#240 - Co-sponsored Sessions

ACS National Meeting
Fall, 2010
Boston, MA

SUNDAY

COMP - Tautomers and Biology Computer Handling of Tautomers
BCEC 157 C
Organized by: Terry Stouch, Yvonne Martin; Presiding: Terry Stouch
8:30 13 Tautomerism in chemical information management systems
Wendy A. Warr M.A., D. Phil. Wendy Warr & Associates, Holmes Chapel, Cheshire, United Kingdom

Tautomerism has an impact on many of the processes in a chemical information management system including novelty checking during registration into chemical structure databases; storage of structures; exact and substructure searching in chemical structure databases; and depiction of structures retrieved by a search. For this talk the approaches taken by a great many different software vendors and database producers have been compared. Since it is important to take account of the nature of the database and the process for which it is designed, and the user requirements vary, it is dangerous to lay down the law about what is right and wrong. The comparison is nevertheless of considerable interest.

9:00 14 Tautomerism in large databases
Dr. Markus Sitzmann, Dr. Wolf-Dietrich Ihlenfeldt, Dr. Marc C Nicklaus. Chemical Biology Laboratory, Center for Cancer Research, National Cancer Institute, National Institutes of Health, DHHS, NCI-Frederick, Frederick, MD, United States; Xemistry GmbH, Königstein, Germany

We are reporting on a comprehensive tautomerism analysis of one of the largest currently existing sets of real (i.e. not computer-generated) compounds. We used the Chemical Structure DataBase (CSDB) of the NCI CADD Group, an aggregated collection of over 150 small-molecule databases totaling 103.5 million structure records. Tautomerism was found to be possible for more than 2/3 of the unique structures in CSDB. A total of 680 million tautomers were calculated from the original structure records. Tautomerism overlap within the same individual database (i.e. at least one other entry was present that was really only a different tautomeric representation of the same compound) was found at an average rate of 0.3% of the original structure records, with values as high as nearly 2% for some of the databases in CSDB. Tautomeric overlap across all constituent databases in CSDB was found for nearly 10% of the records in the collection.

9:30 15 Tautomerism in drug discovery
Bahaa El-Dien M. El-Gendy, Prof. Alan R. Katritzky PhD, Dr. C. Dennis Hall PhD, Bogdan Draghici. Department of Chemistry, University of Florida, Gainesville, Florida, United States; Department of Chemistry, Benha University, Benha, Qalubia, Egypt

The influence of tautomerism on the precise structure of drugs and thus of their potential to interact in biological systems is discussed from thermodynamic and kinetic aspects. The types of tautomerism encountered in the structure of drugs in current use are surveyed together with the effect of pH, solvent polarity, and temperature.

10:00   Intermission
10:15 16 Quantitative forecasts of biological potency of moleculesthat can tautomerize
Dr. Yvonne C Martin. Martin Consulting, Waukegan, IL, United States

Whether one is using ligand-based 2D or 3D QSAR or structure-based estimates of potency of molecules, tautomerism needs to be addressed. This talk will highlight insights as to when one needs to consider tautomerism and how it can be included in potency forecasts.

10:45 17 New questions about tautomerism in cytosine: Quantum chemical and matrix isolation spectroscopic studies
Prof. Geza Fogarasi, Mr Gabor Bazso, Prof Peter G Szalay, Dr. Gyoergy Tarczay. Laboratory of Theoretical Chemistry, Institute of Chemistry, Eotvos University, Budapest, Budapest, Hungary; Laboratory of Molecular Spectroscopy, Institute of Chemistry, Eotvos University, Budapest, Budapest, Hungary

In spite of numerous studies, there is much uncertainty about tautomerism in nucleic acids and specifically cytosine. In the gas phase, form 2 dominates but DG maybe about 1 kcal/mol for both 1 and the “rare” form 3. Spectroscopic studies “see” them but in much smaller abundance. The UV spectrum is normally assigned to 1. Dimerization may also influence tautomerization.

Fig. 1. Selected isomers and a dimer of cytosine

We present infrared and UV spectroscopic measurements in Ar matrix and discuss them by MP2 and CC quantum chemical calculations, including electronic excitations. Contributions from isomers/tautomers and/or dimers to the spectra are discussed.

SUNDAY

CHED - Social Networking: The Next Generation
Seaport Hotel Seaport Ballroom A
Organized by: Laura Pence, Harry Pence
Presiding: AM: Lucille Benedict, PM: Harry Pence
8:30   Introductory Remarks
8:35 10 Construction of topical faculty learning communities by the Center for Workshops in the Chemical Sciences (CWCS) and the use of Drupal as a development platform
Dr. Cianán B. Russell, Dr. David M. Collard. School of Chemistry and Biochemistry, Georgia Institute of Technology, Atlanta, GA, United States

A new national dissemination initiative of the Center for Workshops in the Chemical Sciences (CWCS) is to develop topical faculty learning communities to further spread the adoption of innovative content and to propagate the use of good pedagogical practice in the teaching of undergraduate chemistry. CWCS has provided 88 workshops in a variety of topical areas, hosting over 1400 participants who have then used the workshop materials in a number of ways to improve undergraduate education. In this new initiative, we wish to engage workshop participants as the foundation of online communities that provide access to databases of curricular materials and pedagogies, together with the shared expertise of the group through discussion boards, blogs, etc. The Drupal platform was used to develop a flexible and adaptable interface. The process of developing this interface, and challenges associated with prototyping, assessing, and modifying our approach to the development will be discussed.

8:55 11 Ebooks: A culture shift for academic libraries?
Assisstant Professor Barbara A. Losoff. Science Library, University of Colorado, Boulder, CO, United States

The decline of print materials in academic libraries is a result of changing technology, cost, and plummeting use by patrons. This mobile, Google/YouTube/Facebook, generation acquires their information online. Images are as important as text. Librarians must ask themselves the question: in what ways are these users transforming the very definition of a book, and how can libraries support this cultural shift to digital content, and does anyone know what the book of the future will resemble?

9:15 12 Engaging student discussion: The role of a google jockey
Prof. Laura E Pence, Emily R. Greene. Department of Chemistry, University of Hartford, West Hartford, CT, United States

A challenge to the inclusion of real world applications in a course can be the students' lack of mental images to provide context. PowerPoint images are a solution in a structured lecture environment, but in a first year seminar course with an emphasis on discussion, preselected illustrations constrain the dialogue and reflect only the instructor's mental framework.

A powerful alternative solution employed a senior student embracing the role of Google Jockey, whose purpose is to search and display images from the internet as illustration or counterpoint to an ongoing discussion. The replacement of mental images with visual images enhanced the student engagement in the class and allowed the senior to have a vital, if silent, contribution to the dialogue.

9:35 13 Rip-Mix-Learn (RML): Using Google Docs to create collaborative multimodal class notes
Dr. Lucille A Benedict, Dr. Harry E Pence. Department of Chemistry, University of Southern Maine, Portland, ME, United States; SUNY College at Oneonta, Oneonta, NY, United States

Computer and internet use has become ubiquitous among college students and can be very powerful educational tools when properly incorporated into the course curriculum. The Rip-Mix-Learn (RML) approach applies students' knowledge of surfing the web with course content to create a set of collaborative class notes that incorporate multimodal representations of each concept to make the students more personally invested in the topics. To create these class notes, first-semester general chemistry course students were given a basic set of notes each week in Google Docs focusing on the current course topics. The students' task was to annotate these documents with pictures, videos, or other representations found on the web and then write brief descriptions of how these annotations related to the specific topics. This talk will focus on the implementation, use, and advantages and drawbacks of using this RML approach in a large lecture first-semester general chemistry course.

9:55   Intermission
10:05 14 Smart phones, smart objects, and chemical education
Prof. Harry E. Pence PhD. Department of Chemistry and Biochemistry, SUNY Oneonta, Oneonta, New York, United States

The mobile phone is already changing the way we communicate, but it is also creating new ways to access information. Companies, like Google, Yelp, and Layar, are building a layer of digital information that can augment the photograph a user takes with his/her smartphone. As 2D bar codes become more popular in this country, these symbols can label an object with a URL which, in turn, can cue a smartphone or personal computer to access a web site. This means that a piece of paper can include the equivalent of a hyperlink that may lead to structural, safety, or other information. What new opportunities open up for chemical educators when smartphones offer not only portable access to a massive library of information but also a quick and convenient way to work with smart objects that are connected to the World Wide Web?

10:25 15 How community crowdsourcing and social networking is helpingto build a quality online resource for chemists
Dr Antony J Williams PhD. ChemSpider, Royal Society of Chemistry, Wake Forest, North Carolina, United States

With an intention to provide a free internet resource of chemistry related data for the community, ChemSpider provides an online database of chemical compounds, reaction syntheses and related data. Members of the community can contribute to the database via the deposition of chemical structures, synthesis procedures and analytical data. Data are also aggregated from many other depositors, at present over 400 data sources. The aggregation of data associated with over 25 million chemical compounds does not come without data quality issues. By engaging the community to curate the data the quality continues to improve on a daily basis. The presentation will provide an overview of our ongoing efforts to expand and curate the database. Using a combination of game-based and recognition systems as well as our dependence on societal giveaway by the community ChemSpider continues its path to become a high quality resource and foundation for the semantic web for chemistry.

10:45 16 Chemistry of social media
Scott Jensen. American Chemistry Council, Arlington, VA, United States

The rise of Web 2.0 or social media has created a new frontier in communicating with a variety of audiences on issues directly related to chemistry and how it impacts their lives. Blogs, Twitter, Facebook and even YouTube have created new opportunities to disseminate information in a very direct and targeted fashion. At the same time, social media tools can allow for dialogue or a forum for debate.

This presenttion will discuss how The American Chemistry Council's Chlorine Chemistry Division has entered this new frontier and utilized Web 2.0 tools to engage and educate a range of audiences from policy makers to the general public regarding chlorine related issues.

11:05   Lunch
1:30   Introductory Remarks
1:35 43 Communicating organic chemistry through the internet: Global learning communities
Prof. Philip A Janowicz. Department of Chemistry and Biochemistry, California State University - Fullerton, Fullerton, CA, United States

The power of broadband internet has allowed for instant communication across the world, and opportunities for distance education have been greatly enhanced. In the spring of 2009, students from Peking University in Beijing, China, joined in with students from the University of Illinois at Urbana-Champaign in synchronous discussion sessions for organic chemistry. In the fall of 2009, students from Lahore University of Management Sciences in Lahore, Pakistan, joined the synchronous discussions. Experiences during these semesters will be shared along with an outlook for the future.

1:55 44 Focusing CENtral Science: An overview of C&EN's redesigned blog portal and its usefulness to educators
Editor, C&EN Online Rachel Pepling. Chemical & Engineering News, Washington, DC, United States

In March 2010, Chemical & Engineering News magazine relaunched its blog, C&ENtral Science (http://centralscience.org), as a portal to several content-focused blogs meant for different audiences (and dropped the "&" along the way). This overview will discuss why that decision was made and how the new CENtral Science can be a valuable resource to chemical educators.

2:15 45 Chemistry blogging: From literature to controversy to community to...
Aaron D. Finke. Department of Chemistry, University of Illinois, Urbana-Champaign, Urbana, IL, United States

This talk will focus on my experiences as the co-author of a popular chemistry blog, Carbon-Based Curiosities. Initially, I started blogging as a means to keep up with the literature by forcing myself to read and summarize papers I enjoyed or found interesting. However, as the blog progressed, the audiences increased, and my interests diverged, I found myself using the chemistry blogosphere as a means to a different end, one in which one's personal creative energies, even those that tended to diverge far from chemistry, could be applied to ideas, problems, and controversies in current chemical research. In this personal account, I will draw from not only my own experiences in blogging, but also from others across the chemical blogosphere, and show how this small community has already made some big waves.

2:35 46 Blogging: Ego trip, or sound science? Its role in chemical education and research
Prof Henry S Rzepa D. Sc.. Chemistry, Imperial College, London, United Kingdom

Blogs evolved as a personal statement by an individual, but in science and chemistry have now emerged as a fascinating new way of reviewing the correctness of previously reviewed traditional published science. I will argue they can be much more. In chemical education, they enable the chemist to communicate their accumulated expertise in an accessible manner to both the educational and as it happens the research communities, and indeed to present new and original science that might otherwise be lost. The speaker has posted more than 50 blogs in a year of activity, and a number of these have also been used to enhance taught courses. Others have morphed into published peer-reviewed articles, in traditional journals. The difference between a publication and a blog will be discussed, as well as how a blog can be enhanced with semantic attributes, harvested and aggregated and archived for the longer term.

2:55   Intermission
3:05 47 Teaching scientific communication in pharmaceutical bioinformatics education
Dr. Egon Willighagen. Department of Pharmaceutical Biosciences, Uppsala University, Uppsala, Uppland, Sweden

Communication is a central part in science. Traditionally, students are educated to access scientific literature, but communication channels are changing. The amount of literature has risen sharply, and not even established researchers can keep up with the amount of publications that appear each week. At the same time, new technologies have changed communication as we knew it, and with the introduction of the internet communication world anyone around the world has become as easy as communicating with people at the same department. Research has become so specialized, however, that peers at the same university not always are the best judges of ones work, and the international communication becomes more and more important.

In my education of students doing a 20 week research project in Pharmaceutical Bioinformatics at Uppsala University, we made the use of social websites a core part of their education. Within their projects, the students (two at the moment) report the work they do via their blog; additionally, taking advantage of the programming side of their work, the results of their experiments (source code) is submitted to a central source code repository. This is quite similar to the use of wikis for describing synthesis experiments in organic chemistry. Additionally, reusable components or examples on how their work can be used, is shared via the social MyExperiment.org website, allowing others to download the protocols the students have development, comment on them, and rate them.

The students also take part in a journal club, where we discuss related literature. Goals of these meetings is that the student learns to formulate an opinion on the paper, after which we discuss the theories behind the paper in more detail. For each discussed paper, one or two participants write up a dedicated blog post, which we mark up such that social websites like Chemical blogspace and ResearchBlogging.org can pick up the discussed literature. CiteULike.org is used to share the list of discussed papers using a dedicated hashtag.

