Scientific Report 2005

Molecular Biology

Computation and Visualization in Structural Biology

A.J. Olson, D.S. Goodsell, M.F. Sanner, A. Gillet, Y. Hu, R. Huey, C. Li, S. Karnati, W. Lindstrom, G.M. Morris, A. Omelchenko, M. Pique, B. Norledge, R. Rosenstein, D. Stoffler, Y. Zhao

In the Molecular Graphics Laboratory, we develop novel computational methods to analyze, understand, and communicate the structure and interactions of complex biomolecular systems. This past year, we showed the effectiveness of 3-dimensional molecular models as a tangible human-computer interface in educational and research settings. Within our component-based visualization environment, we continue to develop methods for predicting biomolecular interactions, analyzing biomolecular structure and function, and presenting the biomolecular world in education and outreach.

We have applied these methods to several important systems in human health and welfare. We continue the search for inhibitors of HIV protease to fight the growing problem of drug resistance in HIV disease. We used AutoDock, a suite of programs for predicting bound conformations and binding energies for biomolecular complexes, in the virtual screening of large databases of compounds and ultimately identified new compounds for use in the treatment of cancer. We used methods for predicting protein interactions to probe the mechanism of blood coagulation.

Tangible Interfaces for Structural Biology

We are using the evolving technology of computer autofabrication (“3-dimensional printing”) to produce physical models of complex molecular assemblies (Fig. 1).

Fig. 1. A sample of the molecular models built by using automated fabrication techniques shows a wide range of molecular representations, scales, and sizes.

With this technology, a physical model based on a virtual computer model is built up layer by layer. The great advantage of autofabrication is that nearly any shape can be built; the shape is limited only by the imagination of the researcher and the structural integrity of the building material. We have used 2 technologies: 1 that is much like using a hot glue gun, in which the model is built from layers of molten plastic, and 1 in which gypsum powder and colored binders applied with an ink jet technology are used to create full-color models.

In collaboration with the Human Interfaces Technology Laboratory at the University of Washington, Seattle, Washington, we developed an augmented reality environment that embeds these 3-dimensional models within the virtual environment of the computer. The goal of this technology is to create a sense of user presence in a computational interaction, combining the intuitive tactile interaction of model manipulation with the rich bioinformatics and visualization tools that are available in the computer environment. As shown in Figure 2, the augmented reality environment tracks the position of the model, displaying a video image of the model and user and overlaying a computer-generated image that is spatially registered with the model as the user manipulates and explores the structure.

Fig. 2. Top, The augmented reality environment. The user holds the model under a FireWire camera. Bottom, A video image of the model is displayed on the computer screen, with an overlaid computer-generated image. Here, the electrostatic potential and field of superoxide dismutase are shown with volume-rendered clouds and small animated arrows.

In tests of the model, high school and college students reported that they experienced a compelling sense of realism of the virtual object and enhanced interaction with the subject matter.

We use the program Python Molecule Viewer to create a diverse range of different representations for both our virtual molecular objects and our tangible models, simplifying integration of the models with the virtual environment. Python Molecule Viewer allows us to combine backbone representation, atomic representations, and surfaces and to incorporate markers for spatial tracking. We are also using computer-aided design and manufacturing methods to design mechanical connectors and magnetic fittings that incorporate aspects of flexibility and interaction into the models. Vision, the visual programming interface, is used to integrate nonmolecular features and properties, such as electrostatics and hydrophobicity, into the virtual and physical environment.

Component-Based Visualization Environment

To facilitate the integration and interoperation of computational models and techniques from a wide variety of scientific disciplines, we continue to expand our component-based software environment. The environment is centered on Python, a high-level, object-oriented, interpretive programming language. This approach allows the compartmentalization and reuse of software components. Python provides a powerful “glue” for assembling computational components and, at the same time, a flexible language for the interactive scripting of new applications.

We recently added a visual programming environment, Vision, that supports the interactive and visual combination of computational nodes into networks that correspond to algorithms coded at a high level (Fig. 3).

Fig. 3. Vision, a visual programming environment, allows users to build networks of visualization software, creating new computational pipelines and novel visualizations of data. The canvas is shown at the center, where users interactively combine computational nodes. The network shown is a visualization of an electron micrograph reconstruction of a virus, colored by the radial depth and with a sector removed to show the interior structure.

