#234 - Co-sponsored Sessions

ACS National Meeting
Fall, 2007
Chicago, IL

SUNDAY AFTERNOON

COMP - Phil Magee Memorial Symposium -- QSAR Reborn The Foundations of QSAR
Boston Convention and Exhibition Center (BCEC) 156C
John H. Block, Bob Clark, Organizers
1:30 54 Philip S. Magee: A life in QSAR
Marvin Charton, mcharton@pratt.edu, Pratt Institute, 200 Willougby Avenue, Brooklyn, NY 11205

A brief review of the work of Phil Magee on various aspects of QSAR and of the founding of the Cheminformatics and QSAR Society.

2:00 55 Molecular surfaces, QSAR, QSPR and reactivity
Tim Clark, clark@chemie.uni-erlangen.de, Friedrich-Alexander-Universitaet Erlangen-Nuernberg, Computer-Chemie-Centrum, Naegelsbachstrasse 25, 91052 Erlangen, Germany

In principle, local properties at or near the surface of a molecule should be adequate to define intermolecular interactions and therefore also most properties of interest in QSAR and QSPR. However, properties such as the local hardness, electronegativity and polarizability also allow us to model reactivity effectively. Furthermore, these local properties represent an interesting alternative to interaction energies with probes such as those used in CoMFA or GRID. This lecture will describe some of the more unusual applications of the surface-modeling approach using the local properties derived from semiempirical molecular orbital theory.

2:30 56 CoMFA investigation of Taft Es values
Yvonne C. Martin, yvonnecmartin@comcast.net, 2230 Chestnut St., Waukegan, IL 60087 and Ki H. Kim, pkhkim@gmail.com, Hope Drug Discovery Research Laboratory, 260 Southgate Drive, Vernon Hills, IL 60061.

Es values are the classic linear free energy parameter to describe the steric effect of substituents. With the advent of 3D QSAR, CoMFA in particular, we investigated whether Es is purely steric. We modeled the data on which Es is based, acid-catalyzed hydrolysis of esters, and found that although the majority of the differences between compounds can be explained by steric fields, there is a statistically significant contribution of electrostatic fields. Our work settles the long-standing debate as to whether or this hydrolysis is free of electronic effects Taft originally proposed.

3:00 Intermission
3:20 57 Conformation independent QSAR Descriptor, scaffold hopping with surface property based eHiTS LASSO
Zsolt Zsoldos, Darryl Reid, Bashir S. Sadjad, and Aniko Simon. SimBioSys Inc, 135 Queen's Plate Dr, Suite 520, Toronto, ON M9W 6V1, Canada

The novel Interacting Surface Point Type (ISPT) descriptor used in eHiTS LASSO is independent of the underlying scaffold. Similarity is measured based on the surface properties of potential ligands, disregarding the 2D topology and the conformation of the ligands. This "fuzzyness" makes the descriptor suitable for scaffold hopping applications.

An automated non-linear learning model extracts the key binding patterns from the chemical interaction properties encoded in the ISPT descriptor value vector of a set of compounds with known biological activity. The acquired knowledge is applied to evaluate the surface interaction models of the screened chemical compounds. The advantage of the ISPT descriptor of eHiTS LASSO over the 2D fingerprint based descriptors is the independence from the underlying 2D structural motifs allowing the recognition of structurally diverse ligands with similar interaction profiles. On the other hand, the eHiTS LASSO ISPT descriptor does not dependent on the 3D shape of the surface, thus the descriptor values are independent of the conformation of the ligand, which is an advantage over other surface based descriptors that are biased by the specific input conformation.

The method has been validated on a variety of practical virtual screening tasks. Results will be presented demonstrating the ability of the software to retrieve the majority of actives from the screening essay at the top of the ranking list. Cluster analysis using traditional 2D structural descriptors is used to highlight the scaffold differences in the actives retrieved by eHiTS LASSO with high ISPT similarity scores. The scaffold hopping ability of the descriptor makes LASSO a powerful tool for finding new leads without the same toxicity or potential intellectual property issues as the query.

3:50 58 Approaches to the use of quantum mechanical modeling in QSAR analysis of agrochemicals
Donald W. Boerth, dboerth@umassd.edu1, Todd C. Andrade1, and Erwin Eder2. (1) Department of Chemistry and Biochemistry, University of Massachusetts Dartmouth, North Dartmouth, MA 02747, (2) Institute of Toxicology, University of Würzburg, Versbacher Strasse 9, D-8700 Würzburg, NA, Germany

Theoretical molecular modeling has been widely used in the design and discovery of new pharmaceuticals in medicine and to a somewhat lesser extent in the development of new pesticides for agriculture. Our research has sought to utilize computational modeling as a means of developing molecular descriptors for screening pesticides for genotoxic potential and induction of stress in crop plants, as well as for bioactivity. A combination of semi-empirical and ab initio theory coupled with density functional methods have been employed to characterize these interactions in several classes of pesticides. Electrostatic potentials have been used to establish putative sites for interactions of pesticide molecules with DNA bases and other biomolecules. These studies have been followed by modeling of the energetics of pesticide binding at these sites in the biological systems. A QSAR analysis is constructed by correlation of the computed results with available experimental descriptors.

4:20 59 QSAR without arbitrary descriptors: The electron-conformational method
B Bersuker, bersuker@cm.utexas.edu, Department of Chemistry & Biochemistry, Institute for Theoretical Chemistry, The University of Texas at Austin, 1 University Station A5300, Austin, TX 78712

It is shown that the electronic structure and geometry of the molecular system obtained from conformational analysis and quantum-chemical calculations and presented in a matrix form serves as a unique, most accurate descriptor in its interaction with the bioreceptor instead of arbitrary descriptors in conventional QSAR methods. By processing such electron-conformational (EC) matrices of a set of molecules in relation to their known activities in comparison with EC of inactive compounds by means of special programs, a common submatrix of activity is revealed, the pharmacophore (Pha). The latter is thus unique with a theoretically 100% qualitative (yes, no) prediction of the activity. The second part of the EC method takes into account the influence of Pha flexibility and the out-of-Pha groups on the activity quantitatively by means of regression analysis. Specific calculations are presented for antimitotic antitumor activity, glutamate receptor agonists, and antidiabetics.

MONDAY AFTERNOON

COMP - Phil Magee Memorial Symposium -- QSAR Reborn Theory
Boston Convention and Exhibition Center (BCEC) 156C
John H. Block, Bob Clark, Organizers
176 Paper Withdrawn
Tudor I. Oprea Tharun K. Allu, Dan C. Fara, Oleg Ursu
1:30 172 QSAR applied in systems biology
David E. Patterson, patterson.david.e@gmail.com, Vistamont Consultancy, 571 Vistamont Ave., Berkeley, CA 94708

Given enough training compounds which bind in a consistent manner at a binding site, QSAR is a valuable tool for designing new molecules with suitable activity. Extending this case to a profile of desired target activities, while challenging, is well defined and may be approached as an intersection of independent QSAR models. The advent of "-omics" assays and systems biology based readouts presents new applications and associated questions about whether, and how, to apply QSAR usefully to design molecules with desired phenotypic patterns. The molecular target of the QSAR model is no longer known, and there may be multiple targets in multiple pathways interacting in nonmonotonic fashion. Does QSAR fill a role in this arena? What does one model? Preliminary results of prospective QSAR based selection of compounds for scaffold hopping suggest that QSAR in systems biology holds promise, with substantial room for innovation.

2:00 173 The role of alignment in 3-D QSAR
Robert D. Clark, bclark@tripos.com and Richard Cramer, dcramer@tripos.com. Informatics Research Center, Tripos, 1699 S. Hanley Rd., St. Louis, MO 63144

Some 3D QSAR methods can be applied to conformational ensembles, but most require that each ligand of interest be put into a specific conformation - typically the "bioactive conformation." The well-established and widely-used Comparative Molecular Field Analysis (CoMFA) and Comparative Molecular Similarity Analysis (CoMSIA) methods go a step further and require that the specified conformation of each ligand be put into a common Cartesian frame of reference. Conformer generation and alignment can be done manually or automatically. Shared substructures, pharmacophores, docking modes, surface characteristics and topomeric heuristics can all be used to automate the process, with each yielding a model or models corresponding to a different null hypothesis. Here, we will examine the effectiveness of the various methodologies in creating robust and predictive models.

2:30 174 Realizing Prospective QSAR through data fusion and modern descriptors
Curt M. Breneman, brenec@rpi.edu1, N Sukumar, nagams@rpi.edu2, Mark J. Embrechts3, Kristin P. Bennett, bennek@rpi.edu4, C. Matthew Sundling, sundlm@rpi.edu5, Mike Krein1, and Theresa Hepburn1. (1) Department of Chemistry / RECCR Center, Rensselaer Polytechnic Institute, 110-8th Street, Center for Biotechnology and Interdisciplinary Studies, Troy, NY 12180, (2) Department of Chemistry and Center for Biotechnology, Rensselaer Polytechnic Institute, Cogswell Laboratory, 110 8th Street, Troy, NY 12180-3590, (3) Department of Decision Sciences & Engineering Systems, Rensselaer Polytechnic Institute, 110 8th Street, Troy, NY 12180, (4) Department of Mathematics, Rensselaer Polytechnic Institute, Amos Eaton Building, 110 8th St, Troy, NY 12180, (5) Department of Chemistry and Center for Biotechnology and Interdisciplinary Studies, Rensselaer Polytechnic Institute, 110 8th Street, Troy, NY 12180-3590

The evolution of "prospective" molecular property prediction methods that truly fulfill the promise of QSAR have been paced by the need for parallel development of information-rich molecular descriptors and modern multi-objective machine-learning schemes. By creating multiple models employing data fusion techniques and multiple endpoints, maximum benefit can be derived from the relationship between the chemical information encoded within modern molecular descriptors and several channels of available experimental data. Examples of data fusion QSAR will be discussed, including means for determining domain applicability of the resulting models.