By making the literature reviews and their progress in the 20 week project publicly available, the students engage in a scientific discussion with peers. By having parts of their work publicly available in their blog, it is easier for them to discuss issues on more targeted mailing lists for databases and software libraries they use in their own project. Using these social websites helps the student to put their scientific work in perspective, and learns them to discuss their research with other scientists around the globe.

3:25 48 Developments in chemistry resources on Wikipedia
Prof. Martin A Walker PhD. Department of Chemistry, SUNY Potsdam, Potsdam, NY, United States

In recent years, Wikipedia has become a standard information source for students and researchers alike, but its open nature tends to undermine its reliability. This presentation will explain how to use this immense resource effectively, and also describe efforts made by the Wikipedia chemistry community to address users' concerns. A collaboration with Chemical Abstracts Service has led to validation of Registry Numbers and structures, while other collaborations with ChemSpider and RSC have also brought improvements, yet much remains to be done. The presentation will close with an overview of work that is planned or under way, indicating the direction of likely future developments.

3:45 49 ChemEd DL WikiHyperGlossary
Dr. Robert Belford, Dr. Daniel Berleant PhD, Michael Bauer, Dr. John W. Moore PhD, Roger Hall. Department of Chemistry, UALR, Little Rock, AE, United States; Department of Chemistry, University of Wisconsin-Madison, Madison, WI, United States; Department of Information Sciences, UALR, Little Rock, AR, United States; MidSouth BioInformatics Center, UALR, Little Rock, AR, United States

We will present the new editing interface of the wikihyperglossary generating program being developed for ChemEd DL. We will go over the database design, present several databases, including a non-editable one with IUPAC Gold book definitions, along with several editable ones. We will then discuss our experiences in a general chemistry class where students created definitions for terms in their class textbook using textual and multimedia online resources.

4:05 50 Chempedia Lab: Group meeting on a global scale
Ph. D Richard L Apodaca. Metamolecular, LLC, La Jolla, CA, United States

Online database searches have become the information tool of choice for answering tough experimental chemistry questions. But what if it were possible to answer questions by simply asking the entire experimental chemistry community directly? What would a system that made this possible look like, and how might it work? Chempedia Lab (http://lab.chempedia.com) represents our attempt to answer these questions through a fundamentally new approach to online knowledge-gathering. This talk will discuss how traditional databases have failed the experimental chemistry community, and what Chempedia Lab might teach about the chemical information systems of the future.

SUNDAY AFTERNOON

COMP - Using Waters Explicitly in Drug Discovery Theory and Methods
BCEC 154
Organized by: Veerabahu Shanmugasundaram, Woody Sherman
Presiding: Woody Sherman
1:30   Introductory Remarks - Woody Sherman
1:35 76 Water in protein binding sites: Consequences for ligand optimization
Dr. Julien Michel, Dr. Julian Tirado-Rives, James Luccarelli, Prof. William L Jorgensen. Department of Chemistry, Yale University, New Haven, CT, United States

An efficient molecular simulation methodology, JAWS, has been developed to determine the positioning of water molecules in the binding site of a protein or protein-ligand complex. Occupancies and absolute binding free energies of water molecules are computed using a statistical thermodynamics approach. The importance of determining proper water occupancies is illustrated in Monte Carrlo/free energy perturbation calculations for ligand series that feature displacement of ordered water molecules in the binding sites of scytalone dehydratase, p38-aMAP kinase, and EGFR kinase. The change in affinity for a ligand modification is found to correlate with the ease of displacement of the ordered water molecule. For accurate results, a complete thermodynamic analysis is needed. It requires identification of the location of water molecules in the protein-ligand interface and evaluation of the free energy changes associated with their removal and with the introduction of the ligand modification. Direct modification of the ligand in free-energy calculations is likely to trap the ordered molecule and provide misleading guidance for lead optimization.

2:15 77 Efficient method for computing the free energies of active site waters: Application to drug discovery
Jinming Zou, Sia Meshkat, Zenon Konteatis, Anthony Klon, Charles H. Reynolds. Ansaris, Blue Bell, Pennsylvania, United States

Grand canonical Monte Carlo and systematic free energy methods have been reported previously that allow us to rapidly compute protein-fragment interaction energies. The same methodologies can be employed to compute free energies of binding for water. We have used this approach to identify critical waters in a number of therapeutically interesting protein active sites. Knowledge of the location and affinities of these waters can be useful for designing ligands with improved potency.

2:45 78 Using explicit solvent implicitly
Dr. Christopher J Fennell, Charles W. Kehoe, Prof. Ken A. Dill. Department of Pharmaceutical Chemistry, University of California, San Francisco, San Francisco, CA, United States; Graduate Group in Bioinformatics, University of California, San Francisco, San Francisco, CA, United States

Solvent plays a critical role in biomolecular simulations. It mediates the transfer of small molecules, it bridges interactions between ligands and binding sites, it stabilizes protein stuctures with external hydrophilic groups and buried hydrophobic cores, among others. When solvent is modeled explicitly in simulations, the microscopic interactions can be handled rigorously, but obtaining converged solvation energetics can be time-consuming. Here we describe a process, called Semi-Explicit Assembly, where we precompute the solvation response in simple systems and apply it in complex systems. We show that it is possible to have a detailed/explicit-like treatment of solvation at a computational cost similar to the fastest of implicit solvents.

3:15   Coffee Break
3:30 79 Role of solvent in protein-ligand binding
Robert Abel PhD, Noeris Salam PhD, Thijs Beuming PhD, Woody Sherman PhD, Ramy Farid PhD. Schrodinger Inc., New York, NY, United States

Calculation of protein-ligand binding affinities continues to be an active area of research. Although many techniques for computing protein-ligand binding affinities have been introduced, ranging from computationally very expensive approaches, such as free energy perturbation (FEP) theory to more approximate techniques, such as empirically derived scoring functions, which, although computationally efficient, lack a clear theoretical basis - their remains pressing need for more robust approaches. The recently introduced WaterMap technology, which calculates the locations and displacement free energies of hydration sites in proteins, was developed to bridge the gap between the accuracy of FEP and the computational efficiency of empirically derived scoring functions. In the present work, we apply WaterMap to a number of pharmaceutically relevant targets, and present a generalized approach for accurate predication of binding affinities that combines solvation terms from WaterMap with other important thermodynamic terms.

4:00 80 Compute the contribution of protein-pocket solvation to ligand-binding affinity by explicit water simulations
Dr. Ming-Hong Hao, Dr Ingo Muegge. Department of Medicinal Chemistry, Boehringer Ingelheim Pharmaceuticals, Inc, Ridgefield, CT, United States

A significant fraction of ligand-binding free energy in proteins arises from the replacement of water molecules by the ligand in the binding site of proteins. Continuum solvation models based on surface areas do not treat the short-range correlations of water molecules well in the highly irregular and heterogeneous protein-binding pocket. We have developed a computational procedure to simulate the density distribution and free energy of water molecules in the ligand-binding pocket of proteins using a molecular dynamics procedure (NAMD) with explicit water model (TIP3P). Our results are comparable with literature works (e.g. WaterMap software from Schrodinger Inc.) and show good agreement with crystallized water molecules observed in the X-ray structures of proteins. In our procedure, the distribution of water molecules in the protein-binding pocket is presented as water density on a 3-dimensional grid which we find to provide an intuitive way for visualizing the hydrophobic or polar characteristics of a binding site. The contribution of solvation to ligand-binding free energy is estimated by the difference of free energy of the pocket of water replaced by the ligand in the protein binding site and in the bulk solvent. This contribution is added to the direct ligand-protein interactions in scoring the binding affinity of ligands. We investigated the effects of residue mutations in protein binding-site on ligand binding affinity, including the Tryptophan mutations (W79F, W92F, W108A and W120A) in the high-affinity Streptavidin-Biotin complex and the drug-resistant mutants of HIV protease in complex with the inhibitor U-89360E. In these systems, X-ray crystallography showed no significant differences in the given protein-ligand complex structures between the wild type and mutant proteins. Intermolecular interactions between protein and ligand alone do not fully account for the changes in ligand-binding affinity. The free energy change of solvation in the binding site between wild type and mutants provides a good explanation for the shift in ligand-binding affinity. We also applied the procedure to study the structure-activity relationship of congeneric series of ligands. Our results suggest that binding-pocket solvation is an important factor in understanding the binding affinity of ligands to proteins.

4:30 81 All-atom explicit-solvent fragment-based drug discovery: SILCS ("Site Identification by Ligand Competitive Saturation") molecular dynamics simulations applied to IL-2
Prof. Olgun Guvench M.D., Ph.D.. Department of Pharmaceutical Sciences, University of New England College of Pharmacy, Portland, ME, United States

Two challenges in computer-aided drug discovery are incorporation of protein flexibility and an accurate description of solvation effects. Fast in silico screening methods typically employ rigid or near-rigid protein conformations and continuum descriptions of solvation, while more physical and accurate explicit-solvent all-atom molecular dynamics or Monte Carlo methods are very computationally demanding. Site Identification by Ligand Competitive Saturation (SILCS) is a recently-developed computationally-efficient fragment-based drug discovery method that employs all-atom explicit-solvent molecular dynamics simulations, essentially soaking the target in a 1 molar bath of hydrophobic fragments to compute 3-D probability maps of hot-spots on the protein surface that preferentially bind hydrophobic fragments or water molecules. Applied to the apo crystal structure of IL-2, SILCS identifies two hydrophobic pockets not present in the apo crystal, but later discovered to exist in complexes with small molecule inhibitors and to bind hydrophobic moieties on these molecules.

SUNDAY AFTERNOON

COMP - Scripting & Programming HPC on the Cheap
BCEC 157 A
Organized by: Rajarshi Guha
Presiding: Rajarshi Guha
1:30   Introductory Remarks
1:35 55 FPGA implementation of cheminformatics and computational chemistry algorithms and its cost/performance comparison with GPGPU, cloud computing and SIMD implementations
Dr. Attila Berces PhD, Prof. Bela Feher PhD, Peter Szanto, Imre Pechan, Laszlo Lajko, Zoltan Runyo, Peter Laczko, Janos Lazanyi. Chemistry Logic Kft, Budapest, Hungary; Dept. of Measurement and Information Systems, Budapest University of Technology and Economics, Budapest, Hungary; evopro Kft, Budapest, Hungary

We have developed binary fingerprint based similarity searching, topologial torsional fingerprint based similarity searching, chemical library to library comparison, sphere exclusion and Jarvis Patrick clustering, peptide mass spectrometry fingerprinting, BLAST prefiltering, short read mapping in color space on Silicon Graphics RC100 FPGA card. In addition, we implemented the Autodock docking software on FPGA. We reached 5 to 500 folds acceleartion compared to CPU in these implementations. In this presentation the audience will learn what characteristics an algorithm should have to make it worthwhile to implement it on FPGA. We shall also compare the cost/performance characteristics to other alternatives such as cloud computing, GPGPU, and single-instruction-multiple-data (SIMD) optimization.

2:05 56 Technologies for desktop HPC: Application developer's perspective
Dr. Volodymyr Kindratenko PhD, Guochun Shi. National Center for Supercomputing Applications, University of Illinois, Urbana, IL, United States

In the last few years we have witnessed the emergence of a new computing paradigm: computational accelerators. Most prominent examples of such accelerators include FPGAs, Cell/B.E., and most recently GPUs. While these technologies bring unprecedented computing capabilities to the desktop users at a fraction of the cost of a traditional HPC system, their use comes with substantial difficulties due to the need for software reengineering. We survey the landscape of application accelerators for desktop systems and discuss the challenges of re-implementing computational chemistry applications on some of these systems using Hartree-Fock method and molecular dynamics codes as examples.

2:35 57 Faster, cheaper, and better science: Molecular modeling on GPUs
John E. Stone. Beckman Institute, University of Illinois at Urbana-Champaign, Urbana, IL, United States

Over the past ten years graphics processing units (GPUs) have evolved from fixed-function single-purpose devices into highly programmable massively parallel co-processors. State-of-the-art GPUs support double-precision floating point arithmetic and achieve performance levels approaching one trillion floating point arithmetic operations per second. Modern GPUs enable software development in dialects of familiar C, C++, and Fortran languages, and GPU acceleration extensions exist for Python, Matlab, and other popular languages and computing tools. The high performance of GPUs has created opportunities for acceleration of many computationally demanding molecular modeling algorithms that contain significant parallelism.

We will describe how GPUs are currently employed to accelerate some of the most computationally demanding tasks involved in molecular dynamics simulation, visualization, and analysis in our NAMD and VMD software, and give an overview of how GPUs are expected to evolve in the next few years.

3:05 58 Folding@home: Petaflops on the cheap today, exaflops soon?
Prof. Vijay Pande. Department of Chemistry, Stanford University, Stanford, CA, United States

Over the last 10 years, Folding@home has emerged as a very powerful resource. Today, it has multi-petaflop performance, making it the most powerful supercluster in the world. I will talk about how Folding@home works, both in terms of infrastructure and algorithms, and how one can easily reproduce these sorts of approaches in your own lab. I will also very briefly touch on recent results from Folding@home to highlight what petascale power can do to dramatically change the nature of what simulations can inform us about systems of interest.

3:35   Intermission
3:50 59 Protein-ligand docking on the Cell/BE processor with eHiTS Lightning
Zsolt Zsoldos PhD, Orr Ravitz PhD. SimBioSys Inc., Toronto, Ontario, Canada

The eHiTS flexible docking has proven to be among the most accurate pose prediction tools (http://www.simbiosys.ca/ehits/ehits_validation.html) providing one of the highest enrichment factors based on comparative evaluation studies (http://www.simbiosys.ca/ehits/ehits_enrichment.html). The accurate results of eHiTS have been achieved at the price of longer CPU times in the past, but that has changed with the recent port of the algorithm to the Cell/BE processor (http://www.bio-itworld.com/issues/2008/july-august/simbiosys.html). The revolutionary hardware that powers RoadRunner (the world's current fastest supercomputer) and also available in the low cost SONY PS3 game console, gives eHiTS 30-50 fold speedup compared to a single core Intel/AMD processor. The advantages of the Cell/BE platform over other acceleration techniques (FPGA,GPGPU) will be described, along with the challenges faced during the porting effort. A new proximity data structure is introduced that is optimized for SIMD architectures. It allows efficient evaluation of short range pairwise interactions with optimum cache locality.

