundertofte K., Jorgensen F.S. Eds.,  Plenum Press,  in press (exp. date March 1999).
[3] Ref. 2: forthcoming paper.

Currently this program is based on 11,200 carefully evaluated equations:  6,780 from the field of mechanistic physical-organic chemistry and 4456 based on the interaction of chemicals with biological systems. The latter vary in complexity from DNA to whole animals.  In order of increasing complexity, the Bio-classes are:  macromolecules, enzymes, organelles, single cells, organs/tissues, and finally multi-cellular organisms.  The latter are further subdivided as:  vertebrates, invertebrates, insects, fish, humans, and plants.  The two databases (Bio and Phys) may be searched jointly or separately.  (See Chem. Rev., 96, >1045, [1996])  As your new equations are being derived through the regression portion of C-QSAR, they can be compared with those already present in the databank.  Some parameters--such as hydrophobicity (log P), molar refractivity (and its excess over an alkane of equal size), and molar volume--are calculated for automatic loading into the regression program, while others can be automatically loaded from an extensive database of >electronic and steric substituent constants.  C-QSAR makes it easier to winnow out irrelevant parameters in a biological activity study, and to compare the coefficients (i.e., the relative importance) of those remaining to those in well-established physical chemical processes.
Prof. Hansch is anxious to get more 'feedback' from the field, and we can offer a limited number of users free access to C-QSAR on the internet.  For those having access to VAX/VMS and who are likely to publish results using C-QSAR, we can license the program at a very attractive price.

FlexS is a software tool for superimposing flexible ligands. Either the structural alignment of two rigid molecules is possible within a few seconds, or a flexible molecule can be fitted on a rigid reference structure within a few minutes. The run time performance of FlexS enables interactive usage as wellas screening through databases of nontrivial size.

Contents:

Preface. Comparison of Fish Bioconcentration Models (J. Devillers, D. Domine, S. Bintein, and W. Karcher).

QSAR in Aquatic Toxicology: A Mechanism of Action Approach Comparing Toxic Potency to Pimephales promelas, Tetrahymena pyriformis, and Vibrio fischeri (T.W. Schultz, G.D. Sinks, and A.P. Bearden).

 Comparative QSAR and 3-D-QSAR Analysis of the Mutagenicity of Nitroaromatic Compounds (R.L. Compadre, C. Byrd, and C.M. Compadre).

 Some Novel Approaches to Modeling Transdermal Penetration and Reactivity with Epidermal Proteins (P.S. Magee).

 Contribution of Structure-Odor Relationships to the Elucidation of the Origin of Musk Fragrance Activity (D. Zakarya and M. Chastrette).

 QSAR Analysis of Protoporphyrinogen Oxidase Inhibitors (K.N. Reddy, F.E. Dayan, and S.O. Duke).

 Comparative Quantitative Structure Activity Relationships (QSAR) of the Inhibition of Dihydrofolate Reductase (C.D. Selassie and T.E. Klein).

 A Generalized Approach to Comparative QSAR (C. Hansch, H. Gao, and D. Hoekman). Index.

Ordering Information:
USA: Taylor & Francis, 1900 Frost Road, Suite 101, Bristol, PA 19007-1598. Europe: Taylor & Francis Ltd, Rankine Road, Basingstoke, Hants RG24 8PR, United Kingdom.
 

Positions
 
SCIENTIFIC POSITIONS AT GMD-SCAI.  At GMD-SCAI we are carrying out several exciting bioinformatics projects, most of them together with pharmaceutical industry. In a group of about 15 scientists, we are developing methods for the identification of protein targets for disease therapy, for molecular docking and structure-based drug design, and for protein sequence analysis and protein structure prediction.
Detailed information on our work and a possibility to exercise our software is offered at http://www.gmd.de/SCAI/area-bioinf.html.

Our emphasis for the work of new scientists is placed on

- finding of target proteins for disease therapy, based on sequence
  and gene expression data
- homology-based protein modeling and 3D protein structure prediction based on
  protein threading
- multiple alignment of protein sequences and reconstruction of
  evolutionary trees
- biomolecular docking
- design and scoring of combinatorial libraries
- structural superposition of ligand molecules

Within the postdoc program posted under  http://ik.gmd.de/PD-97-98.html, we
invite applications in these areas. In addition, we have several project positions available for scientists with a doctoral degree as well as for Ph.D. students.