3:00   Intermission
3:20 175 QSAR model assessment
Douglas M. Hawkins, dhawkins@umn.edu, School of Statistics, University of Minnesota, 313 Ford Hall, 224 Church Street SE, Minneapolis, MN 55455 and Jessica J. Kraker, krakerjj@uwec.edu, Department of Mathematics, University of Wisconsin, Eau Claire, Eau Claire, WI 54701.

Fitting a QSAR model involves two vital follow-up steps. Model checking involves ensuring that the data used for model development are compatible with the model fitted. This can fail to be true because of outlier data cases, excessively influential cases, mixtures of populations and failures such as unmodeled curvature. Traditional diagnostics such as plots of observed or residual values versus predicted values are unreliable in the typical high-dimensional QSAR setting. Model validation involves assessing whether, and under what conditions, the model fitted to the data can be applied to future cases. Model validation was traditionally done using ‘hold-out' samples kept out of the entire analysis and used only for assessing the model fitted to the ‘learning' cases. We show that this approach is inefficient in the usual QSAR settings, and should be replaced by newer computation-intensive cross-validation methods making full use of all data for both learning and validation. A vital part of this process is to avoid the potential pitfalls that await if these methods are applied incorrectly; proper and improper implementations are illustrated in the context of a QSAR data set.

3:50 177 Proof of the pudding: How predictive are QSAR models?
Tudor I. Oprea, toprea@salud.unm.edu, Tharun K. Allu, Dan C. Fara, and Oleg Ursu. Division of Biocomputing, University of New Mexico School of Medicine, MSC11 6145, University of New Mexico, Albuquerque, NM 87131

The often-discussed qualities of good QSAR models increasingly include the ability to correctly predict external sets, in particular for objects outside the already-covered descriptor space. Since the proof of the pudding is in the eating, we address the issue of how true are QSAR-based classifier models. The predictivity and applicability of two classifiers, one for drug-likeness and one for aggregation, are discussed in the context of early chemical probe and lead discovery. Each of the two classifier models was extensively investigated using decision trees (CART, WEKA) and support vector machines(LIBSVM), starting from substructures (SMARTS) and "2D"-based properties. We used ensemble methods with randomized training set data to build a committee of subsidiary classifiers (C4.5 decision tree Multiboosting, SVM), then combined the individual outputs to create a single decision from the committee of models as a whole. Thus, 100 models were in fact used for classification. Beyond cross-validation (internal consistency check for model building) and small test-set external predictivity, we used the general accuracy rate as a "blind prediction" evaluation, where significantly larger numbers of previously unknown molecules were externally predicted. Although these methods indicate the classifiers have predictive capability, we found that forward predictions (predictions made before the actual experiments) do not always work as expected, in part because appropriate information is not always made available to the models before the experiments. However, the quality of the experiments can also be questioned, when confidence in the QSAR models allows it.

TUESDAY MORNING

PRES - Going With the Information Flow: Chemical Abstracts Service 100th Anniversary Presidential Symposium
Boston Convention and Exhibition Center (BCEC) 205 A/B
Organized by: Janice E. Mears, Eric Shively, Mary Virginia Orna

Presiding: Mary Virginia Orna
8:30 Introductory Remarks
M. V. Orna

During his multi-decade career at the University of Illinois, R. C. Fuson authored or coauthored an array of books, varying from elementary coverage of organic chemistry to advanced treatises for graduate students and professionals. His first publication in 1939, a set of lectures for the advanced student, ushered in a career of book writing, which has persisted long after his death, in the form of the remarkably enduring editions of “Systematic Identification of Organic Compounds.”

8:35 6 An impressionistic look at the history of CAS
Robert J. Massie, rmassie@cas.org, Director, Chemical Abstracts Service, American Chemical Society, 2540 Olentangy River Road, Columbus, OH 43202-1505

CAS began in 1907 with Chemical Abstracts, produced as a largely volunteer effort for many years. In more recent times, CAS has continued to benefit from a professional staff imbued with the same commitment and dedication exhibited by those hard-working volunteers. CAS' mission to make the world's disclosed chemistry-related information accessible to scientists has entailed many challenges over the years. Two world wars, the post-war information explosion, recurring financial crises, and competitive threats are among the obstacles CAS has weathered. Innovations in indexing, computer-assisted publishing, the creation of the CAS Chemical Registry, creative product development and international cooperation have been crucial elements in CAS' survival and success. Twenty-first century developments, such as new advertising-based business models on the Web and governmental participation in the information industry are no less daunting than the challenges CAS has faced before. A personal view of CAS' history and organizational personality will be presented along with thoughts on its course for the future.

9:10 7 The CAS database: Back to the future
Catharina Maulbecker, cmaulbecker@cas.org, Sales and Marketing, Chemical Abratracts Service, 2540 Olentangy River Road, Columbus, OH 43202

To mark the 100th anniversary of CAS, we will be exploring several research topics covered in the first issue of CA to see what scientists were investigating a century ago and how today's scientists might gain new insights from their discoveries. By exploring the old literature, one gains an appreciation of how the exacting, fundamental work performed by chemists in the early part of the 20th century established truths that inform today's research. The older abstracts contain detailed accounts of the experimental processes and their findings. Information, speculation, and interpretation gleaned from these details can provide new facts and perspectives for today's scientists. The more knowledge researchers possess, the better equipped they are to make new connections and to generate new ideas. Reexamining past discoveries in light of today's knowledge sparks innovative breakthroughs by accelerating the process of serendipity. This survey will provide several examples to illustrate the enduring relevance of the early literature of Chemical Abstracts and the electronic resources to which it gave rise.

9:45 8 The importance of CAS to the world's scientists
Hideaki Chihara, hchihara@jaici.or.jp, Japan Association of International Chemical Information, Nakai Bldg. 6-25-4, Honkomagome, Bunkyo-ku, Tokyo, Japan

CAS has served the world's scientists and engineers over almost 4 generations. When they begin a new research project, the first thing they do is to look for what is known to mankind about the subject using CAS data in a variety of forms. The importance of CAS' products and services is well recognized and appreciated by the R&D scientists as they know there is no other information source that can provide such an exhaustive search that is provided by CAS. The value of CAS databases stems from their comprehensive coverage in terms of subject area and time frame and from a well-defined index structure and depth of indexing. These elements of the value have not changed in this changing world. During these times, CAS survived the most serious crisis for its existence from late 1960's and early 1970's, when a number of other secondary information journals had to disappear, and it was fortunate for the world's scientists that CAS was able to continue. The importance of CAS today may be expressed by a brief single sentence: No scientist, especially chemical scientist, would think he/she can work without CAS services.

10:20 9 Chemical Abstracts service: Its role in the history and evolution of scientific information
Bonnie Lawlor, chescot@aol.com, 276 Upper Gulph Rd, Radnor, 19087

The scientific community has been information-centric for centuries. Initially relying upon oral and hand written methods of knowledge distribution, scientists found the advent of the printing press introduced new distribution channels in the form of books, almanacs and newsletters. Soon scientists were awash in information. Then, in 1655 with the publication of the first scholarly journal focused on abstracts of original research, scientists began to rely upon what are known as abstracting and indexing (A&I) services to manage information overload and ensure the broadest possible distribution of published research. Over the past century, Chemical Abstracts Service (CAS) has filled this critical role, expediting the flow of scholarly information in the chemical and related sciences. Since its inception in 1907, it has evolved with changes in technology and user expectations, and has taken a leadership role in the dissemination of scientific information. The author will discuss CAS' role and take a brief look at the challenges and opportunities that it now faces as young, born-digital researchers gradually come to dominate the scientific community.

10:55 10 SciFinder: It's part of the R&D process
Damon Ridley, d.ridley@chem.usyd.edu.au, School of Chemistry, University of Sydney, Building F11, Universisty of Sydney, New South Wales, 2006, Australia

At times scientists have very specific problems to solve. We may want a particular spectrum, or specific synthetic method, or a key document. At other times we may not know precisely what we want and indeed we are welcome to suggestions, particularly quite novel ones. Historically we achieved this through browsing in the library, or attending lectures outside our immediate fields at conferences. These options are still available to us.

Browsing in the electronic library is possible, although here scientists meet several basic problems including the amount of information in the electronic library and the inability to browse many items (e.g. structures, reaction diagrams, information in tables etc). SciFinder offers solutions in two principal ways. The first is through its ability to Explore topics, substances, properties, and reactions, both separately and in iterative combinations. The second is through its many post-processing functions, and to achieve this SciFinder critically depends on 100 years of intellectually indexed data.

This presentation will present examples on how the functionality in SciFinder allows scientists to browse the electronic literature in a unique and creative way, thereby opening new research opportunities.