4:20 60 Fragment-based druggable hot spot identification in proteins and protein-protein interactions using HPC
Dr. Gwo Yu Chuang, Dr. Ryan Brenke, David R Hall, Dr. Dmitri Beglov, Dr. Dima Kozakov, Dr. Sandor Vajda. Department of Biomedical Engineering, Boston University, Boston, MA, United States

Here we present a highly parallel FFT-based method FTMAP for performing computational fragment mapping. Mapping methods place molecular probes on the surface of proteins in order to identify the most favorable binding positions. Since regions of the protein surface that are major contributors to the binding free energy in drug-protein interactions also bind a variety of small organic molecules, mapping can identify such “hot-spots” and the number of probe molecules bound is a good predictor of druggability. The highly parallel nature of our FFT-based approach allows it to be fully scalable, running efficiently on everything from desktop machines with CUDA enabled graphics adapters to an IBM Blue Gene. The method has been applied to both canonical and protein-protein interaction drug targets, successfully predicting binding hot-spots and target druggability. Our public web server is gaining popularity among academic users and generating significant interest from industry.

4:50 61 GPUs: What is all the fuss about?
Brian Cole, Bob Tolbert, Anthony Nicholls. OpenEye Scientific Software, Santa Fe, NM, United States

High performance computing hardware is undergoing a revolution. The best way to achieve increasing performance is through highly parallelized architectures like the graphics processing unit. However, the GPU requires a new assessment of algorithm design based on different memory versus time tradeoffs. Good performance is no longer gained by simply reducing the number of operations, but by organizing the interaction of those operations with a complex hierarchy of memory with varying latencies. Understanding the changing programming paradigm is critical both to selecting which algorithms will benefit from the GPU and how to achieve optimal performance. We will discuss design principles used when porting ROCS to the GPU. We will compare performance of a GPU implementation of ROCS to the highly-tuned production CPU implementation. We will show that higher performance can be achieved on the GPU at a significantly reduced cost compared to CPU clusters.

MONDAY MORNING

COMP - Tautomers and Biology Predictions of Tautomer Ratios
BCEC 157 C
Organized by: Terry Stouch, Yvonne Martin
Presiding: Yvonne Martin
8:30 101 Predicting tautomer preference: Simple rules and unforeseen complexities
Peter W. Kenny PhD, Peter J Taylor. AstraZeneca (retired), Cheadle, United Kingdom

Tautomer ratio depends on phase, so for coherent analysis this must be chosen first. We settle for water as the biological medium, and show inter alia that the gas phase is still more removed from water than even the least polar of organic solvents. We also point out that, while minor tautomers may bind to receptors, this must entail an energetic penalty.

The 'basicity method' is the main source of quantitative data in water but suffers from systematic errors through its inevitable reliance on model compounds. Elimination of these using correction factors not only improves accuracy but has demonstrated structural regularities that have gone unsuspected till now. Their extrapolation leads to plausible predictions amenable to experiment. The effects of benzofusion, and of intramolecular lone pair and dipolar repulsion, exemplify these regularities and will be discussed.

Central to our approach is the realisation that tautomerism takes two forms, 'C-type' and 'N-type,' which depend on different electronic factors. The apparent inconsistencies that result may have helped to inhibit the comprehensive approach to tautomer ratio that is needed, and hopefully their rationalisation will help in its renewal.

9:00 102 Methods for robust and efficient tautomer enumeration, tautomer searching and tautomer duplicate filtering
József Szegezdi, Zsolt Mohácsi, Tamás Csizmazia, Szilárd Dóránt, Ákos Papp, György Pirok, Szabolcs Csepregi, Ferenc Csizmadia. ChemAxon Ltd., Budapest, Hungary

Tautomerism is an important and difficult problem in cheminformatics, and has gained much attention recently. [1] The presentation will focus on ChemAxon's approaches and algorithms for handling tautomerism.

There are four main topics to cover:

1. The tautomerization calculator plugin [2] is the basis of most methods. It can identify tautomerizable regions, enumerate all or dominant tautomers and predict the distribution of dominant tautomers. Furthermore, it can provide generic and canonical tautomers that are used by the methods discussed. It first identifies possible proton donors and acceptors and finds the tautomerization paths between them. Depending on the desired operation, it then combines the paths into regions (generic tautomer), combinatorially enumerates all possible tautomeric forms (all tautomers), filters and ranks enumerated structures based on pKa and other criteria (dominant tautomers) or canonicalizes using empirical rules (canonical tautomer).

The tautomerization plugin is also used to improve results of other calculations, such as macro pKa and logP.

2. Tautomer duplicate search uses generic tautomers combined with a hash key. This method also allows fast filtering of tautomers in chemical database tables. It will be shown how this method is able to handle tautomeric migration of H isotopes and interactions with stereochemistry.

3. Tautomer substructure search enumerates tautomers of the query, and searches each of them separately. In case of query H constraints (explicit H), the constraint is enforced on the tautomeric region to retrieve only true tautomers.

4. Standardizer is a tool for performing custom and built-in transformations on molecules. It is integrated with the JChem chemical database system, so that database and query structures are automatically transformed by the specified transformations [3]. It will be shown how the canonical tautomer and custom transformations can be used to handle tautomerism. Custom transformations also allow handling of ring-chain tautomerism.

References:

[1] Martin, Y.C.: Let's not forget tautomers J Comput Aided Mol Des (2009) 23:693-704, DOI 10.1007/s10822-009-9303-2

[2] Szegezdi, J.; Csizmadia, F: Tautomer generation. pKa based dominance conditions for generating dominant tautomers.
American Chemical Society meeting, Aug 19-23rd, 2007
http://www.chemaxon.com/conf/Tautomer_generation_A4.pdf

[3] Pirok, G. et al: Standardizer - Molecular Cosmetics for Chemoinformatics.
Drug Discovery Technology, August 7-10th, 2006
http://www.chemaxon.com/conf/standardizer.pdf

9:30 103 Tautomerization approach for drug-like molecules
Dr. John C. Shelley PhD, Arron P. Sullivan, David Calkins, Dr. Jeremy R. Greenwood PhD. Schrodinger, Inc., Portland, Oregon, United States; Schrodinger, Inc., New York, New York, United States

We outline a pragmatic approach for generating the important protonation states, including tautomers, for drug-like molecules in the context of ligand and structure based virtual screening. The emphasis is on generating those states that have significant populations (which we define to be 0.01 mole fraction or more) in solution. These states also encompass the vast majority of those intuited from the examination of more than 2,500 protein-ligand complexes. The overall technology combines the use of many pre-parameterized tautomeric equilibria with Hammett and Taft calculation estimates of pKa values, which in turn can also be used to generate variations in both protonation states and tautomeric states. The overall approach permits the calculation of the mole fractions for the states generated along with their relative free energies. These free energy estimates have been shown to improve the performance of subsequent studies such as docking with Glide.

10:00   Intermission
10:15 104 Acid/base ionization vs. prototropic tautomerism
Dr. Robert Fraczkiewicz PhD, Dr. Marvin Waldman PhD, Dr. Robert D. Clark PhD, Walter S. Woltosz MS, MAS, Dr. Michael B. Bolger PhD. Life Sciences, Simulations Plus, Inc., Lancaster, CA, United States

The most serious difficulty in computational predictive modeling of tautomerism is the lack of a sufficiently comprehensive database of tautomeric constants. [1] Published data on aqueous protonic ionization is, on the other hand, quite abundant to build successful QSPR models. Moreover, prototropic tautomerism is intimately tied to ionization in more than one way. We present compelling examples of how these ties can be explored to make both qualitative and quantitative predictions regarding tautomers using a truly predictive model of ionization constants. We show a very surprising case where the model refuted the widely accepted tautomeric form of one of the most successful drugs on the market today and how all of these predictions were confirmed beyond any doubt, both experimentally and theoretically. We demonstrate how the complex tautomerism of another very well known drug could be explained and quantified from its predicted ionization patterns. A general theoretical treatment of tautomer and ionization equilibria will be presented as well.

1. Martin, Y. C. J. Comput. Aided Mol. Des. 2009, 23, 693-704.

10:45 105 Combinatorial-computational-chemoinformatics approachto finding and analyzing low-energy tautomers
Dr. Maciej Haranczyk, Prof. Maciej Gutowski. Computational Research Division, Larence Berkeley National Laboratory, Berkeley, CA, United States; Chemistry-School of Engineering and Physical Sciencs, Heriot-Watt University, Edinburgh, United Kingdom

Enumeration of low-energy tautomers of neutral molecules in the gas-phase or typical solvents can be performed by applying available organic chemistry knowledge.

However, in esoteric cases such as charged molecules in uncommon, non-aqueous solvents there is simply not enough available knowledge to make reliable predictions of low energy tautomers. We have been developing an approach to address the latter problem and we successfully applied it to discover the most stable anionic tautomers of nucleic acid bases that might be involved in the process of DNA damage by low-energy electrons. The approach involves three steps: (i) combinatorial generation of a library of tautomers, (ii) energy-based screening of the library using electronic structure methods, and (iii) analysis of the information generated in step (ii). In steps i-iii we employ combinatorial, computational and chemoinformatics techniques, respectively. This presentation summarizes our developments and most interesting methodological aspects of our approach.

11:15 106 Comparison of pattern-based and algorithm-based approaches to tautomer informatics
Ben Ellingson, Robert Tolbert, A. Geoffrey Skillman. OpenEye Scientific Software, Inc, Santa Fe, NM, United States

Tautomers are an important consideration for cheminformatics and molecular modeling. In cheminformatics, a unique tautomer is stored as the singular registration key where it is vital that the unique key can be generated from any tautomer as well as that all tautomers can be generated from the unique key. The stored tautomer is often chosen for aesthetics or computational ease, but chemical implications such as the loss or gain of aromaticity or stereochemistry through tautomerization must also be addressed. Molecular modelers are often concerned with small ensembles of low energy tautomers. Unfortunately, determining the low energy tautomers is a complex task, for which sub-kcal/mol accuracy remains computationally intensive [1]. Thus, tautomer prediction for large-scale modeling or cheminformatics remains the domain of approximate. We will discuss two such approximate methods, pattern-based tautomer recognition and atom-type tautomer recognition. The advantages and disadvantages of these approaches will be examined.

1. Geballe, M. T.; Skillman, A. G.; Nicholls, A.; Guthrie, J. P.; Taylor, P. J. The SAMPL2 Blind Prediction Challenge: Introduction and Overview. Journal of Computer-Aided Molecular Design 2010, 24, XX.

MORNING

COMP - The Community Structure-Activity Resource (CSAR) Scoring Challenge
BCEC 157 B
Organized by: Heather Carlson
Presiding: AM: Heather Carlson, PM: James Dunbar
9:00   Introductory Remarks
9:05 120 Community structure-activity resource: Collecting, curating, and generating protein-ligand data to improve docking and scoring
Dr. James B. Dunbar Jr, Prof. Heather A. Carlson. Department of Medicinal Chemistry, University of Michigan, Ann Arbor, Ann Arbor, MI, United States

The Community Structure-Activity Resource (CSAR) is a center at the University of Michigan funded by the National Institute of General Medical Sciences. The function of this center is to collect, curate, and disseminate protein-ligand data sets of crystal structures, biological binding affinities, and thermodynamic data to aid in the refinement of docking and scoring methodologies. These data sets are to come from in-house projects at the University of Michigan, other academic labs, and most importantly from industrial, pharma sources. Part of our remit is to augment the deposited data with synthesis, crystallography, and assays to expand the range of properties, binding affinities, and other relevant characteristics involved in docking and scoring. Here, we present CSAR's capabilities and summarize our current in-house project and potential future targets. We also outline the creation of a dataset (based on the PDB, Binding MOAD, and PDBbind) used in our first community-wide benchmark exercise.

9:35 121 Results of CSAR's 2010 Benchmark Exercise
Dr. James B. Dunbar, Dr. Richard D. Smith, Prof. Heather A. Carlson. Department of Medicinal Chemistry, University of Michigan, Ann Arbor, Ann Arbor, MI, United States

The goal of CSAR's Benchmark Exercises is not to declare winners and losers! Instead, we combine the results of all participants to provide a wider assessment of the field. Here, we present an analysis of which protein-ligand complexes score poorly across the majority of submissions (“globally bad” complexes) and compare their properties to the set of complexes that score well across the majority of methods (“globally good”). It may be tempting to draw conclusions by simply examining the characteristics of the globally bad set, but those characteristics must be rarely observed in the globally good set to gain true insight. Lastly, each participant was asked to submit a standard method and an alternative approach. Several groups showed that the correlation to experiment was the same for vdw/fit-based scores as for full scoring functions that included electrostatics and hydrogen bonding. To help the field overcome this limitation, CSAR will focus on creating datasets that provide a range of hydrogen-bonding characteristics. The overarching goal of our benchmark exercises is to provide insight into what data is most needed to move our field ahead.

10:05 122 Scoring performance of eHiTS on the CSAR dataset
Zsolt Zsoldos PhD, Orr Ravitz PhD. SimBioSys Inc., Toronto, Canada

Numerous studies have pointed out at the inability of scoring functions to perform uniformly well accross all biological systems of interest. Some studies suggest guidelines for choosing the best method for a specific problem, others advocate consensus techniques.

An alternative solution is to tailor the scoring function for the system of interest. eHiTS uses a novel scoring method consisting of statistical knowledge focused on interacting surface points and physical terms combined with an adaptive parameter scheme. During the automated tuning of eHiTS-score, receptor targets are clustered according to the chemical and shape similarity of the active site, and weight sets are optimized for each family.

The performance of eHiTS on the CSAR dataset was evaluated using the default parameters (pre-tuned on other data). In addition, the automatic tuning utility was run on one subset of the CSAR data and tested on the other. Results will be presented from both studies.