We are looking for people with scientific backgrounds in computer science, biology, or chemistry, with significant computing expertise and experience, especially on Unix platforms, and with interest and - preferably - knowledge and experience in bioinformatics and/or molecular modeling.

More information on our institute (SCAI, Institute for Algorithms and Scientific Computing at GMD) can be found in the URL below. Information can be obtained from the following address.

This VCH journal is considered to be the "home" journal of THE QSAR AND MODELLING SOCIETY. Editors are Prof. Michael Wiese, University of Halle, and Prof. Gerd Folkers, ETH Zurich.

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).

1999
We want to announce a Workshop on Virtual Screening, Schloss Rauischholzhausen, March 15-18, 1999, Germany.

Virtual Screening has emerged as a prospective alternative to high-throughput screening. This computer method is by no means yet mature, however it is important to seriously develop and expand its scope and to learn about its limitations. Different techniques are involved. First of all, the handling and screening of large databases that involves clustering and similarity searching. Once a reduced selection has been obtained either docking or molecular superposition methods come into play. Alternatively, libraries based on combinatorial principles from combinatorial chemistry or NMR evidence can be screened for complementarity with a given binding site or similarity with a set of given lead structures. Computational speed and reliable scoring functions either for docking or molecular superposition are essential in virtual screening.

We think time is ripe to hold a workshop on this topic. The idea is to get leading experts (about 70) together at a pleasant place in a very informal atmosphere. We have reserved a castle near Marburg, about 80km from Frankfurt/M.at March 15th to 18th, 1999. If you want to know more about this meeting and you are interested to sign up please consider the web page http://pc1664.pharmazie.uni-marburg.de/workshop99/

For further information Hans-Joachim Boehm,  Thomas Lengauer and Gerhard Klebe.
 

The 1999 Gordon Research Conference on Quantitative Structure-Activity Relationships will examine topics at the forefront of 2-D and 3-D QSAR and molecular modeling.  Current plans are to include sessions on:  Similarity, Diversity, and Library Design; Data Mining and Knowledge Base Development; QSAR Variable Selection and Improved QSAR Methods; The Nature of Biologically Active Molecules; Computing Ligand-Receptor Binding Affinity; Applications of Computer-Aided Molecular Design; 3-D Pharmacophore Perception; and New Developments.

All participants are encouraged to present a poster describing recent work.  Poster abstracts which are received by April 15, 1999 will be given consideration for oral presentation in the New Developments session.  The speakers for this session will not be chosen until closer to the meeting to ensure inclusion of late-breaking scientific developments.

In order to make this conference the best possible, we ask that you contact the chairs with your suggestions concerning topics, speakers, poster presentations, conference format, etc.  A preliminary program will be published here by the end of November with the final program available at the end of this year.

We look forward to another exciting meeting including a stimulating exchange of ideas.  Hope to see you there!

While "3D Searching" has established itself as one of the essential tools in Computer-Aided Drug Design and combinatorial library focusing, the ability to ask better questions to retrieve better hit lists from 3D databases is still a challenging task.  The questions in 3D Searching involve a search queries that represent "pharmacophore models."  Pharmacophore development may involve a range of activities from a simple visual pattern recognition, to fully automated model generation, to receptor-based methods.

In this symposium, we will try to cover new developments in the area of pharmacophore development as well as validation studies on the existing methods, experimental prototypes, and in short, full spectrum of techniques in pharmacophore development.

If you are interested in contributing to this symposium, the due date  for 150 word abstract is November 15, 1998.  Electronic version of ACS abstract forms are prefered and can be obtained from http://www.acs.org/meetings/abstract/absdown.html

As Internet is rapidly becoming an essential part of our day-to-day life more and more modeling and analysis tools are also becoming available through the Internet.

This symposium will show off the current developments in this area: corporate-intranet applications, on-line modeling services, educational initiatives on the Internet, modeling and analysis tools through the Internet, and others.

If you are interested in contributing to this symposium, the due date for 150 word abstracts is November 15, 1998.  Electronic version of ACS abstract forms are prefered and can be obtained from http://www.acs.org/meetings/abstract/absdown.html

All members are invited to contribute our Newsletter and to our website. 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!

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.

Comments to these pages to the site manager