11:30 Concluding Remarks
C.T. Hunt, ACS President

TUESDAY AFTERNOON

COMP - Phil Magee Memorial Symposium -- QSAR Reborn Methods I
Boston Convention and Exhibition Center (BCEC) 156C
John H. Block, Bob Clark, Organizers
1:30 249 New pharmacophore constrained Gaussian shape/electrostatic/ olored force field similarity searching tools: Feeding the synthetic beast with KIN

Andrew C. Good, andrew.good@bms.com, Structural Biology and Modeling, Bristol-Myers Squibb, 5 Research Parkway, Wallingford, CT 06492, Andrew Tebben, andrew.tebben@bms.com, Computer-Assisted Drug Design, Bristol Myers Squibb, Pharmaceuticals Research Institute, P. O. Box 4000, Princeton, NJ 08543-400, and Brian Claus, Computer-Assisted Drug Design, Bristol-Myers Squibb, P.O. Box 4000, Princeton, NJ 08543-4000.

A major historical limitation of many QSAR analyses has been a reliance on retrospective analysis. This has typically limited their extension to the design of novel compounds, a central requirement for many computational chemists. Molecular similarity calculations are in essence QSAR models driven by a molecular template, and have the advantage of easy extensibility to prospective analysis. With this in mind, the DOCK program has been extensively modified to permit its application in ligand-based de novo design. A Gaussian-based scoring function have been incorporated to permit shape, electrostatic potential and weighted colored force-field similarity searching. In addition Gaussian-based exclusion volumes and r-group linker constraints have been added to permit inclusion of steric constraint SAR and fragment screening. When combined with fragment databases culled from the existing chemistry space, the resulting program KIN provides a highly flexible tool for de novo core / head group replacement. Illustrations of the software's utility are highlighted with a number of search examples.

2:00 250 Application of pharmacophore fingerprint QSAR to 7TM drug design
Zheng Yang, Zheng.P.Yang@gsk.com, Computational and Structural Chemistry, Molecular Discovery Research, GlaxoSmithKline Pharmaceuticals, 1250 South Collegeville Road, Collegeville, PA 19426

This presentation will discuss application of the GSK in-house pharmacophore fingerprint quantitative structure-activity relationship (pFPQSAR) method to 7TM drug design. pFP is a GSK in-house implementation of three- and four-point pharmacophore fingerprinting that incorporates proprietary GSK physiochemical featurization and utilizes imported molecular conformers. The pFPQSAR method uses the binary pharmacophore bits as 3D molecular descriptors, and applies a nonlinear iterative partial least squares algorithm to correlate these descriptors with compound biological activities to construct QSAR models. The pFPQSAR methodology has been validated on GSK in-house kinase datasets, and has been used to build predictive QSAR models to support prospective 7TM ligand-based drug design for lead optimization.

2:30 251 Hierarchical QSAR technology on the base of simplex representation of molecular structure
Eugene N. Muratov, murik@ccmsi.us1, Victor E. Kuz'min, victor@2good.org2, and Anatoly G. Artemenko2. (1) Computational Center for Molecular Structure and Interactions, Jackson State University, 1400 J.R. Lynch Street, Jackson, MS 39217, (2) Laboratory of Theoretical Chemistry, A.V.Bogatsky Physical-Chemical Institute NAS of Ukraine, Lustdorfskaya Doroga 86, Odessa, 65080, Ukraine

The Hierarchical QSAR technology (HiT QSAR) is developed for solution of any structure-activity/property task and especially for the optimization of new effective pharmaceutical agents creation process. On each stage of this technology QSAR task is solved with the use of information received from a previous stage (system of permanently improved solutions). Simplex representation of molecular structure (SiRMS) is the basis of the developed technology. In SiRMS any molecule can be represented as the system of different simplexes (tetratomic fragments of fixed composition, structure, chirality and symmetry). Such representation allows unifying description of spatial structure of compounds with saving of complete stereochemical information and determination of molecular fragments increasing or decreasing investigated properties. The advantage of HiT QSAR over several well-known QSAR approaches was shown on example of the sets of acetylcholinesterase and angiotensin converting enzyme inhibitors. The successful application of HiT QSAR was confirmed by solution of different QSAR problems.

3:00 Intermission
3:20 252 Informatics-based to structure-based ADME/tox modeling
Anton J Hopfinger, hopfingr@gmail.com, College of Pharmacy, University of New Mexico, MSC 09 5360, Albuquerque, NM 87131-0001

The modeling of an ADME/Tox endpoint is highly dependent upon the complexity of the molecular mechanism involved. In cases where the molecular mechanism is complex, and/or pharmacological understanding is quite limited, an empirical informatics approach to develop predictive models is the preferred methodology to apply. We have developed a set of universal descriptors, called 4D-fingerprints, which capture the three-dimensional size, shape, chemical composition, reactive state and molecular flexibility of a molecule for informatics type ADME/Tox modeling. These descriptors have been applied to skin sensitization and eye irritation. For ADME/Tox endpoints where cellular membrane permeation and diffusion are involved, a pseudo structure-based design approach called membrane-interaction (MI-) QSAR analysis can be applied. Here descriptors derived from the simulation of an organic molecule passing through a phospholipid membrane assembly are used with intramolecular descriptors derived from the organic molecule to build MI-QSAR models. MI-QSAR simulation modeling has revealed that some organic compounds pass directly through the membrane and, presumably, into the interior of a cell, while other organic molecules use the membrane bilayer as a two-dimensional ‘sea', hopping from cell membrane to cell membrane in order to cross tissue composed of the cells. We will discuss the MI-QSAR modeling of blood-brain barrier penetration by organic compounds.

3:50 253 A novel technique for virtual discovery for study of multistage bioprocesses
Vladimir Potemkin, pva@csu.ru, Department of Chemistry, Chelyabinsk State University, Br. Kashirinych 129, Chelyabinsk, 454021, Russia

A lot of modern methods for virtual discovery predicts a bioactivity at the stage of receptor – ligand interaction. At the same time a biological action of a drug includes more than one stage of action even in cases of in vitro experiments. Therefore, a new method for virtual discovery is proposed. The method creates pseudo-atomic receptor model and allows to simulate a movement of a drug molecule to the receptor through water and membrane. The interaction with competitive sites is taken into account. Also the method allows to presuppose a metabolism of drug. The method has been used for detailed elucidation of stages of action for membranotropic dihydrofolatereductase inhibitors. It has been shown that the process of biological action of the drugs includes 3 critical stages: penetration through membrane, diffusion and interaction with the target. Some of compounds play a role of pro-drugs and their metabolism yields to an active molecule. The quantitative relationships for each of the stages are obtained. The importance of every stage is estimated for each molecule. Now the algorithm is used for virtual screening of more than 20 kinds of biological activities.

4:20 254 Workflows based quantitative structure-activity relationship modeling
Shaillay Kumar Dogra, shaillay@strandls.com and Ramesh Hariharan. Cheminformatics, Strand Life Sciences Pvt. Ltd, No. 237, Sir C. V. Raman Avenue, Raj Mahal Vilas, Bangalore, India7

Quantitative Structure-Activity Relationship (QSAR) modeling has now acquired complex dimensions from its humble beginnings. At times the focus of modelers is on fitting some model equation compromising comprehensibility in the process. For the purpose of interpretable models a two-pronged approach can be followed. One is to use intuitive descriptors in QSAR modeling. Another approach, that is presented here, advocates using simpler models over more complex ones. Model complexity can be defined in terms of algorithm complexity, number of descriptors used in the model, computation time required for training, model interpretability, etc. However, learning better models would obviously be preferred over simpler models. We present an approach wherein the user follows simple flowcharts that guide him in the modeling process, taking him from simple to complex algorithms as and when suitable model cannot be fit on the data. Several experience-based guidelines and technical tips that facilitate QSAR modeling are also presented.

WEDNESDAY MORNING

CHED - Writing to Learn: Using Writing to Engage Students in the Chemistry Classroom
Seaport Plaza A
Lorena Tribe, Ike Shibley, Organizers & Presiding
8:30 Introductory Remarks
8:35 422 Grading writing assignments without investing an inordinate amount of time
Ike Shibley, ias1@psu.edu, Division of Science, Penn State Berks, Tulpehocken Rd., Reading, PA 19610

Many teachers do not include written assignments as part of a course grade because they do not feel qualified to grade the assignments. Writing can be a critical component of student learning in a variety of chemistry courses and grading these assignments can be accomplished by any chemistry teacher. Several styles of grading will be presented to demonstrate the effectiveness of grading written work in a science course without a large time commitment. Several types of writing assignments will be presented to explore the variety of ways that writing can be used to facilitate thinking. Because writing often helps students to think more clearly, an argument will also be made for including ungraded writing as well.

9:00 423 Controversial science: A writing assignment in general chemistry
Bryan W. May, maybw@cctech.edu, Department of Science, Central Carolina Technical College, 506 North Guignard Drive, Sumter, SC 29150

Students in a freshman level general chemistry class write a paper on a controversial science topic. Topics are tailored to meet specific interests of each student. The paper consists of at least three parts: both sides of the specific controversy and a conclusion section where the student describes their own conclusions or describes how a controversy was ultimately resolved. Particular emphasis is placed on research methodology and proper formatting.