10:35   Intermission
10:50 123 Hydrophobic complementarity: A dominant term in affinity and binding mode prediction
Dr. Leslie A. Kuhn, Matthew E. Tonero. Biochemistry & Molecular Biology, Michigan State University, East Lansing, MI, United States

Empirical scoring functions designed for high-throughput docking, containing linear combinations of terms measuring protein-ligand interactions, were tested for affinity prediction. Scoring functions that best predicted affinity were dominated by hydrophobic or shape complementarity terms. Similarly, a scoring function containing only polar terms compensated for the absence of a hydrophobic term by heavily weighting the polar term that correlated most with hydrophobic complementarity. These results are consistent with Eisenberg & McLachlan's observation that the solvation component of the change in Gibbs free energy upon binding is proportional to the surface area and degree of hydrophobicity of atoms buried in the interface. Scoring functions that perform best at affinity prediction are not necessarily optimal for binding mode prediction, though hydrophobic burial is important in both. In other words, tuning scoring functions only to predict the affinity of good ligands in the correct binding mode can limit their applicability, suggesting a broader approach.

11:20 124 Docking and scoring for 2010 CSAR benchmark using an improved iterative knowledge-based scoring function with MDock
Sheng-You Huang, Xiaoqin Zou. Department of Physics, Department of Biochemistry, Dalton Cardiovascular Research Center, Informatics Institute, University of Missouri-Columbia, Columbia, MO, United States

Based on a physics-based iterative method (Huang & Zou, J. Comput. Chem., 2006, 27, 1865-75; 1876-82), we have extracted a set of distance-dependent all-atom potentials for protein-ligand interactions (ITScore2.0) using a large training set of 1300 protein-ligand complexes. The iterative method circumvents the long-standing reference state problem in traditional knowledge-based scoring functions. ITScore2.0 has been tested with the 2010 CSAR dataset of 345 diverse protein-ligand complexes, and achieved a correlation coefficient of 0.73 between the calculated binding scores and experimental affinity data, compared to 0.58 for the van der Waals (VDW) scoring function and 0.32 for the force field (FF) scoring function consisting of VDW and electrostatic terms. For rigid-ligand docking, ITScore2.0 achieved a success rate of 86.7% in identifying native binding modes, compared to 80.0% and 64.1% for FF and VDW. For flexible-ligand docking, ITScore2.0 yielded a success rate of 79.7%, compared to 71.0% and 52.8% for FF and VDW. The moderate performance of VDW suggests that VDW alone may serve as a benchmark for evaluation of scoring functions. What we have learned through participating in CSAR scoring will be shared.

11:50   Lunch
1:30 145 Lead Finder in the CSAR scoring challenge
Victor Stroylov MD, Dr Ghermes Chilov, Dr Oleg Stroganov, Fedor Novikov, Val Kulkov MD, MBA. "Molecular Technologies", Ltd, Moscow, Russian Federation; BioMolTech, Corp., Toronto, Ontario, Canada

Lead Finder is a specialized software package for ligand docking, binding energy evaluation and virtual screening. The standard approach in estimation of binding affinities of protein-ligand complexes of the CSAR test set was the use of Lead Finder v.1.1.14 scoring mode that estimates free energy of protein-ligand binding for the fixed ligand coordinates for each protein-ligand complex. No pre-optimization of either protein or ligand structures were performed.

The improvements in the scoring protocol included corrections of protein's and ligand's protonation states, positions of functional hydrogen atoms (for proteins only), and local geometry of nitrogen atoms (for ligands only). No other improvements of Lead Finder's the standard scoring function have been performed.

The RMSD of estimated vs experimentally obtained protein-ligand binding energies was found to be equal to 2.07 kcal/mol and 1.98 kcal/mol for the standard and improved protocols correspondingly.

2:00 146 Benchmark of solvated interaction energy (SIE) scoring function on the CSAR-2010 dataset
Traian Sulea, Qizhi Cui, Herve Hogues, Christopher R Corbeil, Enrico O Purisima. Biotechnology Research Institute, National Research Council Canada, Montreal, QC, Canada

Solvated interaction energy (SIE) is a first-principle function for predicting absolute binding affinities from force-field non-bonded terms, continuum solvation, and scaling for configurational entropy. Standard SIE parametrization applied to the CSAR dataset with binding interfaces refined by constrained minimization predicted absolute affinities with 2.5 kcal/mol mean-unsigned-error, but with correlation outperformed by buried surface or van der Waals interaction alone. Re-training SIE on CSAR subsets led to increased solute dielectric and reduced electrostatic interactions, stressing the weak signal carried by calculated electrostatics in this heterogeneous dataset. Overestimated complexes implicate highly negatively-charged ligands interacting via metals. Underestimated outliers reveal alternate protonation states that significantly improve SIE predictions. In an upgraded version of the CSAR dataset with reassigned protonation states, 10% of ligands and 20% of proteins are affected. Among other investigated aspects are the sensitivity to polar hydrogens orientation, incorporation of MD-generated ensembles, different solvent models and entropy estimates, and ligand strain.

2:30 147 Protonation states and scoring receptor-ligand poses: It's always the details
Emilio Xavier Esposito PhD. exeResearch LLC, East Lansing, Michigan, United States

The protonation state of the receptor - ligand complex has a large influence over the correct approximation of the binding interactions. Using the CSAR dataset, various methods of assigning the complex's protonation state are used to explore the abilities of several scoring functions with respect to protonation state. In conjunction with the complex's protonation state, the 'standard' protocols employed to prepare a receptor for a docking simulation, along with the post-dock refinement of poses, are explored.

3:00   Intermission
3:15 148 Role of active-site solvent in protein-ligand binding affinity calculations
Dr. Ye Che, Dr. Veerabahu Shanmugasundaram. Groton Structural Biology, Antibacterials Chemistry/Discovery Technologies, Pfizer PharmaTherapeutics Research & Development, Groton, CT, United States

Accurate methods for computing binding affinities of a small molecule to a protein are needed to speed the discovery and optimization of new medicines. An assessment of six scoring functions commonly applied at Pfizer using the CSAR (Community Structure-Activity Resource) set of protein-ligand complexes will be presented. A current weakness amongst these various scoring functions is the treatment of active-site water molecules. Here, we quantitatively estimate the thermodynamic properties of active-site water molecules and capture the effects of solvent displacement from the protein active site. Water inclusion shows promise in improving current scoring functions and we propose that this could be used more extensively in virtual screening and lead optimization applications.

3:45 149 Flexible docking using a stochastic rotamer library of ligands
Dr. Feng Ding, Dr. Shuangye Yin, Prof. Nikolay V. Dokholyan. Biochemistry and Biophysics, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States

Uncovering structures of molecular complexes via computational docking is at the heart of many structural modeling efforts and virtual drug screening. Modeling both receptor and ligand flexibility is important in order to capture receptor conformation changes induced by ligand binding, but is a major challenge in computational drug discovery. Many flexible docking approaches model the ligand and receptor flexibility either separately or in a loosely-coupled manner, which captures the conformational changes inefficiently. Here, we propose a truly flexible docking approach, MedusaDock, which models both ligand and receptor flexibility simultaneously using sets of discrete rotamers. We developed an algorithm which allows for the building of the ligand rotamer library “on the fly” during docking simulations. MedusaDock benchmarks demonstrate a rapid sampling efficiency and high prediction accuracy in both self-docking (to the co-crystallized state) and cross-docking (to a state co-crystallized with a different ligand), the latter of which mimics the virtual screening procedure in computational drug discovery. We also perform a virtual-screening test for a flexible protein target, cyclin-dependent kinase 2. We find a significant improvement in virtual screening enrichment when compared to rigid-receptor methods. The high predictive power of MedusaDock comes from several innovations, including the generation of a stochastic rotamer library of ligands, the efficient docking protocol, and the novel ligand pose-ranking method. We expect a broad adaption of these methodologies and the application of MedusaDock in ligand-receptor interaction predictions and drug discovery.

4:15 150 Cheminformatics meets molecular mechanics: A combined application of knowledge based pose scoring and physical force field-based hit scoring functions improves the accuracy of virtual screening
Jui-Hua Hsieh, Shuangye Yin, Xiang S. Wang, Shubin Liu, Nikolay V. Dokholyan, Alexander Tropsha. University of North Carolina at Chapel Hill, United States

Many scoring functions fail to discriminate between true binders and non-binders (binding decoys), leading to a large number of false positive hits in virtual screening (VS) studies. We have developed a novel binary QSAR-like approach that discriminates geometrical pose decoys from native-like poses for each ligand. We have applied it for filtering (presumed) decoy poses from a library of docked ligand conformations followed by scoring the remaining poses with the MedusaScore physical force field-based scoring. We have demonstrated that this pre-filtering affords a significant improvement of hit rates in virtual screening studies for 5 of the 6 benchmark sets from the Database of Useful Decoys (DUD). Moreover, the top 10 hits in these 5 sets were found to include chemically diverse ligands while yielding high true positive rates (60-100%). We will discuss the methodology as well as the results of applying this approach to CSAR datasets.

MONDAY AFTERNOON

COMP - Using Waters Explicitly in Drug Discovery Characterization and Applications
BCEC 154
Organized by: Veerabahu Shanmugasundaram, Woody Sherman
Presiding: Veerabahu Shanmugasundaram
1:30   Introductory Remarks
1:35 151 Application of free energy methods to water molecules in protein binding sites
Prof. Jonathan W. Essex D.Phil., Dr Caterina Barillari PhD, Mr Michael Bodnarchuk, Dr Russell Viner PhD. School of Chemistry, University of Southampton, Southampton, Hampshire, United Kingdom; Jealott’s Hill International Research Centre, Syngenta, Bracknell, United Kingdom

Water molecules play a crucial role in mediating the interaction between a ligand and a macromolecular receptor. An understanding of the nature and role of each water molecule in the active site of a protein could efficiency of rational drug design approaches. In this presentation, a range of different simulation methods, including double decoupling with replica exchange thermodynamic integration, Grand-Canonical Monte Carlo, and JAWS, are used to calculate the absolute binding free energies of a number of water molecules in protein-ligand complexes. The relative merits of each of these methods are discussed. In addition, the development of a number of descriptor-based QSAR models for calculating water binding free energies is described, with a view to reducing the need for expensive free energy simulations.

2:15 152 Which waters are important and how do weidentify them?
Dr Simon Bowden, Dr Jason C Cole, Dr Oliver Korb, Dr Tjelvar Olsson, Dr John Liebescheutz, Dr Colin Groom. Cambridge Crystallographic Data Centre, Cambridge, United Kingdom

The important role waters play in ligand binding both in terms of thermodynamics and selectivity is well known but identifying which waters are important for the success of a docking experiment is still difficult. Given that consideration of waters involved in primary and secondary mediated protein-ligand contacts has been shown to improve success rates in both native docking and virtual screening, experimenters need tools to help them decide which waters are important and which are not even real.

In this talk we will describe tools which may be of use to identify important waters and to highlight dubious waters. Conserved water structures can also be identified which may have an important influence on ligand binding. The effect of this information when applied to molecular docking will be demonstrated.

2:45 153 Free energies and entropies of water molecules at protein-ligand interfaces
Prof. Steve W Rick PhD, Mr. Hongtao Yu. Chemistry, University of New Orleans, New Orleans, LA, United States

Water molecules are commonly found at he protein-ligand interface. The thermodynamics of these water molecules plays an important role in ligand affinity. In particular, the entropic cost of localizing a water molecule at the binding site can be significant. From the database of crystal structures, it is evident that the local environments of water molecules at the protein-ligand interface can vary considerably. We use molecule dynamics simulations and thermodynamic integration to calculate the free energy, enthalpy, and entropy changes associated with localizing a water molecule at a wide variety of sites at protein-ligand interfaces. Results analyzing how the free energies, enthalpies, and entropies depend on the details of the local environment, including the number of hydrogen bonds and the cavity size, will be presented.

3:15   Coffee Break
3:30 154 Role of water molecules in docking studies of Cytochromes P450
Dr. Chris Oostenbrink. Institute of Molecular Modeling and Simulation, BOKU University, Vienna, Austria; Chemistry and Pharmaceutical Sciences, VU University, Amsterdam, The Netherlands

Active-site water molecules form an important component in biological systems facilitating promiscuous binding, or an increase in specificity and affinity. Taking water molecules into account in computational approaches to drug design or site-of-metabolism prediction is far from straightforward. The effect of including water molecules in molecular docking simulations of metabolic Cytochrome P450 enzymes is investigated, focusing on pose prediction, virtual screening and free energy estimates. The structure and dynamics of water molecules that are present in the active site simultaneously with selected ligands are described. The transferability of hydration sites between different ligands is investigated. The role of water molecules appears to be very dependent on the protein conformation and the substrate, further enhancing the versatility of these metabolic enzymes.

4:00 155 Modeling explicit waters in docking and scoring
Dr. Niu Huang. National Institute of Biological Sciences, Beijing, Beijing, China

Water molecules play an important role in protein-ligand recognition. However, incorporating explicit waters during docking is challenging in both the sampling and scoring aspects. We explored a method to switch ordered water molecules “on” (retained) and “off” (displaced) during docking screens. This method assumes additivity and scales linearly with the number of waters sampled despite the exponential growth in configurations. We tested this approach for ligand enrichment in screens of a large compound database against 24 DUD targets, exploring up to 8 waters in 256 configurations. Compared to calculations where the water positions were not sampled, enrichment factors increase substantially for 12 of the targets and are largely unaffected for most others. However, in our previous study, the positions of the water molecules were obtained from the x-ray structures, and all waters were treated as equally displaceable without the consideration of the differential energy of water binding. Our recent work in improving the treatment of waters during docking and scoring will be presented.