9:25 424 Learning curve: Reflections from exam rewrites
Karen Anderson, Department of Chemistry, Madison Area Technical College, Madison, WI 53704

Requiring students to re-correct their first exam of the semester can be quite an eye-opener. With limited instructor comments on the exam, students make use of collaborative learning with peers, peer tutors, and other course resources in order to properly annotate a guided-discovery correction's exercise. As part of this reflection, a personal assessment of their performance is required which captures highlights of their performance, areas that could be improved towards passing the next exam, and any major discovery revealed through the writing up of this exercise. The exam rewrite also gives an instructor a better understanding of why students make the mistakes they do, with some surprising and often humbling findings.

9:50 Intermission
10:00 425 Development of content understandings using student journals in an elementary education majors course
Donald J. Wink, dwink@uic.edu and Michael Dianovsky, dianovsk@uic.edu. Department of Chemistry (MC 111), University of Illinois at Chicago, 845 W. Taylor Street, Chicago, IL 60607

This paper describes the use of journals in a general education chemistry course for elementary education majors. The particular structure of the journal involves students' description of their understanding of a topic, its development, and its connection to an aspect of their lives. This becomes the basis of instructor feedback to correct student misunderstandings, validate their efforts to use metacognition, and shape their understanding of the meaning of content for themselves. The ways in which this aligns with principles of student identity development presented by Baxter Magolda and others will be discussed, with a particular understanding of how the writing changes to reflect an improved "voice" in the content domain.

10:25 426 Role of nontraditional texts in organic chemistry
Katie E. Amaral, kea12@psu.edu, Division of Science, Penn State Berks, Reading, PA 19610-6009

A work of non-fiction was used in a fundamental organic chemistry course to relate chemistry principles to everyday life. Three class days were set aside to discuss topics ranging from agriculture to preservatives to organic foods. A dialogue evolved from these discussions which, while not always germane to chemistry curriculum, increased student interest and engagement. Students then presented posters on self-chosen topics relating to the book and subsequent discussions, which encouraged them to explore appealing concepts in greater depth. The effect of this process on students' attitudes toward and perceptions of chemistry and learning will be discussed.

10:50 427 Nanotechnology as a topic for teaching freshman writing
Anderson L. Marsh, marsh@lvc.edu, Department of Chemistry, Lebanon Valley College, 101 N. College Ave., Annville, PA 17003

First year seminar courses have become popular on college campuses, and generally provide freshman with a detailed look at a specific topic. At Lebanon Valley College these courses also serve as an alternative to freshman English. This talk will be centered on one of those courses that specifically targeted science majors: nanotechnology. The overall structure of the course will be discussed, as well as the informal and formal writing assignments that were completed. During the semester students learned to read and think critically about nanotechnology using various texts. Class discussion over the semester focused on a variety of issues surrounding nanotechnology, ranging from the science and technology behind specific discoveries to ethical questions surrounding the application of these discoveries. The course allowed the students to explore certain aspects about scientific research that are not usually covered in freshman chemistry lecture or laboratory courses.

11:15 Discussion
11:25 Concluding Remarks

WEDNESDAY AFTERNOON

COMP - Phil Magee Memorial Symposium -- QSAR Reborn Applications I
Boston Convention and Exhibition Center (BCEC) 156A
Organized by: John H. Block, Bob Clark
8:30 387 Potency and selectivity of hydroxy hydantoins, a novel class of MMP-12 inhibitors: Structure-based QSAR analysis
Bo O. J. Nordén, Bo.Norden@AstraZeneca.com1, Igor Shamovsky2, Balint Gabos2, Magnus Munck af Rosenschöld2, Matti Lepistö2, Göran Carlström2, Johan Evenäs2, Djordje Musil3, and Kristina Stenvall, Kristina.Stenvall@astrazeneca.com2. (1) Department of Medicinal Chemistry, AstraZeneca R&D Lund, Scheelevagen, S-22187 Lund, Sweden, (2) Department of Medicinal Chemistry, AstraZeneca, S-221 87 Lund, Sweden, (3) Merck KGaA, Frankfurter Str. 250, D-64293 Darmstadt, Germany

Matrix metalloproteinases (MMPs) are a large family of zinc-containing enzymes that regulate the turnover of extra-cellular matrix proteins and activity of a number of pro-inflammatory mediators. Abnormal enzymatic activity of macrophage metalloelastase (MMP-12), one member of the MMP superfamily, is implicated in the development of cigarette smoke-induced emphysema, a hallmark of chronic obstructive pulmonary disease (COPD). Inhibition of MMP-12 activity therefore represents an attractive therapeutic strategy for the treatment of COPD. Since the enzymatic sites of MMPs contain Zn(II), the vast majority of MMP inhibitors carry a Zn(II)-chelating group to specifically target those sites. Most known MMP inhibitors possess a hydroxamic acid moiety, a strong Zn(II)-binding group which leads to their high-affinity binding to the enzymatic sites of MMPs. Correspondingly, such compounds generally exhibit potency across a wide range of MMPs. Hydroxy hydantoins, a novel class of MMP-12 inhibitors, consist of a weak Zn(II)-binding group, a hydantoin, and a lipophilic biphenyl P1' moiety, which is connected to the hydantoin via a carbinol linker. The binding mode of hydroxy hydantoins in MMPs is revealed by X-ray crystallography and solution state NMR, and is consistent with those of other Zn(II) binding MMP inhibitors. Selectivity of hydroxy hydantoins for MMP-12 against MMP-2, MMP-8 and MMP-9 is studied by different multivariate approaches and 3D-QSAR. Lipophilic substituents tend to increase potency of hydroxy hydantoins at all MMPs, whereas direct interactions of para- and meta-substituents of the terminal ring of the P1' domain with Lys-241, a distinctive residue of MMP-12, drive the selectivity.

9:00 388 PepT1 substrate QSAR and pharmacophore definition
Terry R. Stouch, tstouch@lexpharma.com, Computational Chemistry, Lexicon Pharmaceuticals, 350 Carter Road, Princeton, NJ 08540 and Balvinder S. Vig, balvinder.vig@bms.com, Biopharmaceutics R&D, Bristol-Myers Squibb, One Squibb Drive, New Brunswick, NJ 08903.

A designed series of mono, di, tri and tetra peptides and a functional membrane depolarization assay were used to assemble a consistent set of data that allowed the definition of a true substrate pharmacophore and QSAR. Straight-forward molecular modeling and the charge, size, hydrophobicity and flexibility of the constituent amino acids as well as total molecular charge and hydrophobicity were found sufficient to explain the unexpected range in transport seen for dipeptides.

9:30 389 3-D-QSAR study of submandibular gland tripeptide FEG and its analogs
Essam Metwally, emetwall@tripos.com1, Ronald D Mathison, rmathiso@ucalgary.ca2, Joseph S Davison, jdavison@ucalgary.ca2, and Robert D. Clark, bclark@tripos.com1. (1) Informatics Research Center, Tripos, 1699 S. Hanley Rd., Saint Louis, MO 63144, (2) Department of Physiology and Biophysics, University of Calgary, 3330 Hospital Dr. NW, Calgary, AB T2N-4N1, Canada

Submandibular tripeptide FEG (Phe-Glu-Gly) and its analogues are potent anti-inflammatory peptides. 3D-QSAR-CoMFA (comparative molecular field analysis) and CoMSIA (comparative molecule similarity indices analysis) were employed on the compound series after a GALAHAD model was constructed using a subset of the 10 most active compounds. The remaining compounds were flexibly aligned to this model. The models obtained were used to predict the activities of the 20 compound test set. Biological activity was used as a measure of the ability of the compound to bind to the “active-site” with compounds which exacerbate the response also treated as good-binders. The 3D-QSAR models obtained were in good statistical agreement with the experimental results despite having relatively low q2. Contour plots generated for the models are in good qualitative agreement with experimentally observed changes in behaviour.

10:00 390 Inorganic QSAR and imaging
David E Reichert, reichertd@wustl.edu, Mallinckrodt Institute of Radiology, Washington University School of Medicine, 510 S. Kingshighway Blvd, Campus Box 8225, St Louis, MO 63110

Imaging modalities commonly used in modern radiology include gamma scintigraphy, positron emission tomography (PET) and magnetic resonance imaging (MRI). All three of these rely heavily on biologically stable metal coordination complexes for providing signal or contrast. As part of our research in the design of new or improved imaging agents, we have utilized both 2D-QSAR and 3D-QSAR techniques such as CoMFA and CoMSIA on various classes of coordination complexes. These studies have allowed the prediction of important physiochemical parameters such as logP, as well as predictions of pharmacokinetic behavior and receptor binding. A particularly successful application is the use of CoMFA or CoMSIA, to serve as scoring functions for receptor docking studies of targeted metal complexes. The studies to be presented will focus primarily on copper complexes, but will also discuss technetium and rhenium containing complexes.

10:20 Intermission
10:40 391 3-D-QSAR models and activity predictions of human TRPV1 channel antagonists: Comparative molecular field analysis (CoMFA) and comparative molecular similarity analysis (CoMSIA) of cinnamide analogs
Vellarkad N. Viswanadhan, visv@amgen.com1, Yaxiong Sun1, and Mark H. Norman2. (1) Molecular Structure, Amgen, Inc, MS 29-M-B, Small Molecule Drug Discovery, 1, Amgen Center Drive, Thousand Oaks, CA 91360, (2) Small Molecule Drug Discovery, Amgen, Inc, 1, Amgen Center Drive, Thousand Oaks, CA 91320

Three Dimensional Quantitative Structure Activity Relationship (3D-QSAR) models for human TRPV1 channel antagonists were developed based on Comparative Molecular Field Analysis (CoMFA) and Comparative Molecular Similarity Analysis (CoMSiA), using a training set of 61 cinnamide TRPV1 antagonists. These models were tested on an independent test set of 47 antagonists not included in the training set. Alignment of these compounds followed a procedure described recently, which included weights for both internal energy and atom-to-atom matching against a reference molecule in its minimum energy conformation. Dependence of results on partial charge assignment was explored using Gasteiger-Huckel, Gasteiger-Marsili and AM1-BCC charge calculation methods. AM1-BCC charge assignments gave superior results overall, for both CoMFA and CoMSiA models. Comparison of the CoMSiA and CoMFA results showed that both gave very similar results. For CoMFA, the best cross-validated correlations included steric and electrostatic fields (r2 = 0.96, q2 = 0.58, n = 61 for the training set and r2 = 0.50, n = 47 for the test set). For CoMSiA, the best cross-validated correlations included steric, electrostatic and hydrophobic fields (r2 = 0.95, q2 = 0.57, n = 61 for the training set and r2 = 0.48, n = 47 for the test set). Docking of these molecules in a homology model of TM3/4 helical region of TRPV1 showed consistency between the homology model and 3D-QSAR models. Additionally, molecular alignment used in these 3D-QSAR models was also consistent with the proposed binding modes of known activators of the TRPV1 channel, such as capsaicin and reseniferatoxin.

11:10 392 MAO-A safety profile of oxazolidinone antibacterials
C Eyermann, Joe.Eyermann@astrazeneca.com1, P. Fleming, paul.fleming@astrazenica.com1, M. Gravestock, michaelgravestock@verizon.net1, T. Jones, rrr@st-and.ac.uk2, G. Kern, gunther.kern@astrazenica.com1, R Ramsay, rrr@st-and.ac.uk2, F. Reck, folkert.reck@astrazenica.com1, and F. Zhou, fei.zhou@astrazenica.com1. (1) Infection Discovery, Cancer and Infection Research Area, AstraZeneca, R&D Boston Inc, 35 Gatehouse Drive, Waltham, MA 02451, (2) University of St. Andrews, North Haugh, St. Andrews, Fife KY16 9ST, England

Oxazolidinones are a new class of synthetic antibacterial agents that show good activity against Gram-positive bacteria. A concern with oxazolidinones as a drug class has been inhibition of MAO, especially Type A (MAO-A), due to structural similarity to MAO inhibitors like toloxatone. Inhibition of MAO-A could potentially lead to severe hypertensive crises as a result of ingestion of tyramine-containing food together with an oxazolidinone drug. It is therefore desirable to develop novel oxazolidinone drugs, which have improved activity against linezolid-resistant Gram-positive bacteria and show an improved safety profile with regard to MAO-A inhibition. Using the published crystal structure of MAO-B, a homology model of MAO-A has been built and used to interpret experimental studies on the orientation of oxazolidinones in the MAO-A active site. In addition, the homology model has been used to guide the design of new oxazolidinones which have reduced MAO-A inhibition while maintaining good anitbacterial activity.

11:40 393 Neural network-based QSAR and the discovery of the next generation spinosyn insecticide: Spinetoram (DE-175)
Thomas C. Sparks, tcsparks@dow.com1, Gary D. Crouse, gdcrouse@dowagro.com2, James E. Dripps1, Peter B. Anzeveno1, Jacek Martynow1, and James Gifford1. (1) Discovery Research, Dow AgroSciences, 9330 Zionsville Rd., Bldg. 306/G1, Indianapolis, IN 46268, (2) Dow AgroSciences LLC, 9330 Zionsville Road, Indianapolis, IN 46268

Improvements in the efficacy and spectrum of spinosad, a novel fermentation derived insecticide, has long been a goal. As a large complex, fermentation product, identifying specific modifications to spinosad likely to result in improved activity was difficult since most modifications decreased activity. Along the path to spinetoram (DE-175), a variety of approaches to spinosyn QSAR were examined including multiple regression, CoMFA and others, all unsuccessful. However, application of artificial neural networks (ANN) to the spinosyn QSAR problem identified new synthetic directions for improved activity, which subsequent synthesis and testing confirmed. This information coupled with other information on spinosyn structure activity relationships directly lead to the discovery of spinetoram. Scheduled for launch in late 2007, spinetoram provides both improved efficacy and expanded spectrum while maintaining the exceptional environmental and toxicological profile already established for the spinosyn chemistry. Details of the ANN-based QSAR and subsequent identification of spinetoram will be discussed.

WEDNESDAY AFTERNOON

COMP - Phil Magee Memorial Symposium -- QSAR Reborn Methods II
Boston Convention and Exhibition Center (BCEC) 156C
John H. Block, Bob Clark, Organizers
1:30 414 A new paradigm for pattern recognition of drugs
Maria A. Grishina, maria_grishina@csu.ru, Vladimir Potemkin, pva@csu.ru, and Elena S. Pereyaslavskaya. chemical department, Chelyabinsk State University, Ul. Br. Kashirinych, 129, Chelyabinsk, 454021, Russia

Methods of pattern recognition play an important role in analysis and prognosis of biological activity. The most of existing methods of pattern recognition doesn't consider the conformational state of biological active molecules, electronic structure etc. Therefore, in this work a new paradigm for pattern recognition of biological active compounds that takes into account the problems of the existing methods has been suggested. The method is established on the combination of two 3D QSAR algorithms BiS/MC and ConGO which will be described in the presentation. The sets of anti-tumor, anti-inflammatory etc. drugs have been considered within this approach. It has been shown that the suggested paradigm allows to recognize active drug molecules with quality not less than 0.90.

2:00 415 Intrinsic descriptors
George D. Purvis III, gpurvis@CACheSoftware.com, Scigress Development, Biosciences Group, Fujitsu, 15244 NW Greenbrier Pkwy, Beaverton, OR 97006

Non-linear effects in structure-activity relationships are sometimes incorporated by transforming the descriptors with mathematical functions such as the square root or logarithm. Less commonly products of two descriptors are used to account for cross dependencies. We have found that simple division of descriptors by molecular weight creates "intrinsic descriptors" that are independent of the weight of chemical described. Intrinsic descriptors are frequently selected as the best descriptors in our automated QSARs. This talk will show examples of their appearance and suggest a physical explanation as to why they are preferred.

2:30 416 The development of novel fragment descriptors of molecular structure using frequent common subgraph mining approach: applications to QSAR and protein structure function relationship modeling
Alexander Tropsha, alex_tropsha@unc.edu, Laboratory for Molecular Modeling, School of Pharmacy, University of North Carolina at Chapel Hill, CB # 7360, Beard Hall, School of Pharmacy, Chapel Hill, NC 27599-7360

We discuss a novel approach to generating fragment-based molecular descriptors. The molecules are represented by labeled undirected graph. As applied to organic molecules the nodes are atoms labeled by their chemotypes and edges are bonds; for protein structures the nodes are residue Cα atoms linked by physical proximity edges. Fast Frequent Subgraph Mining (FFSM) is used to find common chemical fragments (subgraphs) that occur in at least a subset of all molecules in a dataset. The collection of frequent subgraphs forms a dataset-specific descriptor set. We present examples of application of these novel fragment descriptors in QSAR modeling of several chemical datasets of biologically active molecules. Concurrently we discuss the application of the same methodology to identifying protein family specific residue patterns. This study presents an example of expanding QSAR-like approaches towards novel areas such as structural bioinformatics and highlights the breadth of QSAR modeling and the legacy of its pioneers such as Phil Magee.

Intermission
3:20 417 Molecular topology as a tool for the design of new drugs
Jorge Galvez Sr., jorge.galvez@uv.es, Department of Physical Chemistry. Faculty of Pharmacy, University of Valencia, Avenida V.A. Estelles s.n, 46100-Burjasot (Valencia), Spain

The efficiency of molecular topology in the design and selection of new chemical compounds and particularly new drugs has been solidly presented and demonstrated in recent years. Literature sources account for many discoveries of new leads in different therapeutic areas so illustrating , beyond any reasonable doubt, such efficiency. Since the procedure is commonly based on the search of topological patterns of activity by similarity with known drugs but also yields non-obvious (dissimilar chemical structures and classes) , the formalism has been criticised for appearing to act as a black box compared to the conventional methods based on the knowledge of the drug-receptor interaction. Procedures based upon such interaction are frequently referred to as rational drug design. The aim here is to show that although both approaches are alternative, they are also complementary; thus, examples are shown in which the topological characterization of the isolated drugs clearly agrees with what is expected from the actual knowledge of the drug-biological target interaction, which in turn may also be depicted topologically in different ways. Such interaction should be better named as targeted –rather than rational- drug design.

3:50 418 Evaluation of descriptors and classification schemes to predict drug metabolism in terms of chemical information
John H. Block, John.Block@oregonstate.edu, Department of Pharmaceutical Sciences, Oregon State University, College of Pharmacy, Corvallis, OR 97331 and Douglas Henry, BIOSAR Research, San Leandro, CA.