4:30 156 Desolvation/resolvation: A revolving door that controls the rates of association/dissociation of protein-ligand complexes? Analysis of PCSK9-EGF-A binding kinetics using WaterMap
Dr. Robert A. Pearlstein Ph.D., Dr. Qi-Ying Hu Ph.D., Dr. Jing Zhou Ph.D., Dr. David Yowe Ph.D., Dr. Julian Levell Ph.D., Bethany Dale, Virendar Kaushik, Dr. Doug Daniels Ph.D., Susan Hanrahan, Dr. Woody Sherman Ph.D., Dr. Robert Abel Ph.D.. Novartis Institutes for BioMedical Research, Cambridge, MA, United States; Schrodinger, Inc., New York, NY, United States

We hypothesize that desolvation and resolvation processes can constitute rate-determining steps for protein-ligand association and dissociation, respectively. We tested this hypothesis using proprotein convertase subtilisin-kexin type 9 (PCSK9) bound to the epidermal growth factor-like repeat A (EGF-A) of low density lipoprotein cholesterol receptor (LDL-R). We analyzed and compared predicted desolvation properties of wild-type vs. gain-of-function mutant Asp374Tyr PCSK9 using WaterMap, a new method for calculating preferred locations and thermodynamic properties of water solvating proteins (“hydration sites”). We propose that fast kon and entropically driven thermodynamics observed for PCSK9-EGF-A binding is due to functional replacement of water occupying stable PCSK9 hydration sites (exchange of water for polar EGF-A groups). We further propose that relatively fast koff observed for EGF-A unbinding results from limited displacement of unstable water. Slower koff observed for EGF-A and LDL-R unbinding from Asp374Tyr PCSK9 may be due to destabilizing effects of this mutation on PCSK9 hydration sites.

5:00 157 Biophysics-based library design: Discovery of “non-acid” inhibitors of S1 DHFR
Veerabahu Shanmugasundaram, Kris Borzilleri, Jeanne Chang, Boris Chrunyk, Mark E Flanagan, Seungil Han, Melissa Harris, Brian Lacey, Richard Miller, Parag Sahasrabudhe, Ron Sarver, Holly Soutter, Jane Withka. Groton Structural Biology, Antibacterials Chemistry/Discovery Technologies, Pfizer PharmaTherapeutics Research & Development, Groton, CT, United States; AntiBacterials Chemistry, Pfizer PharmaTherapeutics Research & Development, Groton, CT, United States; AntiBacterials Research Unit, Pfizer PharmaTherapeutics Research & Development, Groton, CT, United States

Methicillin-resistant Staphylococcus aureus (MRSA), the causative agent of many serious nosocomial and community acquired infections, and other gram-positive organisms can show resistance to trimethoprim (TMP) through mutation of the chromosomal gene or acquisition of an alternative DHFR termed "S1 DHFR" To develop new therapies for health threats such as MRSA, it is important to understand the molecular basis of TMP resistance and use that knowledge to design and develop novel inhibitors that are effective against S1 DHFR. This presentation will highlight and illustrate an effort using a multi-pronged biophysics based strategy that utilizes NMR, thermodynamic, kinetic, structural, computational and medicinal chemistry information in developing an understanding of the mechanism of resistance in S1 DHFR as well as using this prospectively in drug discovery. Specifically this presentation will illustrate computational studies using WaterMap (WM) that developed an understanding of a key element of the mechanism of resistance that was supported by a variety of biophysical experiments and use of these WM calculations in a prospective fashion in library design.

TUESDAY MORNING

COMP - The Community Structure-Activity Resource (CSAR) Scoring Challenge
BCEC 157 B
Organized by: Heather Carlson
8:30 174 Predicting relative binding affinities in the CSAR ScoringChallenge
Prof. Matthew P Jacobson, Dr. Chakrapani Kalyanaraman. Department of Pharmaceutical Chemistry, UCSF, San Francisco, CA, United States

We have been interested in evaluating whether all-atom force fields combined with implicit solvent models can be also used as a docking scoring function. Our prior experience has suggested that such energy functions can be used for, at best, predicting relative binding affinities to a particular binding site, with the best results being achieved for chemically related compounds, such as congeneric series generated in lead optimization. Thus, although predicting absolute binding affinities is a noble challenge, we have not attempted to do so in the CSAR exercise. Instead, with the assistance of the organizers, we focused on series of compounds bound to the same target. The results using the protein-ligand structures as provided showed essentially no ability to rank order compounds by binding affinity. However, complete energy minimization, and in some cases correcting protonation states, significantly improved the results, to the point where there was some ability to distinguish more potent from less potent compounds, as we have also shown in other work on congeneric series. I will also discuss our attempts to characterize and correct some of the many limitations of this simple scoring scheme.

9:00 175 Surflex:Docking and scoring on CSAR
Prof. Ajay N Jain PhD. Bioengineering and Therapeutic Sciences, UCSF, San Francisco, CA, United States

One of the most challenging aspects of structure-based drug design is binding affinity prediction, since it embeds both the pose determination problem as well as requiring accuracy in estimation of energetic contributions where differences on the order of 1 kcal are large enough to matter. Even in the artificial case where a bound ligand/target structure is known, this remains a challenging problem. We present results for the Surflex family of methods for making predictions on the CSAR 2010 benchmark data set. Results will include straight docking-based pose prediction and scoring, tuned scoring approaches through scoring function optimization and protein structure optimization, and ligand-based approaches.

9:30 176 What we can learn from very large panel docking screens
Kong T Nguyen, John J Irwin, Brian K Shoichet, Michael M Mysinger. Department of Pharmaceutical Chemistry, University of California San Francisco, San Francisco, CA, United States

Whereas molecular docking is the most practical way to leverage structure for ligand discovery, the method retains important weaknesses. Among the more confounding problems is that docking can work well one target yet fail completely on the next, yet predicting in advance which will succeed or fail is challenging. To investigate the strengths and weaknesses of docking we have assembled a very large panel of experimental information with which to test it. We have used our automated docking program, DOCK Blaster1. , to study the performance of DOCK 3.5.54 against many protein targets for which experimental control information is available2. We have focused on two of the seven stated goals of the 2010 CSAR Workshop: to provide a baseline assessment of current scoring functions and to document which targets are most difficult. This approach has enabled us to comprehensively test the effect of changes in sampling, scoring and library composition.

References

1. Irwin, J.J. et al. Automated docking screens: a feasibility study. J Med Chem 52, 5712-20 (2009).

2. Overington, J. ChEMBL. An interview with John Overington, team leader, chemogenomics at the European Bioinformatics Institute Outstation of the European Molecular Biology Laboratory (EMBL-EBI). Interview by Wendy A. Warr. J Comput Aided Mol Des 23, 195-8 (2009).

10:00   Community Discussion: Evaluation Criteria
10:30   Intermission
10:45 177 Docking and scoring of fragments
Dr Marcel L Verdonk PhD. Astex Therapeutics Ltd, Cambridge, United Kingdom

Through the application of fragment-based drug discovery, Astex have produced >1,400 in-house X-ray crystal structures of fragments and >2,500 structures of lead-like compounds against a range of drug targets. From this wealth of structural data, we have constructed two test sets, each containing ~100 complexes, representing 10 drug targets. In the first test set the ligands are fragments, whereas in the second test set the ligands are lead-like compounds. By applying docking and virtual screening on these sets, we will discuss whether fragments are harder to dock and score than larger compounds, and present our latest experiences on docking and scoring fragments. In addition, we will show how structural data on fragments obtained early on in drug discovery projects can be used to improve docking and scoring during the hit-to-lead phases. Finally, we will show examples of the application of docking and scoring of fragments on actual drug discovery programs.

11:15   Community Discussion: Dataset Criteria

TUESDAY MORNING

COMP - Tautomers and Biology Tautomers and Macromolecule-ligand Complexes
BCEC 157 C
Organized by: Terry Stouch, Yvonne Martin
Presiding: Yvonne Martin
8:30 170 Computational evaluation of tautomers and zwitterions of D-amino acid oxidase (DAAO) inhibitors
Scot Mente. Neuroscience Chemistry, Pfizer Global Research and Development, Groton, CT, United States

Quantum mechanical calculations and molecular docking were used in to design novel inhibitors of D-amino acid oxidase (DAAO). Using available x-ray structural information and simple tautomer enumeration tools, reasonable docked poses of a set of small ligands have been obtained. Use of these tools have helped lead to the optimization of the novel non-acidic 3-hydroxyquinolin-2(1H)-one Series (I), as well as the identification of structurally similar 3-hydroxyquinoline (II) and benzotriazole (III). Despite their small sizes, all three of these molecular scaffolds are capable of adopting multiple tautomer or zwitterionic states. The ability to accurately predict these states with quantum mechanical methods will be discussed.

9:00 171 Defining states of ionization and tautomerization of thiamin diphosphate at individual reaction intermediates on enzymes: Enzymes that use a rare tautomeric form
Prof. Frank Jordan PhD, Dr. Natalia S. Nemeria PhD, Mr. Anand Balakrishnan, Mr. Siakumar Paramasivam, Prof. Tatyana Polenova PhD. Chemistry, Rutgers University, Newark, NJ, United States; Chemistry and Biochemistry, University of Delaware, Newark, DE, United States

The author and coworkers demonstrated on several thiamin diphosphate (ThDP) enzymes that the 1',4'-iminopyrimidine tautomer of ThDP participates at several reaction steps. Hence, ThDP has dual function: an electrophilic covalent catalyst - a function long accepted- and an acid-base catalyst facilitating the ionization of the weak carbon acid to generate the C2 ylide.

It is proposed that ThDP exists in these forms on enzymes: the N1'-protonated 4-aminopyrimidinium (APH+) in protolytic equilibrium with its three conjugate bases, the canonical 4-aminopyrimidine (AP), its 1',4'-iminopyrimidine (IP) tautomeric form, and the C2 carbanion or ylide (Yl). The first three forms have been observed on multiple enzymes in the absence of substrate. In the presence of substrate and analogs, the IP form has been seen on several enzymes along with the APH+ state. Circular dichroism and solid-state NMR methods are being used for the first time to characterize different species. Supported by NIH-GM-050380 and 5P20RR017716.

9:30   Intermission
9:45 172 Do tautomers matter in calculating molecular similarity?
Dr. Steven W Muchmore PhD, Isabella Haight, Dr. Scott Brown. Cheminformatics, Abbott Laboratories, Abbott Park, IL, United States

Compounds that have multiple tautomeric forms, which typically account for about 25% of pharmaceutical company corporate collections, present a challenge in cheminformatic analysis. While widely recognized, their manipulations are often ignored in database registration, substructure searching and similarity searching due to incremental increases in computation time and data management. However, clustering and diversity selection, which are based on similarity calculations, could yield erratic results if they include or exclude molecules that happen to be encoded as different tautomers. We enumerated tautomers for a data set of more than 66,000 compound pairs with associated activity against protein targets used in the assessment of similarity programs (Muchmore et al. J. Chem. Inf. Model. 2008, 48, 941). The similarity value for the highest scoring tautomer pair was compared to the original data to determine if its similarity score increased. These tautomer similarity values were also applied to single representation results to determine if tautomer enumeration would yield a better estimate of the probability that two compounds will be equipotent.

10:15 173 Automated prediction of tautomeric states in protein-ligand complexes
Sascha Urbaczek, Stefan Bietz, Prof. Dr. Mathias Rarey. Center for Bioinformatics, University of Hamburg, Hamburg, Hamburg, Germany

Hydrogen bonding plays a mayor role in the stabilization of protein-ligand complexes. Unfortunately, the positions of hydrogen atoms are not resolved in most structures present in the PDB. This makes it particularly hard to predict adequate tautomeric and protonation states for the atoms and groups involved in the binding. To overcome this difficulty many approaches have been developed to predict the correct protonation of either the ligand or the protein separately using a variety of different methodologies. We present a new method that predicts the tautomeric and protonation states as well as the resulting hydrogen atom positions of both the protein and the ligand simultaneously. The optimization of these states is based on an empirical scoring scheme used also in docking methods. Assuming an optimal hydrogen bonding network, the obtained results indicate that the most stable tautomeric forms in solution do not always correspond to those found in binding modes.

TUESDAY AFTERNOON

COMP - Using Waters Explicitly in Drug Discovery Hybrid Explicit/Implicit Methods
BCEC 154
Organized by: Veerabahu Shanmugasundaram, Woody Sherman
Presiding: Woody Sherman
1:30   Introductory Remarks - Woody Sherman
1:35 217 Molecular dynamics studies of water-protein interactions
Gerhard Hummer, Jayendran C. Rasaiah, Hao Yin, Guogang Feng. Laboratory of Chemical Physics, National Institutes of Health, Bethesda, MD, United States; Department of Chemistry, University of Maine, Orono, ME, United States

We use molecular dynamics simulations to study the interaction of water with proteins. With the help of a semi-grand canonical formalism, we determine the structure, dynamics, and thermodynamics of water in the protein interior and at buried sites. We find that water filling of weakly polar protein cavities from the solvent is governed by a subtle balance between the loss in bulk hydrogen bond interactions, the gain in strong hydrogen-bond interactions between confined water molecules, weakly attractive interactions between water and the cavity, and the entropic gain from filling a void space. The simulation results will be compared to X-ray crystallography and NMR experiments. The effects of interfacial and cavity water on protein function and ligand binding will be discussed.

2:15 218 Addressing limitations with the MM-GB/SA scoring procedure using the WaterMap method and free-energy perturbation calculations
Dr. Cristiano R. W. Guimaraes. CVMD Chemistry, PharmaTherapeutics Research and Development, Pfizer, Inc., Groton, Connecticut, United States

The MM-GB/SA scoring technique has become an important computational approach in lead optimization. Despite showing good accuracy, much work is necessary before the method can be applied to rank multiple chemical series. Here, we investigate the poor estimation of protein desolvation provided by GB/SA and the large dynamic range in the MM-GB/SA scoring compared to that of the experimental data. In the former, replacing the GB/SA protein desolvation by the WaterMap free energy liberation of binding-site waters provides the best results. However, the improvement is modest over results obtained with the MM-GB/SA and WaterMap methods individually, apparently due to the high correlation between the free energy liberation and protein-ligand van der Waals interactions. As for the large dynamic range, comparisons between MM-GB/SA and FEP calculations indicate that it has its origin in the lack of dynamical screening of protein-ligand electrostatic interactions and the incomplete description of enthalpy-entropy compensation effects.