Using a small database of defined substrates in humans for cytochrome P450 mixed function oxidases, a series of descriptors were evaluated with respect to how well they correctly classified substrates. The descriptors ranged from structural keys to topological to stereochemical to electronic. A variety of classification schemes were examined in terms of their ability to point out which descriptors are important for predicting the cytochrome P450 specificity for a substrate. Results illustrate the relative effectiveness of the various kinds of descriptors and classification methods, as well as the value of using as well-defined a data set as possible.

4:20 419 Gaussian processes: A method for automatic QSAR and ADME modelling
Olga Obrezanova, olga.obrezanova@glpg.com1, Joelle MR. Gola, joelle.gola@glpg.com2, and Matthew D. Segall, matthew.segall@glpg.com2. (1) In Silico ADMET, BioFocus DPI, 127 Cambridge Science Park, Milton Road, CB4 0GD, Cambridge, United Kingdom, (2) ADMET, BioFocus DPI, 127 Cambridge Science Park, Milton Road, CB4 0GD, Cambridge, United Kingdom

We will discuss the application of the Gaussian Processes method to predictive QSAR modelling of Absorption, Distribution, Metabolism and Excretion (ADME) properties. The method has overcome many of the problems of existing QSAR modelling techniques and is sufficiently robust to enable automatic model generation - one of the demands of modern drug discovery. The method is suitable for modelling nonlinear relationships, does not require subjective determination of model parameters, such as variable importance or network architectures, is inherently resistant to overtraining, and has an ability to select important descriptors. The method is based on a Bayesian probabilistic approach. Originating in the machine learning field, it has not yet been widely used in QSAR or ADME modelling. We will show application of Gaussian Processes to modelling of several ADME properties, discuss how the method is used as part of an automatic process and compare it with other modelling techniques.

WEDNESDAY AFTERNOON

COMP - Phil Magee Memorial Symposium -- QSAR Reborn Methods II
Boston Convention and Exhibition Center (BCEC) 156C
John H. Block, Bob Clark, Organizers
1:30 414 A new paradigm for pattern recognition of drugs
Maria A. Grishina, maria_grishina@csu.ru, Vladimir Potemkin, pva@csu.ru, and Elena S. Pereyaslavskaya. chemical department, Chelyabinsk State University, Ul. Br. Kashirinych, 129, Chelyabinsk, 454021, Russia

Methods of pattern recognition play an important role in analysis and prognosis of biological activity. The most of existing methods of pattern recognition doesn't consider the conformational state of biological active molecules, electronic structure etc. Therefore, in this work a new paradigm for pattern recognition of biological active compounds that takes into account the problems of the existing methods has been suggested. The method is established on the combination of two 3D QSAR algorithms BiS/MC and ConGO which will be described in the presentation. The sets of anti-tumor, anti-inflammatory etc. drugs have been considered within this approach. It has been shown that the suggested paradigm allows to recognize active drug molecules with quality not less than 0.90.

2:00 415 Intrinsic descriptors
George D. Purvis III, gpurvis@CACheSoftware.com, Scigress Development, Biosciences Group, Fujitsu, 15244 NW Greenbrier Pkwy, Beaverton, OR 97006

Non-linear effects in structure-activity relationships are sometimes incorporated by transforming the descriptors with mathematical functions such as the square root or logarithm. Less commonly products of two descriptors are used to account for cross dependencies. We have found that simple division of descriptors by molecular weight creates "intrinsic descriptors" that are independent of the weight of chemical described. Intrinsic descriptors are frequently selected as the best descriptors in our automated QSARs. This talk will show examples of their appearance and suggest a physical explanation as to why they are preferred.

2:30 416 The development of novel fragment descriptors of molecular structure using frequent common subgraph mining approach: applications to QSAR and protein structure function relationship modeling
Alexander Tropsha, alex_tropsha@unc.edu, Laboratory for Molecular Modeling, School of Pharmacy, University of North Carolina at Chapel Hill, CB # 7360, Beard Hall, School of Pharmacy, Chapel Hill, NC 27599-7360

We discuss a novel approach to generating fragment-based molecular descriptors. The molecules are represented by labeled undirected graph. As applied to organic molecules the nodes are atoms labeled by their chemotypes and edges are bonds; for protein structures the nodes are residue Cα atoms linked by physical proximity edges. Fast Frequent Subgraph Mining (FFSM) is used to find common chemical fragments (subgraphs) that occur in at least a subset of all molecules in a dataset. The collection of frequent subgraphs forms a dataset-specific descriptor set. We present examples of application of these novel fragment descriptors in QSAR modeling of several chemical datasets of biologically active molecules. Concurrently we discuss the application of the same methodology to identifying protein family specific residue patterns. This study presents an example of expanding QSAR-like approaches towards novel areas such as structural bioinformatics and highlights the breadth of QSAR modeling and the legacy of its pioneers such as Phil Magee.

Intermission
3:20 417 Molecular topology as a tool for the design of new drugs
Jorge Galvez Sr., jorge.galvez@uv.es, Department of Physical Chemistry. Faculty of Pharmacy, University of Valencia, Avenida V.A. Estelles s.n, 46100-Burjasot (Valencia), Spain

The efficiency of molecular topology in the design and selection of new chemical compounds and particularly new drugs has been solidly presented and demonstrated in recent years. Literature sources account for many discoveries of new leads in different therapeutic areas so illustrating , beyond any reasonable doubt, such efficiency. Since the procedure is commonly based on the search of topological patterns of activity by similarity with known drugs but also yields non-obvious (dissimilar chemical structures and classes) , the formalism has been criticised for appearing to act as a black box compared to the conventional methods based on the knowledge of the drug-receptor interaction. Procedures based upon such interaction are frequently referred to as rational drug design. The aim here is to show that although both approaches are alternative, they are also complementary; thus, examples are shown in which the topological characterization of the isolated drugs clearly agrees with what is expected from the actual knowledge of the drug-biological target interaction, which in turn may also be depicted topologically in different ways. Such interaction should be better named as targeted –rather than rational- drug design.

3:50 418 Evaluation of descriptors and classification schemes to predict drug metabolism in terms of chemical information
John H. Block, John.Block@oregonstate.edu, Department of Pharmaceutical Sciences, Oregon State University, College of Pharmacy, Corvallis, OR 97331 and Douglas Henry, BIOSAR Research, San Leandro, CA.

Using a small database of defined substrates in humans for cytochrome P450 mixed function oxidases, a series of descriptors were evaluated with respect to how well they correctly classified substrates. The descriptors ranged from structural keys to topological to stereochemical to electronic. A variety of classification schemes were examined in terms of their ability to point out which descriptors are important for predicting the cytochrome P450 specificity for a substrate. Results illustrate the relative effectiveness of the various kinds of descriptors and classification methods, as well as the value of using as well-defined a data set as possible.

4:20 419 Gaussian processes: A method for automatic QSAR and ADME modelling
Olga Obrezanova, olga.obrezanova@glpg.com1, Joelle MR. Gola, joelle.gola@glpg.com2, and Matthew D. Segall, matthew.segall@glpg.com2. (1) In Silico ADMET, BioFocus DPI, 127 Cambridge Science Park, Milton Road, CB4 0GD, Cambridge, United Kingdom, (2) ADMET, BioFocus DPI, 127 Cambridge Science Park, Milton Road, CB4 0GD, Cambridge, United Kingdom

We will discuss the application of the Gaussian Processes method to predictive QSAR modelling of Absorption, Distribution, Metabolism and Excretion (ADME) properties. The method has overcome many of the problems of existing QSAR modelling techniques and is sufficiently robust to enable automatic model generation - one of the demands of modern drug discovery. The method is suitable for modelling nonlinear relationships, does not require subjective determination of model parameters, such as variable importance or network architectures, is inherently resistant to overtraining, and has an ability to select important descriptors. The method is based on a Bayesian probabilistic approach. Originating in the machine learning field, it has not yet been widely used in QSAR or ADME modelling. We will show application of Gaussian Processes to modelling of several ADME properties, discuss how the method is used as part of an automatic process and compare it with other modelling techniques.

THURSDAY AFTERNOON

COMP - Phil Magee Memorial Symposium -- QSAR Reborn Applications II
McCormick Place Lakeside E353 C, Level 3
John H. Block, Bob Clark, Organizers
1:00 483 Random forest ensembles applied to MLSCN screening data for toxicity prediction and feature selection
Rajarshi Guha, rguha@indiana.edu, School of Informatics, Indiana University, 1130 Eigenmann Hall, 1900 E 10th Street, Bloomington, IN 47406 and Stephan Schurer, sschurer@scripps.edu, Chemical Informatics, The Scripps Research Institute Florida, 5353 Parkside Drive, RF-A, Jupiter, FL 33458

We present a study describing the use of a random forest ensemble to predict the cytotoxicity of MLSCN screening data. The models were built using a training set taken from MDL ToxNet using LeadScope and 1052 bit BCI fingerprints as the feature set. The real valued toxicity data was discretized using a cutoff and the class assignments were used to build the ensemble, which exhibited 85% accuracy for the prediction set. We also investigated the use of the models for the purposes of feature selection. With all 1052 bits, a Naive Bayes classifier exhibited 74% accuracy on the prediction set. Using the random forest to select a set of important bits allowed us to achieve 73% accuracy using only 171 bits. We also analyzed the most important fragments and evaluated their frequency of occurence and they correspond well with previously known toxic fragments and could be used as a measure of domain applicability for the ensemble when applied to the MLSCN screening data.