2:45 219 Prediction of potency of protease inhibitors by GBSA simulations with polarizable quantum mechanics-based ligand charges and a hybrid water model
Dr. Debananda Das, Dr. Hiroaki Mitsuya, Dr. Yasuhiro Koh, Yasushi Tojo, Dr. Arun Ghosh. HIV and AIDS Malignancy Branch, National Cancer Institute, Bethesda, MD, United States; Departments of Hematology and Infectious Diseases, Kumamoto University Graduate School of Medical and Pharmaceutical Sciences, Kumamoto, Japan; Departments of Chemistry and Medicinal Chemistry, Purdue University, West Lafayette, Indiana, United States

Reliable and robust prediction of binding affinity for drug molecules continues to be a daunting challenge. We have simulated the binding interactions and free energy of binding of several protease inhibitors (PIs) with wild-type and various mutant proteases by performing GBSA simulations, in which each PI's partial charge was determined by quantum mechanics and the partial charge accounts for the polarization induced by the protease environment. We employed a hybrid solvation model that retains selected explicit water molecules in the protein with surface generalized Born implicit solvent. We examined the correlation of the free energy with antiviral potency of PIs. The free energy showed a strong correlation with experimentally determined anti-HIV-1 potency. The present data suggest that the presence of selected explicit water in protein, and protein polarization-induced quantum charges for the inhibitor, compared to lack of explicit water and a static force field-based charge model, can serve as an improved lead optimization tool, and warrants further exploration.

3:15   Coffee Break
3:30 220 Continuum theory and the analysis of active sites
Dr. Anthony Nicholls PhD, Dr. Mike Word. Department of Research and Development, OpenEye Scientific Software, Inc, Santa Fe, NM, United States

Continuum theory for electrostatics free energies at the molecular level was never supposed to work- water is discrete and the very idea of treating its properties as a mean field was considered inappropriate. Yet Poisson-Boltzmann (PB) theory continues to perform as well as, if not better than, explicit water treatments in the estimation of small molecule solvation or macromolecular biophysics. However, it is still assumed PB will fail to correctly describe the physics of the active sites of proteins. As this remains a focus for predictive drug discovery, is this assumption correct? And if it is, can we improve continuum theory by going beyond the mean field limit, i.e. producing a 'virial' expansion of PB? This talk will cover our attempts to date and the physical insight gained.

4:00 221 Prediction of consistent water networks in uncomplexed protein binding sites based on knowledge-based potentials
Michael Betz, Gerd Neudert, Professor Gerhard Klebe PhD. Institute of Pharmaceutical Chemistry, Philipps-University Marburg, Marburg, Germany

Within the active site of a protein water fulfills a variety of different roles. Solvation of hydrophilic parts stabilizes a distinct protein conformation, whereas desolvation upon ligand binding may lead to a gain of entropy. In an overwhelming number of cases, water molecules mediate interactions between protein and the bound ligand. Therefore, a reliable prediction of water molecules participating in ligand binding is essential for docking and scoring, and is necessary to develop strategies in ligand design. We require some reasonable estimates about the free energy contributions of water to binding.

Useful parameters for such estimations are the total number of displaceable water molecules and the probabilities for their displacement upon ligand binding. These parameters depend on specific interactions with the protein and other water molecules, and thus the positions of individual water molecules.

The high flexibility of water networks makes it difficult to observe distinct water molecules at well defined positions in structure determinations. Thus, experimentally observed positions of water molecules have to be assessed critically, bearing in mind that they represent an average picture of a highly dynamic equilibrium ensemble. Moreover, there are many structures with inconsistent and incomplete water networks.

To address these deficiencies we developed a tool that predicts possible configurations of complete water networks in binding pockets in a consistent way. It is based on the well established knowledge-based potentials implemented into DrugScore, which also allow for a reasonable differentiation between "conserved" and "displaceable" water molecules. The potentials used were derived specifically for water positions as observed in small molecule crystal structures in the CSD.

To account for the flexibility and high intercorrelation we apply a clique-based approach, resulting in water networks maximizing the total DrugScore.

To incorporate as much known information as possible about a given target, we also allow to include constraints defined by experimentally observed water positions.

Our tool provides a useful starting point whenever a possible configuration of water molecules need to be estimated in an uncomplexed protein, and suggests their spatial positions and their classification with respect to some kind of affinity prediction.

In first tests we were able to get classifications and positional predictions which are in good agreement with crystallographically observed water molecules with remarkably small deviations.

4:30 222 Explicit-water modeling of a model protein-ligand binding site predicts the non-classical hydrophobic effect
Demetri T. Moustakas PhD, Phil W Snyder PhD, Woody Sherman PhD, Prof. George M Whitesides. Department of Infection, Computational Sciences, AstraZeneca R&D Boston, Waltham, MA, United States; Department of Chemistry and Chemical Biology, Harvard University, Cambridge, MA, United States; Schrödinger, Inc., New York, NY, United States

This work reports a study of the thermodynamics of hydrophobic interactions between human carbonic anhydrase II and a series of structurally analogous heteroaromatic sulfonamides. Isothermal titration calorimetry (ITC) established that increasing the non-polar surface area of the ligands resulted in a large enthalpy-dominated increase the binding affinity - the so-called non-classical hydrophobic affect. Subsequent X-ray crystallography studies reveal no significant changes in protein-ligand interactions as a function of increasing the ligand non-polar surface area, suggesting that solute-solvent interactions are responsible for the observed thermodynamic effects. Modeling studies using explicit solvent models suggest that the larger ligands alter both the structure and thermodynamic characteristics of water molecules in the binding site, which contributes significantly to the observed non-classical hydrophobic effect.

5:00 223 New coarse-grained model for water: The importance of electrostatic interactions
Zhe Wu, Prof. Qiang Cui, Prof. Arun Yethiraj. Department of Chemistry, UW Madison, Madison, WI, United States

A new coarse-grained (CG) model for water is developed based on the properties of clusters of four water molecules in atomistic simulations. CG units interact via a soft non-electrostatic interaction. Electrostatic interactions are incorporated via three charged sites with the charges and model topology chosen to reproduce the dipole moment and quadrupole moment tensor of 4-water clusters. The parameters in the model are optimized to reproduce experimental data for the compressibility, density, and permittivity of bulk water, and the surface tension and interface potential for the air-water interface. This big multipole water (BMW) model represents a qualitative improvement over existing CG water models, e.g., it reproduces the dipole potential in membrane-water interface when compared to experiment, with modest additional computational cost.

WEDNESDAY AFTERNOON

COMP - Scripting & Programming Cross Pharma Collaboration in High Performance Computing
BCEC 157 C
Organized by: Eugene Fluder, Zheng Yang
Sponsored by: Dell Incorporated, Intel Corporation
1:30 359 Introduction to cross pharma high performance computing forum
John C Morris MBA, Dr Zheng Yang. Massachusetts Research Business Technology, Pfizer, Cambridge, MA, United States; Department of Computational and Structural Chemistry, GlaxoSmithKline Pharmaceuticals, Collegeville, PA, United States

High Performance Computing (HPC) within the pharmaceutical industry is a growing and critical component of research due to the large scale analytical demands driven by modern research methods and advancements in computational chemistry and bioinformatics methods to model biological systems. HPC has become a necessary capability to facilitate the analysis of the terabytes of scientific data being generated from technologies such as Next Generation Sequencing, modeling complex drug-target interaction, and statistical analysis. To support the industrialization of scientific research, integrated and coordinated HPC information technology tools, methods, and capabilities are needed. The Cross Pharma HPC forum is a group of scientists, engineers, and key stakeholders within the pharmaceutical industry working together to promote best practices, coordinate activities, optimize methods, and leverage experience in the non-competitive areas within HPC. In this talk, the history, current status, and future directions of HPC in the pharmaceutical industry will be discussed.

2:00 360 Applications and use of cloud computing in the pharmaceutical industry
Dr. Michael D Miller PhD, David M Powers, Gregory Stiegler, Dr Jeremy Martin M PhD. Research Business Technlogy, Pfizer, Groton, CT, United States; Research and Development IT, Eli Lilly, IIndianapolis, Indiana, United States; Scientific Computing, Bristol-Myers Squibb, Princeton, New Jersey, United States; System Support Department, Information Technology, GlaxoSmithKline R&D Ltd, Harlow,, Essex, United Kingdom

Technological advances across the sciences have enabled basic drug research with an unprecedented amount of data. As a result, the application of computational methods are becoming an increasingly important approach in drug discovery and development. The need for increased computing capacity has reached the point where, today it can become rate limiting. As a result Pharmaceutical companies have begun exploring the use of cloud computing to address these needs. We will present on some of the challenges Pharmaceutical companies have faced in using cloud resources and the different approaches that have been taken to address them.

2:30 361 Current trends of high performance computing in Pharma
Dr. Stephen Litster, Dr. Jeremy Martin. NITAS Scientific Computing, Novartis Institutes of BioMedical Research, Cambridge, MA, United States; Department of System Support, Information Technology, GlaxoSmithKline R&D Ltd, Harlow, Essex, United Kingdom

The world of high performance computing (HPC) has evolved quickly, as exemplified by recent developments in hardware (e.g. Intel Nehalem multi-core CPUs with integrated memory controller), software (e.g. NAMD, a highly scalable molecular dynamics program), computing services (e.g. cloud computing), and storage (TB+ scale file systems). Given these recent developments and much lower cost of entry into HPC, Pharma based Scientific Computing groups are beginning to apply traditional HPC techniques to “non-traditional” (e.g. High Content Screening) and emerging areas of research (e.g. Next Generation Sequencing).

We present here a number of case studies highlighting the current trends of HPC in the pharmaceutical industry and its to impact scientific workflows.

3:00 362 Challenges of HPC and collaboration opportunities in Pharma
Robert Stansfield PhD, MBA, Michael D Miller PhD. R&D Information Solutions, sanof-aventis U.S., Bridgewater, NJ, United States; Research Business Technologies, Pfizer, Groton, CT, United States

High Performance Computing (HPC) in Pharmaceutical R&D is well established in computational chemistry and computational biology for drug discovery, but is increasingly seeing broader application across research and development. In addition, internal capacity is being supplemented by external “cloud computing”. In consequence, the issues around providing HPC services to in-house scientists in an optimal way for the entire company become more visible and critical. From a technical perspective, HPC requires a holistic view across compute, network, and storage capabilities. From an organizational perspective, effective governance - roles, responsibilities, prioritization and decision making across multiple different groups, operations, and support to end-user scientists - makes all the difference. For these reasons at least, HPC deserves a place in strategic planning. These issues will be explored, as well as the opportunities afforded by pre-competitive collaboration in the Cross-Pharma HPC Forum for identifying best practices.

3:30   Intermission
3:45   Open Panel Discussion

THURSDAY

COMP - Targeting Gram-Negative Pathogens
BCEC 157 A
Organized by: Veerabahu Shanmugasundaram, Jeremy Starr
Presiding: AM: Veerabahu Shanmugasundaram, PM: Jeremy Starr
8:00   Introductory Remarks - Veer Shanmugasundaram
8:05 376 Approaches to the treatment of multidrug resistant gram negative infections
Dr. Mark C Noe PhD, Dr. Steven J Brickner PhD, Dr. Thomas Gootz PhD, Michael Huband, Dr. Mark E Flanagan PhD, Dr. John Mueller PhD. Department of Antibacterials Research, Pfizer Global Research and Development, Groton, CT, United States

Each year, over 4.3 million people worldwide contract hospital-based bacterial infections, approximately half of which are caused by Gram negative organisms. The widespread emergence of genes that confer multidrug resistance in these pathogens threatens to undermine the clinical utility of several antibiotic classes, including the fluoroquinolones, cephalosporins, carbapenems and aminoglycosides. Particularly concerning are the extended spectrum beta lactamases, including carbapenemases, which are advancing at an alarming rate and compromise the effectiveness of the most widely used classes to treat Gram negative infections. This talk will review the medical need for new antibacterial agents, some of the challenges associated with discovering new antibiotics, examples of potentially enabling technologies and recent advances in our understanding of privileged targets for antibacterial therapy. An example of one antibacterial drug discovery program will be presented.

8:45 377 Physicochemical property space of antibiotics
Heinz E Moser PhD. Department of Chemistry, Achaogen, South San Francisco, California, United States

While there have been enormous discovery efforts during the past decades to identify novel classes of antibacterials with clinical utility against Gram-negative pathogens, no first-in-class compounds have been successfully developed to use in humans for roughly half a century, and none is currently in clinical evaluation. Predictably, this lack of success has been met by an increasing prevalence of Gram-negative pathogens causing serious infections in hospitals and critical care settings. Recent outbreaks caused by multi-drug resistant (MDR) or pan-resistant organisms such as K. pneumoniae have been reported recently and leave physicians with few to no treatment options. This presentation focuses on the physico-chemical property space of antibacterial drugs and how an understanding of this property space can assist in the discovery and lead optimization of antibiotics, in particular that of antibacterial drugs active against Gram-negative bacteria. Specific examples will be presented and discussed in detail.

9:15 378 Physicochemical properties correlated with Gram-negative antibacterial activity of compounds in the Pfizer corporate library
Jeremy T Starr PhD, Rishi Gupta PhD, Veerabahu Shanmugasundaram PhD. Department of Antibacterials and Discovery Technologies, Pfizer Pharmatherapeutics Research and Development, Groton, CT, United States

Correlation of computed physicochemical properties of Pfizer proprietary compounds with their respective E. coli or P. aeruginosa MICs has led to the identification of a physicochemical fingerprint associated with higher probability of whole cell activity with a cytosolic target and presumed passive cell penetration. A computational tool has been designed to calculate a desirability quotient based on these parameters which demonstrates positive differentiation of higher scoring compound classes.

9:45 379 Combining lessons from computational design of gram positive antibacterials with datamining to aid the design of novel gram negative antibacterials
Charles J. Eyermann. Infection Discovery, AstraZeneca, Waltham, MA, United States

Our approach to address the emergence of resistant bacterial strains has been to identify new chemotypes with a novel mode of action. A significant effort has been made to develop novel inhibitors against gram positive strains like Methicillin-resistant Staphylococcus aureus (MRSA) These efforts have provided a number of key lessons related to target isozyme specificity and drug safety margins. Work to identify novel MurI inhibitors of H. pylori has also provided insights into the physiochemical properties that impact gram negative antibacterial activity. Combining the lessons learned from the above research efforts with datamining of existing gram negative agents provides a framework to aid in the optimization of novel leads for gram negative antibacterials.