1:30 484 On the importance of topological descriptors in understanding structure-property relationships in QSAR and QSPR models
David T. Stanton, stanton.dt@pg.com, Modeling & Simulations - CADMol Group, Procter & Gamble, Miami Valley Laboratories, 11810 East Miami River Road, Cincinnati, OH 45252

It has been generally observed in our work that topological descriptors play an important and often key role in many QSAR and QSPR models we have developed. This is found to be true even when a broad selection of descriptor types is evaluated. These descriptors do not only provide the means to generate a good fit to the observed data used to train the models, but they also provide information that is needed to generate a clear physical interpretation of the underlying structure-activity or property relationships. Examples of this will be presented using two properties that have well understood mechanisms, skin penetration and critical micelle concentration.

2:00 485 Predicting allergic contact dermatitis: alternative statistical approaches to chemical classification
Subhash C. Basak, sbasak@nrri.umn.edu1, Denise Mills, dmills@nrri.umn.edu1, Brian D. Gute, bgute@nrri.umn.edu1, and Douglas M. Hawkins, dhawkins@umn.edu2. (1) Center for Water and the Environment, Natural Resources Research Institute, University of Minnesota, 5013 Miller Trunk Hwy, Duluth, MN 55811, (2) School of Statistics, University of Minnesota, 313 Ford Hall, 224 Church Street SE, Minneapolis, MN 55455

Allergic contact dermatitis (ACD) is believed to be the most prevalent form of immunotoxicity found in humans, the adverse effect arising out of the interaction of immunoregulatory cytokines and discrete subpopulations of T lymphocytes. As such, ACD is a major impediment to the development of new cosmetics, personal hygiene products and topical medications. Calculated molecular descriptors, based solely on chemical structure, may be used to develop models for the prediction of ACD. Such models can be used to evaluate new chemicals, synthesized or hypothetical. In the current study, various statistical modeling approaches, including ridge linear discriminant analysis and tailored similarity, have been used to classify chemicals with respect to ACD. Results obtained from the different modeling methods will be discussed.

2:30 486 A comparison of the chemical properties of drugs and FEMA/FDA notified GRAS chemical compounds used in the food industry
DG. Sprous, dennis.sprous@redpointbio.com and FR. Salemme, frs@redpointbio.com. Chemistry, Redpoint Bio, 1 Graphics Drive, Ewing, NJ 08628

The molecular properties for Generally Recognized As Safe (GRAS) compounds (food/flavoring approved) are compared to marketed drugs and used to develop a QSAR model. It was observed that log(P) and log(S), which provide computed estimates of compound solubility in organic and aqueous solvents respectively, have significant overlap in the two populations. On the whole, GRAS compounds are seen to be more flexible, smaller, and composed of a more restricted set of elements than marketed drugs. A multivariable binary Quantitative Structure-Activity Relationship (QSAR) model can correctly identify 94% of the GRAS compounds and 92% of the pharmaceutical compounds. The performance of the model was such that training sets comprised of as little as 10% of the whole dataset could predict the 90% reserved as a test set with an accuracy >90%. To summarize, the majority of the historical GRAS compounds are not “druglike”, and easily distinguished from pharmaceuticals.

2:50 Intermission
3:10 487 A QSAR model for hERG based on mulitple 1-D alignments
David J. Diller, ddiller@pharmacop.com, Molecular Modeling, Pharmacopeia, Box 5350, Princeton, NJ 08543-5350

The 1D-molecular representation (Dixon & Merz, JMC 2001) is a misnomer in that it contains more structural information than most commonly used 2D descriptors. Furthermore, it does not suffer from the conformational ambiguities of traditional 3D methods. Thus the 1D representation has great promise to fill the void where 2D descriptors are not sufficiently rich and 3D methods are not practical such as large or noisy data sets. We describe techniques to generate 1D multiple alignments of molecules with similar biological activity. These alignments effectively isolate biologically critical regions much like a multiple sequence alignment isolates conserved motifs within a protein family. Particular emphasis in this presentation will be on our efforts to develop a QSAR model to predict hERG inhibition. Finally, we will discuss the implications of this work on 3 dimensional methods.

3:40 488 Modeling fluorophilicity: A hybrid method
Marvin Charton, mcharton@pratt.edu, Pratt Institute, 200 Willougby Avenue, Brooklyn, NY 11205

Modeling fluorophilicity: A hybrid method. There has been considerable interest in recent years in modeling fluorophilicity, lnPF, defined as the natural logarithm of the partition coefficient of a compound between perfluromethylcyclohexane and toluene. Approaches used to model this property include the use of surface area parameters, the Abraham modification of the Taft-Kamlet equation for solvent effects, and a completely empirical model that is based on the choice of a mix of available parameters that best fit the data. Here we assume that fluorophilicity depends on the difference in intermolecular forces between a compound and solvent 1 and those between the compound and solvent 2. Our model represents the intermolecular forces of interest by a count of the number of groups X of that type in the molecule and by the sum of the polarizabilities of each group Y that is not represented by a term in the number of groups present in the molecule. Correlations of ln PF with this equation have been carried out. The regression equation obtained on the exclusion of 8 outliers is: ln PF = 0.309 (± 0.00965) n (CF2) + 0.637 (± 0.0369) n (CF3) – 0.201 (± 0.0140)n (CH2) - 1.73(± 0.0969)(n Ph / Pn) -1.03 (± 0.224) n(OH) – 6.94 (± -0.857) α - 0.734 (± 0.163) 100R2, 95.83; A100R2, 95.64; F, 409.9; Sest, 0.555; S0, 0.211; Ndp, 114.

4:10 489 Inductive descriptors: Ten successful years in QSAR
Artem Cherkasov, artc@interchange.ubc.ca, Division of Infectious Diseases, University of British Columbia, 2733 Heather Str, Vancouver, BC V5Z 3J5, Canada

In a series of recent studies we reported the development of ′inductive′ QSAR descriptors which are related to atomic electronegativity, covalent radii and intramolecular distances, and that have all been derived from the original equations for steric and inductive substituent constants we have published 10 years ago. Since that time a variety of related QSAR parameters including ′inductive′ electronegativity, ′inductive′ partial charge and ′inductive′ analogues of chemical hardness and softness have been introduced. To date, 50 global and local ′inductive′ descriptors have been elaborated; possible interpretation of their physical meaning has been suggested by considering a neutral molecule as an electrical capacitor formed by charged atomic spheres. Consequent studies demonstrated successful application of ′inductive′ QSAR parameters in quantification of antibacterial activity of organic compounds and peptides, in QSAR modelling of metabolic and drug-like substances, in comparative docking studies and in the discovery of novel drug leads.

THURSDAY MORNING

AGRO - Agricultural Biomass, Biobased Products, and Biofuels
McCormick Place South S103D, Level 1
Organized by: Justin R. Barone, Cathleen J. Hapeman, Joseph H. Massey, James N. Seiber
Presiding: Joseph H. Massey
8:30 Introductory Remarks
8:35 195 Properties of biodegradable feather keratin polymers
Justin R. Barone, jbarone@vt.edu, Biological Systems Engineering, Virginia Polytechnic Institute and State University, 0303 Seitz Hall, Blacksburg, VA 24061

The properties of a recent class of polymers created from poultry feather keratin are described. A “cradle to grave” approach is employed and production of the polymers, uses, and finally biodegradation characteristics will be described. Properties are dependent on the amino acid composition of the feather keratin and modification of the amino acids to elicit new properties. Melt-state properties of the feather keratin such as viscosity can be modified with the use of reducing agents such as sodium sulfite and lubricants such as poultry fat. Solid-state properties can be modified using divalent transition metal ions to affect stiffness and smell.

9:00 196 Biopolymers from polylactic acid and milk proteins
Charles Onwulata, Charles.Onwulata@ars.usda.gov and Peggy Tomasula. Dairy Processing and Products Research Unit, USDA-Agricultural Research Service, Eastern Regional Research Center, 600 E. Mermaid Lane, Wyndmoor, PA 19038

Polylactic acid (PLA) is a commercially-available biodegradable polymer derived from lactic acid and is used in many products as an alternative to petrochemical-derived polymers. However, the physical properties limit its use in many applications. Using dairy proteins to substitute for portions of PLA in a formulation may extend its use and prevent shortages of PLA. This work reports on the mechanical and thermal properties of composites made from PLA substituted with starch-whey concentrates and casein blends (DPB). The blends were extruded under the following conditions: mass flow rates (27 to 102 g/min), solids feed rates (0.43 to 2.85 g/sec), moisture (30 to 75%); extrusion melt profiles were: 75, 90, 100, 100, 90, 80°C; and molding conditions at 200°C and 12,000 psi. The physical properties of the extruded DBP were moisture 14-18%, peak tensile strength 4.5 mPa, thickness 3.9 mm, elongation at break 45%, and storage modulus 5.0 mPa. Injection molded product peak melt temperature shifted down in order: PLA 132.8°C, DBP/PLA (10/90%) 149.4/130.3°C, DBP/PLA (20/80%) 148.8/128.2°C, indicating softening of PLA when combined with DBP. Dairy proteins, whey and casein, may provide an advantage by lowering the peak molding temperature of PLA allowing for more biomaterials to be used. Further work is needed to improve the extrusion compounding and miscibility of this high-temperature melting PLA and high-temperature burning-DBP blend.