10:15   Intermission
10:30 380 Targeting gram-negative pathogens: Drug design to improve antibiotics permeation?
Eric Hajjar, Amit Kumar, Paolo Ruggerone, Matteo Ceccarelli PhD. Department of Physics, Universita degli Studi di Cagliari and Sardinian, Monserrato, Italy

Gram-negative bacteria are protected by an outer membrane and to function, antibiotics have to diffuse passively through outer membrane channels, known as porins, such as OmpF in E.coli (Pages, J. M. et al. Nat. Rev. Microbiol. 2008, 6, 893). Bacterial strains can modulate their susceptibility to antibiotics by under-expressing or mutating the structures of porins, becoming resistant, in the worst case, to different antibiotics families. These multidrug resistant bacteria are now ubiquitous in both hospitals and the larger community and the resurrection of tuberculosis provides one ominous example highlighting the risk associated with evolved drug resistance (Cars, O. et al. Brit. Med. J. 2008, 337, 726). Moreover, many pharmaceutical companies abandoned this field and no truly novel active antibacterial compounds are currently in clinical trials. A major current dilemma for the pharmaceutical industry is whether to develop drugs for new targets or promote those drugs presently on the market (Weiss, D. et al. Nat. Rev. Drug. Discov. 2009, 8, 533.), identifying bottlenecks of existing antibiotics to suggest chemical modifications. Following such a strategy, we revealed the complete permeation pathways of b-lactams and fluoroquinolones antibiotics through porins using metadynamics simulations and found that experimental results remarkably confirmed the computational predictions. Further, simulations revealed its potentiality to overcome experimental limitations and provide microscopic details on the permeation process (Hajjar, E. et al. Biophys. J. 2010, 98, 569; Mahendran K. et al. J. Phys. Chem. B, IN PRESS).

Here we follow the paradigm for selecting antibiotics with better permeation properties using computer simulations only. Taking advantage of the atomic level of detail that the simulations provide we find that the diffusion of ampicillin through OmpF is governed by a subtle balance of interactions with partners in the porin channel: we draw, for the first time, the complete inventory of the rate-limiting interactions and map them on both the porin and antibiotics structure. Our methodology, which can be conveniently employed to study other porins/antibiotics, allows identifying the functional groups that govern optimal translocation. Such findings will directly benefit rational antibiotics design, by defining for example, some appropriate pharmacophores within high throughput screening strategies.

11:00 381 Structure-based lead optimization of novel bacterial type II topoisomerase inhibitors
Dr Neil D Pearson, Dr Zheng Yang, Dr Benjamin D Bax, Michael N Gwynn. Department of Antibacterial Chemistry, Infectious Diseases Center of Excellence in Drug Discovery, GlaxoSmithKline Pharmaceuticals, Collegeville, Pennsylvania, United States; Department of Computational and Structural Chemistry, GlaxoSmithKline Pharmaceuticals, Collegeville, Pennsylvania, United States; Department of Antibacterial Microbiology, Infectious Diseases Center of Excellence in Drug Discovery, GlaxoSmithKline Pharmaceuticals, Collegeville, Pennsylvania, United States; Department of Computational and Structural Chemistry, GlaxoSmithKline Pharmaceuticals, Stevenage, Hertfordshire, United Kingdom

The emergence of multi drug resistant Gram negative pathogens is a major concern given the paucity of new therapies in clinical development. GSK has discovered a novel series of inhibitors of both DNA gyrase and topoisomerase IV (NBTIs) with a unique mechanism and no target based cross resistance to established classes of antibacterials including the fluoroquinolones. Optimisation of the Gram positive selective early leads led to new series which afforded good activity versus Gram negative pathogens. GSK subsequently solved the first X-ray structure of a NBTI inhibitor in complex with S.aureus DNA gyrase and DNA providing unprecedented knowledge for lead optimization and the design of novel inhibitors. This talk will discuss how the structural information enabled the medicinal chemistry team to design new subunits as well as illustrating when optimization of interactions with the binding site have been well served by traditional medicinal chemistry.

11:30 382 Fragment-based development of tetrazole inhibitors against class A beta-lactamase
Yu Chen PhD. Department of Molecular Medicine, University of South Florida, Tampa, FL, United States

The production of beta-lactamases is the predominant cause of resistance to beta-lactam antibiotics, such as penicillins, in Gram-negative bacteria. Whereas high through-put screening has appeared insufficient for the development of new beta-lactamase inhibitors, fragment-based methods provide an effective approach in sampling novel chemical space in antibiotics discovery. We have previously used fragment-based molecular docking to identify mM range tetrazole inhibitors against CTX-M Class A beta-lactamase and to subsequently evolve their affinities to ~10 micromolar. New compounds have now been synthesized using the micromolar-affinity tetrazole scaffold, based on some similarities between this scaffold and beta-lactam antibiotics or on X-ray crystal structures of the inhibitor-bound complexes. Other fragment compounds have also been tested to probe regions of the active site not sampled by existing inhibitors. Combining the fragment-based approach with molecular docking, X-ray crystallography and chemical synthesis, we hope to eventually develop these tetrazole compounds into nM inhibitors.

12:00   Lunch
1:00   Introductory Remarks
1:05 419 Utilizing organicsyntheses and microbial iron assimilation processes for the development of newantibiotics
Prof. Marvin J. Miller. Chemistry and Biochemistry, University of Notre Dame, Notre Dame, IN, United States

Pathogenic microbes have rapidly developed resistance to all known antibiotics. To keep ahead in the “microbial war,” extensive interdisciplinary effort is needed. Resistance develops primarily to overuse of antibiotics that can result in alteration of microbial permeability, alteration of drug target binding sites, induction of enzymes that destroy antibiotics (ie, beta-lactamases) and even cause efflux of antibiotics. A combination of chemical syntheses, microbiological and biochemical studies will demonstrate that the known critical dependence of iron assimilation by microbes for growth and virulence can be exploited for the development of new approaches to antibiotic therapy. Iron recognition and active transport relies on the biosyntheses and use of microbe-selective iron chelating compounds called siderophores.

Our studies demonstrate that siderophores and analogs can be used for

-Iron transport-mediated drug delivery (“Trojan Horse”).

-Induction of iron limitation (Development of new agents to block microbial iron assimilation).

-Converting microbe-induced chemistry of iron into a process that is lethal to microbes.

1:45 420 Utilization of bacterial iron transport systems fordrug delivery
Dr. Ute Moellmann, Dr. Lothar Heinisch. Department of Molecular and Applied Microbiology, Leibniz Institute for Natural Product Research and Infection Biology - Hans Knoell Institute, Jena, Germany

The outer membrane permeability barrier is an important resistance factor of bacterial pathogens. In combination with other factors like drug inactivating enzymes, target alteration and efflux, it can increase resistance dramatically. A strategy to overcome this membrane mediated resistance is the misuse of bacterial transport systems. Most promising systems are those for iron transport. They are vital for virulence and survival of bacteria in the infected host, where iron depletion is a defense mechanism against invading pathogens. We synthesized biomimetic siderophores as shuttle vectors for active transport of antibiotics through the bacterial membrane. Structure activity relationship studies resulted in ampicillin siderophore conjugates highly active against Pseudomonas aeruginosa and other Gram-negative pathogens, which play a crucial role in destructive lung infections in cystic fibrosis patients and in severe nosocomial infections. The mechanism of action, in vitro and in vivo efficacy were demonstrated.

2:15 421 Activity of BAL30072, a novel siderophore sulfactam
Prof. Malcolm G P Page PhD. Basilea Pharmaceutica International Ltd, Basel, Switzerland

BAL30072 is a monocyclic b-lactam antibiotic belonging to the sulfactams. BAL30072 showed potent activity against multidrug-resistant (MDR) Pseudomonas aeruginosa and Acinetobacter spp., including many carbapenem-resistant strains. BAL30072 was bactericidal against both Acinetobacter spp. and P. aeruginosa, even against strains that produced metallo-b-lactamases that conferred resistance to all other b-lactams tested, including aztreonam. It was also active against many species of MDR Enterobacteriaceae, including isolates that had a class A carbapenemase or a metallo-b-lactamase. Unlike other monocyclic b-lactams, BAL30072 was found to trigger spheroplasting and lysis of E. coli, rather than the formation of extensive filaments. The basis for this unusual property is its inhibition of the bifunctional penicillin-binding proteins PBP 1a and PBP 1b in addition to its high affinity for PBP 3, which is the target of monobactams such as aztreonam.

2:45 422 Targeting bacterial multidrug efflux pumps
Olga Lomovskaya PhD, Scott Hecker PhD. Mpex Pharmaceuticals, San Diego, California, United States

Powerful techniques of modern drug discovery such as comparative genomics, ultra-high-throughput screening, structure-guided drug design and combinatorial chemistry have been used to identify novel targets and optimize novel, preferentially broads-spectrum antibiotics to combat antibiotic resistance. However, despite the fact that these employed targets are broadly conserved in bacteria, no drug candidate advanced using these methods has demonstrated relevant activity against most gram-negative bacteria. Thus, the outlook for new antibiotics appears unchanged from present in that of all approved classes of antibiotics, representatives of only three classes (fluoroquinolones, b-lactams and aminoglycosides) have clinical utility for the treatment of gram-negative bacteria such as Pseudomonas aeruginosa.

Multidrug resistance (MDR) efflux pumps play a prominent and proven role in gram-negative intrinsic resistance. Moreover, these pumps also play a significant role in acquired clinical resistance. Together, these considerations make efflux pumps attractive targets for inhibition in that the resultant efflux pump inhibitor (EPI)/antibiotic combination drug should exhibit increased potency, enhanced spectrum of activity and reduced propensity for acquired resistance. To date, at least one class of broad-spectrum EPI has been extensively characterized. While these efforts indicated a significant potential for developing small molecule inhibitors against efflux pumps, they did not result in a clinically useful compound. Stemming from the continued clinical pressure for novel approaches to combat drug resistant bacterial infections, a second-generation programs have been initiated based on a number of recent developments in the field, including structural elucidation of all three individual components of MDR efflux pumps and ligand-based insights into the mechanism-of-action of drug transporters. Building upon previous efforts, these new approaches show early promise to significantly improve the clinical usefulness of currently available and future antibiotics against otherwise recalcitrant gram-negative infections.

3:15   Intermission
3:30 423 Interaction of β-peptides with membranes
Jagannath Mondal, Dr. Xiao Zhu, Prof Qiang Cui, Prof Arun Yethiraj. Department of Chemistry, UW Madison, Madison, WI, United States

A new class of anti-microbial agents named b-peptides have recently been reported that show interesting sequence dependent activity and selectivity. In this work we investigate the interaction of these molecules with a model membrane in an effort to obtain physical insight into the mechanism of anti-microbial activity. We investigate the effect of sequence on the adsorption of these b-peptides to a membrane using computer simulations with both implicit and explicit solvent and membrane. Two classes of molecules are investigated: 10-residue oligomers of 14-helical sequences, and four sequences of random co-polymeric b-peptides. The oligomers of interest are two isomers, globally amphiphilic (GA) and non-GA, of two 10-residue 14-helical sequences. The penetration of the molecules into the membrane and the orientation of the molecules at the interface depend strongly on the sequence. We attribute this to the propensity of the b-phenylalanine (bF) residues for membrane penetration. The membrane adsorption studies are consistent with potential of mean force calculations using the same model. Results are similar when the membrane and solvent are treated in an implicit or explicit fashion. For the four sequences of random-co-polymeric b-peptides, the extent of stabilization of free-energy correlates with their efficiency to segregate the hydrophobic and cationic residues. The simulations are in qualitative accord with experiments on the minimum inhibitory concentration, and suggest simple strategies for the design of candidates for anti-microbial beta-peptides.

4:00 424 Molecular modeling of beta-lactamase inhibitors
Sookhee Nicole Ha, T. Blizzard, H. Chen, S. Kim, J. Wu, K. Young, Y. Park, A. Ogawa, S. Raghoobar, R. Painter, N. Hairston, S. Lee, A. Misura, T. Felcetto, P. Fitzgerald, N. Sharma, Jun Lu, E. Hickey, J. Hermes, M. Hammond. Merck & Co., Inc, Whitehouse Station, New Jersey, United States

Resistance against new antibiotics usually appears within few years after their marketing. Expression of the beta-Lactamase is the most common mechanism of resistance to the beta-Lactam antibiotics in Gram-negative bacteria. To maximize delaying the drug resistance, we have developed a beta-Lactamase inhibitor for combination therapy. We report our efforts on optimization of bridged mono-bactam analogs.

4:30 425 Assembly and function of large Gram-negative bacterial machines studied by molecular simulation integrated with experimental data
Prof. Matteo Dal Peraro. Institute of Bioengineering, Swiss Federal Institute of Technology, EPFL Lausanne, Lausanne, Switzerland

Gram-negative bacteria have evolved several means to attack their hosts and defend themselves from external attacks. Here, we use molecular simulations closely integrated with new experimental data to dissect the structural and dynamic features of the assembly mechanism of three large bacterial machines.

(i) We propose a four-helix model of E.coli PhoQ two-component system transmembrane domain, which is consistent with new experimental cross-linking data, and can explain the bacterial response to divalent cations and antimicrobial peptides. (ii) We study, with the aid of site-directed mutagenesis, the role of the pore-forming loop and the C-terminal pro-peptide for the heptamerization of pore-forming toxin aerolysin from A.hydrophila. Finally, (iii) we model the needle formation and regulation for the type III secretion system from Y.enterocolitica (injectisome) based on fresh genetic and mutagenesis results.

The full comprehension of the structural assembly of these bacterial machines can contribute, on one side, to unveil their fundamental biological function, and, on the other, will permit to develop rational strategies to specifically interfere with them for therapeutic intervention.