9:25 197 Extraction and electrospinning of zein extracted from corn gluten meal using acetic acid
G. W. Selling, sellingg@ncaur.usda.gov and Kristen K. Woods, woodskk@ncaur.usda.gov. Plant Polymer Research, USDA-Agricultural Research Service, National Center for Agricultural Utilization Research, 1815 N. University St., Peoria, IL 61604

It has been demonstrated that zein fibers can be produced using the electrospinning technique. Fibers electrospun from acetic acid solution under suitable conditions provide fibers with a more consistent morphology (round, 0.5-2.0μ fibers) compared to fibers produced form aqueous ethanol solutions. Spinning continuity of zein acetic acid solutions is significantly improved as well. Commercial zein is produced via extraction of corn gluten meal using aqueous alcohol solvents. In order to better model a possible commercial process, acetic acid was used to extract zein from corn gluten meal. It was found that acetic acid removes more protein than the more traditional solvent systems. The impact of time, temperature, and other solvents on extractability will be presented. The zein acetic acid solution obtained from corn gluten meal was successfully electrospun producing fibers of similar quality to that produced from commercial zein.

9:50 198 Improved physical properties of zein using glyoxal as a crosslinker
Kristen K. Woods, woodskk@ncaur.usda.gov and Gordon W. Selling, sellingg@ncaur.usda.gov. Plant Polymer Research, USDA-Agricultural Research Service, National Center for Agricultural Utilization Research, 1815 N. University St., Peoria, IL 61604

The effect of the crosslinkers, glyoxal, methylglyoxal, and formaldehyde, on the physical properties of zein films was studied. Crosslinker concentrations varied from 0.3-6% by zein weight. Films crosslinked with glyoxal and formaldehyde showed a significant increase in tensile strength under certain pH conditions. Films of glyoxal reactions conducted at basic pH gave the highest overall tensile strength, with a 52% increase compared to the control film. Formaldehyde films had improved tensile strength when reacted at acidic or neutral pH. Methylglyoxal had no effect on the tensile strength of zein films. Zein films crosslinked with glyoxal or formaldehyde were found to swell, rather than degrade, when placed in three compatible solvents. Films crosslinked with glyoxal were resistant to boiling water. Denaturing gel electrophoresis of glyoxal and formaldehyde reactions showed the presence of high molecular weight moieties when compared to control reactions.

10:15 Intermission
10:30 199 Arthropod repelling constituents from a southern folk remedy: Investigations of the American beautyberry, Callicarpa americana
Charles L. Cantrell, clcantr1@olemiss.edu1, Charles T. Bryson, cbryson@ars.usda.gov2, Stephen O. Duke, sduke@olemiss.edu1, Jerome A. Klun, klunj@ba.ars.usda.gov3, and John F. Carroll, jcarroll@anri.barc.usda.gov4. (1) Natural Products Utilization Research Unit, USDA-Agricultural Research Service, P.O. Box 8048, University, MS 38677, (2) Southern Weed Science Research Unit, USDA-Agricultural Research Service, P.O. Box 350, Stoneville, MS 38776, (3) Chemicals Affecting Insect Behavior Laboratory, USDA-Agricultural Research Service, Beltsville Agricultural Research Center, 10300 Baltimore Ave., Bldg. 007, Rm. 301, BARC-West, Beltsville, MD 20705, (4) Animal Parasitic Diseases Laboratory, USDA-Agricultural Research Service, Beltsville Agricultural Research Center, Bldg. 1040, BARC-East, Beltsville, MD 20705

Based on botanical lore of insect repellent properties, essential oil extracts from Callicarpa americana and Callicarpa japonica were investigated. Bioassay-guided fractionation of C. americana extracts using the yellow fever mosquito, Aedes aegypti, led to the isolation of α-humulene, humulene epoxide II, and intermedeol, and a newly isolated terpenoid (callicarpenal). Similar work involving C. japonica resulted in the isolation of an additional compound, spathulenol, as well as the four compounds isolated from C. americana. Heretofore, 13,14,15,16-tetranor-3-cleroden-12-al, callicarpenal, has never been identified from natural sources. In bite-deterrent studies spathulenol, intermedeol, and callicarpenal showed significant bite-deterring activity against Aedes aegypti and Anopheles stephensi. The repellency of callicarpenal and intermedeol against workers of red imported fire ants, Solenopsis invicta and black imported fire ants, Solenopsis richteri will also be reported. In addition, callicarpenal and intermedeol were evaluated in laboratory bioassays for repellent activity against host-seeking nymphs of the blacklegged tick, Ixodes scapularis, and lone star tick, Amblyomma americanum and results will be presented.

10:55 200 Biobased Herbicides
Franck E. Dayan, fdayan@olemiss.edu and Stephen O. Duke, sduke@olemiss.edu. Natural Products Utilization Research Unit, USDA-Agricultural Research Service, P.O. Box 8048, University, MS 38677

Herbicides amount to more than half of all agricultural pesticides used in the developed world. The availability of insect and disease resistant transgenic crops that reduces the reliance on other synthetic pesticides will contribute to a further increase in the relative proportion of herbicides in the pesticide market. The use of biobased herbicides, either in the form of phytotoxic natural products applied conventionally or in the form of allelopathic crops that would repress the growth of weeds by releasing their own phytotoxins, can potentially be used as low input alternatives. Simple biobased herbicides such as acetic acid, fatty acids, and oils are commonly used as alternative to synthetic compounds. However, the most widely-used, natural herbicide is the microbial secondary metabolite bialaphos (a natural form of phosphinothricin or glufosinate). This glutamine synthase inhibitor was first introduced in Japan in 1984 and is now used under one form or another in more than 40 countries. Its synthetic counterpart is most commonly used on genetically engineered glufosinate-resistant crops that either express the bar or PAT genes. Other natural products have served as templates for the development of commercial analogues. For examples, the p-hydroxyphenylpyruvate dioxygenase inhibitors, sulcotrione and mesotrione, were derived from leptospermone, a natural triketone, isolated from bottlebrush. Allelopathy is an often-overlooked approach to reduce synthetic pesticide output. However, work in allelopathic rice demonstrated that excellent weed control could be achieved using half the normal rate of herbicide. Selection of highly allelopathic crop varieties, either through traditional breeding or using genetic engineering techniques, may also provide novel and low input environmentally-friendly approaches to weed control.

11:20 201 Single-use, disposable food containers: Starch-based alternatives to petroleum-based plastics
Gregory M. Glenn, gmg@pw.usda.gov, Charles N. Ludvik, Artur P. Klamczynski, arturk@pw.usda.gov, William J. Orts, Syed H. Imam, and Delilah Wood. Bioproduct Chemistry and Engineering Research, USDA-Agricultural Research Service, Western Regional Research Center, 800 Buchanan Street, Albany, CA 94710

The use of valuable petroleum resources to make single-use, disposable plastic foodservice containers has raised concerns among environmental and consumer groups. Billions of single-use food service containers are used each year in the U.S. alone to dispense beverages and serve food. Starch is an abundant, inexpensive, renewable resource derived primarily from cereal and tuber crops. A baking technology has been developed to produce degradable food containers with functional properties similar to those of polystyrene foam products. The product is a composite material consisting of a vapor barrier film, starch, fiber, and other minor ingredients. Starch/fiber foam composites have also been made using extrusion technology. The extruded composite materials containing fiber have improved tensile strength and modulus and are more stable during aging than materials that do not contain fiber.

11:45 202 Incorporation of bacteriocin in edible pectin films for antimicrobial packaging
LinShu Liu, lsliu@errc.ars.usda.gov1, Tony Jin, Tony.Jin@ras.usda.gov2, Cheng-Kung Liu, Chengkung.Liu@ars.usda.gov3, Kevin B. Hicks, kevin.hicks@ars.usda.gov1, Amar K. Mohanty, mohantya@msu.edu4, Rahul Bhardwaj, bhardwa5@msu.edu4, and Manjusri Misra, misraman@egr.msu.edu5. (1) Crop Conversion Science and Engineering Research Unit, USDA-Agricultural Research Service, Eastern Regional Research Center, 600 E. Mermaid Lane, Wyndmoor, PA 19038, (2) Food Safety Intervention Technology Research Unit, USDA-Agricultural Research Service, Eastern Regional Research Center, 600 E. Mermaid Lane, Wyndmoor, PA 19038, (3) Fats, Oils and Animal Co-Products Research Unit, USDA-Agricultural Research Service, Eastern Regional Research Center, 600 E. Mermaid Lane, Wyndmoor, PA 19038, (4) School of Packaging, Michigan State University, 130 Packaging Building, East Lansing, MI 48824, (5) Composite Materials and Structures Center, Michigan State University, 2100 Engineering Building, East Lansing, MI 48824

Edible, antimicrobial films were prepared by extrusion blown film process. Blends of pectin, fish skin gelatin or soybean flour protein, and a bacteriocin, nisin were chosen to prepare the edible, antimicrobial films. The blends were prepared using a ZSK-30 twin-screw extruder. The compounded pellets were then used to prepare blown film using a Killion-KLB-100 extruder. The films retained activity against the indicator bacterial, L. plantarum. The resulting films also possess appropriate mechanical properties for food packaging.

12:10 Lunch Break

 

 

Publishing software