5:00 426 Design of potent, broad-spectrum AccC inhibitors
Li Xiao PhD, Cliff Cheng, Gerald W Shipps, Aileen Soriano, Peter Orth, Todd Black. Merck Research Laboratory, Kenilworth, New Jersey, United States; Merck Research Laboratory, Cambridge, Massachusetts, United States

The biotin carboxylase (AccC) is part of the multi-component bacterial acetyl coenzyme-A carboxylase (ACCase) and is essential for pathogen survival. We identified and validated AccC as an antibacterial drug target for our in-house AS/MS screen. An initial hit, 2-(2-chlorobenzylamino)-1-(cyclohexylmethyl)-1H-benzo[d]imidazole-5-carboxamide (1), was identified, and x-ray crystallography and computer modeling were utilized in its optimization. In this presentation we report our biology, chemistry and structure based drug design efforts in discovering a novel series of AccC inhibitors, exemplified by (R)-2-(2-chlorobenzylamino)-1-(2,3-dihydro-1H-inden-1-yl)-1H-imidazo[4,5-b]pyridine-5-carboxamide (2). These inhibitors are potent and selective for bacterial AccC with good cell-based activity against a sensitized strain of E. coli (HS294 E. coli).

THURSDAY MORNING

SCHB - Computer Modeling: The Wave of the Future and its Benefits for Small Business Owners
BCEC 212
Organized by: Masha Petrova
Presiding: Masha Petrova
8:30   Introductory Remarks
8:35 44 Best practices in scientific computer modeling
Dr. Masha V Petrova. Department of Research, MVP Modeling Solutions, LLC, Springfield, IL, United States

Computer modeling can help research organizations save a lot of money and time, if the modeling program is implemented correctly. Are you sure that your research group is making the most out of computer modeling? Attend this session to learn:

How companies and research groups tend to shoot themselves in the foot when setting up a computer modeling project;

What measures you can take to make sure that you don't spend a lot of time going down the wrong path or purchasing the wrong software;

The best way to take a scientific/engineering problem and translate it into computer modeling terms.

9:05 45 New wave of computational tools for the leads selection in biomedical industry
Dr. Aurora D. Costache PhD, Prof. Doyle D. Knight PhD, Prof. Joachim Kohn. New Jersey Center for Biomaterials, Rutgers - The State University of New Jersey, Piscataway, NJ, United States; Mechanical and Aerospace Engineering, Rutgers - The State University of New Jersey, 98 Brett Rd, Piscataway, NJ, United States

The high cost and intensive labor of developing new polymeric biomaterials for tissue engineering, drug delivery and other medical applications highlights the need for a change in the discovery process. As large corporations continuously look to cut costs, individual contractors or small businesses that can provide them with lead materials for given biomedical applications are expected to thrive. With this business niche in mind, the New Jersey Center of Biomaterials (NJCBM) created “Biomaterials StoreTM”- a computational tool specifically designed for development of new biomaterial leads. This integrated database and datamining tool allows the user to create/use large databases of virtual polymer libraries and to apply modeling tools to predict relevant polymer properties and biological responses to biomaterials. Based on the requirements for a specific application, the most promising candidates are selected for synthesis and complete experimental evaluation, thus accelerating the discovery process and cutting costs at the same time.

9:35 46 Computational modeling of soft condensed matter and biomaterials
Dr. Jayeeta Ghosh. New Jersey Center for Biomaterials, Rutgers, Piscataway, NJ, United States

Computational modeling helps understand chemistry starting from quantum level to process dynamics length and time scales.

This presentation will discuss the application of atomistic and mesoscale modeling for soft condensed matters as well as combinatorial computational approach to biomaterials invention. The main objective is to show the relevance and importance of detailed molecular modeling versus approximate surrogate modeling.

Molecular modeling of soft condensed matters including glasses, polymers and lipids will be discussed in the context of industrial application and drug delivery.

Quantitative structure property relation (QSPR) modeling approach for identifying suitable biomaterials starting from a large combinatorial library of polymers for tissue engineering and biomedical applications, can help reduce the experimental cost and time and advance business.

10:05 47 First-principles computational approach for the characterization and design of novel organic electronic materials
Roel S Sanchez-Carrera PhD, Prof. Alan Aspuru-Guzik. Deparment of Chemistry and Chemical Biology, Harvard University, Cambridge, MA, United States

Organic electronics have recently emerged as a technology that will revolutionize the way in which we visualize information, generate energy from renewable resources, and communicate with people around the world. Thus, various international academic laboratories and major chemical companies are actively involved in the fine-tuning and development of the molecular materials used in the field of organic electronic devices.

To highlight the potential of current computational methodologies, in this study, on the basis of quantum chemistry calculations and molecular dynamics simulations, we investigate the microscopic charge transport parameters of one of the most outstanding candidates, the dinaphtho-thieno-thiophene organic semiconductor. The good agreement found in this work between observed and computed properties, stresses the importance of using computational chemistry techniques to identify suitable molecular materials for the emerging field of organic electronics.

10:35   Intermission
10:50 48 Recent advances in structure-based drug design
Woody Sherman. Schrodinger, New York, NY, United States

Structure-based drug design is an important part of the drug discovery process and recent methodological advancements, as well as increased computing resources have resulted in a growing number of success stories. In this presentation, we highlight some of the most promising methods and applications, including the accurate assessment of water free energies, incorporation of protein flexibility into docking algorithms, and structure-based modeling of GPCRs. In addition, we describe the most significant limitations in the existing methods and provide a development roadmap to overcome these limitations.

11:20 49 Computer simulation of ligand binding to a flexible protein target
Dr Philip W Payne. Consulting, InterBiotics LLC, Sunnyvale, CA, United States

A research-based biotechnology or pharmaceutical business must focus capital and labor on the experiments that will most rapidly discover or refine intended products. Computer simulations are useful adjuncts to an experimental program when they provide structural insights that suggest how a protein or ligand structure should be modified to improve a measured outcome - enzymatic rate, receptor activation, or ligand affinity; the successful modeling program means that fewer proteins need to be mutated or fewer ligands synthesized during a product development campaign.

Important biological functions often entail large displacements of protein main chains or loops, and industrially useful modeling of protein structure needs to assess such motion and its impact on protein-ligand affinity or ligand-directed signaling. Unfortunately, there is little commercial software that can cost-effectively predict important protein motions. Therefore we have developed a strategy (Inverse Docking) for analyzing main chain movements that conform a G-Protein Coupled Receptor (Dopamine D2S) to a nanomolar D2 antagonist, spiperone.

11:50 50 FAST Predictions of protein stability and flexibility
Prof. Dennis R. Livesay, Dr. Hui Wang, Prof. Donald J. Jacobs. Department of Bioinformatics and Genomics, University of North Carolina at Charlotte, Charlotte, NC, United States; Department of Physics and Optical Science, University of North Carolina at Charlotte, Charlotte, NC, United States

Accurate descriptions of stability and flexibility are necessary for a complete understanding of protein structure and function. As such, we have developed “FAST” to provide a Flexibility And Stability Test on proteins in aqueous solutions. Herein, all intramolecular interactions are assigned enthalpy and entropy values. Total enthalpy is the sum of all components, whereas efficient graph-rigidity algorithms account for entropy nonadditivity. FAST has been designed from the ground-up to account for dependence on temperature, pressure, pH, salt concentration, etc. As such, free energy landscapes as a function of multiple thermodynamic variables can be quickly calculated. FAST also calculates a wide variety of mechanical properties related to structural rigidity and flexibility with virtually no increase in computational expense. This talk will summarize our general approach, and recent improvements in regards to speed and accuracy. Support for this work has been from grants from the NIH (R01-GM073082) and the Charlotte Research Institute.

12:20 51 Patentability of computer simulations and models
Noah Malgeri. Law Office of Noah V. Malgeri, Uxbridge, Massachusetts, United States

In recent years, several companies and individuals, including major industry leaders, have filed patent applications for computer models, particularly in the area of control systems, project management simulations and for modeling pathologies. This presentation will address the subject of scientific software patentability.

12:50   Concluding Remarks

THURSDAY AFTERNOON

COMP - Data Analysis: Statistics on Chemicals
BCEC 157 C
Organized by: Emilio Esposito
Presiding: Alemayehu Abebe
1:30 433 Exploring protein conformational changes with accelerated molecular dynamics in NAMD
Dr. Yi Wang, Prof. J. Andrew McCammon. Chemistry and Biochemistry, Howard Hughes Medical Institute, University of California, San Diego, La Jolla, CA, United States

Accelerated molecular dynamics (aMD) enhances conformational space sampling by reducing energy barriers separating different states of a system. Here we present the implementation of aMD in the highly efficient parallel molecular dynamics program NAMD and offer exemplary applications performed on systems up to 60,000 atoms. Our results indicate that while providing significantly enhanced sampling, aMD simulations have only a small overhead in comparison to classical MD simulations. A 10-ns aMD simulation performed on the bacterial enzyme RmlC successfully revealed its transition from apo- to holo- state, which is not observed in a 50-ns classical MD simulation. We demonstrate that aMD can be applied efficiently to explore the conformational changes of complex biomolecules, especially when little is known about their alternative structures and transition reaction coordinates.

2:00 434 Pseudo-chair conformation of carboxyphosphate
Venkata S Pakkala, Steven M Firestine, Jeffrey D Evanseck. Department of Chemistry and Biochemistry, Duquesne University, Pittsburgh, Pennsylvania, United States; Eugene Applebaum College of Pharmacy and Health Sciences, Wayne State University, Detroit, Michigan, United States

For over 40 years, carboxyphosphate has been postulated as a key intermediate in several carboxylase enzymes. Unfortunately, this compound is extremely unstable (t1/2 of 70 ms), thus precluding direct experimental studies. Therefore, we have utilized high level ab inito (MP2 and CCSD(T)), DFT (B3LYP, BB1K, M05-2X, M06-2X and MPW1K) and ONIOM(DFT:AMBER) methods to investigate the structure and energetics of carboxyphoshpate in vacuum, in a PCM continuum solvation model and in the active site of N5-CAIR synthetase, an enzyme shown to proceed via the formation of carboxyphosphate. We report here, for the first time, that carboxyphosphate adopts a “pseudo-chair” conformation and calculations reveal that this conformation is found to be the most stable in vacuum, solvent and the active site. This study has implications in the development of the carboxyphosphate analogs as potential inhibitors, in understanding the instability of the compound, and in elucidating the mechanisms of enzymes utilizing this compound.

2:30 435 Analysis of vibrational spectra of polypeptides in terms of localized vibrations
Dr. Christoph R Jacob, Prof. Markus Reiher. Center for Functional Nanostructures, Karlsruhe Institute of Technology (KIT), Karlsruhe, Germany; Laboratorium für Physikalische Chemie, ETH Zurich, Zurich, Switzerland

While nowadays efficient quantum chemical methods allow for the calculation of vibrational spectra of large (bio-)molecules, such calculations also provide a large amount of data. In particular for the vibrational spectra of polypeptides, a large number of close-lying normal modes contribute to each of the experimentally observed bands, which hampers the analysis of the calculated spectra considerably.

Here, we discuss how vibrational spectra obtained from quantum chemical calculations can be analyzed by transforming the calculated normal modes contributing to a certain band in the vibrational spectrum to a set of localized modes [1]. We demonstrate that these localized modes are more appropriate for the analysis of calculated vibrational spectra of polypeptides and proteins than the delocalized normal modes.

We apply this methodology to investigate the influence of the secondary structure on infrared and Raman spectra of polypeptides [2]. As a model system, a polypeptide consisting of twenty (S)-alanine residues in the conformation of an a-helix and of a 310-helix is considered. In particular, we show how the use of localized modes facilitates the analysis of the positions and of the total intensities of the bands in the vibrational spectra, and how the couplings between localized modes determine the observed band shapes. Finally, this analysis is applied to analyze the Raman optical activity (ROA) spectra of these helical polypeptides, which provides a detailed picture of the generation of ROA bands in proteins [3].

[1] Ch. R. Jacob and M. Reiher, J. Chem. Phys. 130 (2009), 084106.

[2] Ch. R. Jacob, S. Luber, M. Reiher, J. Phys. Chem. B 113 (2009), 6558.

[3] Ch. R. Jacob, S. Luber, M. Reiher, Chem. Eur. J. 15 (2009), 13491.

3:00   Intermission
3:15 436 Conformational coupling between LOV and kinase domains in phototropins: A computational perspective
Dr. Marco Stenta PhD, Prof. Matteo Dal Peraro PhD. Department of Bioengineering, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Vaud, Switzerland

Phototropins constitute an important class of plant photoreceptors playing key roles in many physiological responses to light, including phototropism, chloroplast movement and stomata opening. Phototropins feature, along with a serine-threonine kinase domain, two LOV (light-, oxygen- or voltage-regulated) domains, each binding a FMN (flavin mononucleotide). Blue light affects the kinase domain by triggering, in the LOV domain, the formation of a covalent intermediate between the FMN cofactor and a nearby cysteine residue. Despite X-ray structures provided solid ground for mechanicistic hypothesis, the molecular details of the inter-domain communication process are still unknown. By using accurate QM/MM (quantum mechanics/molecular mechanics) calculations we investigated the formation/breaking of the FMN/Cys covalent intermediated. We investigated the coupling between the LOV and kinase domains by means of long MD (molecular dynamics) simulations and detailed PES (potential energy surface) explorations (MM level).

Zoltowski, B. D.; Vaccaro, B.;

Crane, B. R. Nat Chem Biol 2009, 5, 827-834.

3:45 437 Conformational sampling of macrocycles through accelerated molecular dynamics simulation
S. Roy Kimura Ph.D.. Department of Computer Assisted Drug Design, Bristol Myers Squibb, Wallingford, CT, United States

Macrocyclization is a strategy used in medicinal chemistry to lock a molecule in its bioactive conformation. The resulting decrease in conformational flexibility often leads to higher potencies due to the reduced entropy loss upon binding, and sometimes improved physical chemical properties such as bioavailability. Conformational searches of macrocycles are usually performed by temporary ring opening and Monte Carlo (MC) sampling to overcome the energy barriers between low energy states. However, widely available MC algorithms can only be used in conjunction with simplified continuum solvents such as dielectrics or Generalized Born-related models. In this study, we assess the use of molecular dynamics simulation in explicit solvent with periodic high-temperature pulsing as a method to overcome the characteristic energy barriers of macrocycles. The pros and cons of this methodology versus MC sampling are discussed.

4:15 438 WITHDRAWN

 

 

Publishing software