Vision provides nonprogrammers an intuitive interface for building networks that describe new computational pipelines and novel visualizations of data. The basic molecular visualization methods of Python Molecule Viewer, a molecular symmetry generator, and a volume-rendering method are a few of the currently available nodes, and new nodes are easy to create in the Python language. The combination of the visual programming model and the ability to interactively inspect and edit nodes written in a high-level language creates an unprecedented number of levels at which users can interact with the program. The software tools developed by using our software components have been distributed to more than 10,600 users, with an average of 250 downloads a month during the past year.

We released a new version of our software tools in December 2004 that contains a large number of improvements and additions. In particular, we streamlined our distribution mechanism and included concurrent versioning system entries that allow users to update the software once it has been installed. We fixed several bugs and added new packages, including mesh decimation algorithms and support for manipulating and visualizing volumetric data. In addition, we increased the number of tests that are run on a nightly basis to more than 2500.

Modeling of Flexibility

In a project funded by the National Institutes of Health, we developed Flexibility Tree, a hierarchical and multiresolution representation of the flexibility of biological macromolecules that can be used in computational simulations. With this software, a user can encode a small subset of a protein’s conformational subspace. After implementing the core infrastructure of Flexibility Tree and integrating it with Python Molecule Viewer and Vision, we are building such trees for molecular systems, including HIV type 1 protease and protein kinases.

A number of laboratories around the world have developed software tools for extracting the information that describes how the various parts of proteins move relative to each other. We are now using Flexibility Tree to assess the quality of the decomposition of the protein structure into rigid bodies provided by these tools as well as the accuracy of the motions calculated by using these methods. Early results indicate that when small local perturbations are allowed in addition to the motions predicted by these tools, the Flexibility Tree covers a conformational space that includes both open and closed conformations of our test systems with accuracy sufficient for docking experiments. Our next step will be to design prototype docking tools that can include protein flexibility based on the Flexibility Tree.

Virtual Screening with Autodock

We have developed new interactive tools to streamline the process of virtual screening in AutoDock. With these tools, users can perform docking experiments to evaluate the binding of a database of molecules with a particular macromolecule of interest. In collaboration with I.A. Wilson, Department of Molecular Biology, we used the method to discover new inhibitors for aminoimidazole carboxamide ribonucleotide transformylase, a target for new cancer chemotherapeutic agents. The diversity set from the National Cancer Institute was screened, and 44 potential candidates were identified. In vitro inhibition assays indicated that 8 of the 44 were soluble compounds, had chemical scaffolds that differed from the general folate template, and caused inhibition when used in micromolar concentrations. Currently, we are optimizing the lead candidates; our goal is to obtain novel nonfolate inhibitors.

AutoDock is currently used in more than 3200 academic and commercial laboratories worldwide. We continued development of AutoDock by testing a new empirical free-energy force field. The force field incorporates a charge-based model for evaluation of hydrophobicity and an improved method for evaluating the geometry of hydrogen bonding. The force field was calibrated by using a set of 138 protein complexes of known structure taken from the Ligand Protein Database from the laboratory of C.L. Brooks, Department of Molecular Biology. We anticipate that the revised AutoDock, which incorporates this new force field and methods for selective flexibility in the protein target, will be released in 2005.

We also used AutoDock to predict intermolecular interactions in several biological systems. In collaboration with C.F. Barbas, Department of Molecular Biology, we investigated the binding of peptides to the catalytic aldolase antibody 93F3. To explore the large conformational space available to these peptides, we used a divide-and-conquer approach that separates the search space into searchable blocks. In studies with G. Legge, University of Texas, Austin, Texas, we explored the interaction between the cytoplasmic tail of tissue factor and the WW domain of proline isomerase PIN1, focusing on the interaction of several key phosphoserine residues.

Fighting Drug Resistance in HIV Disease

We are continuing our work on inhibitors to fight drug resistance in the treatment of AIDS (Fig. 4).

Fig. 4. The predicted bound conformation of sanguinarine, a potential lead compound for the development of novel HIV protease inhibitors.

In collaboration with K.B. Sharpless and C.-H. Wong, Department of Chemistry, we have focused on the design of inhibitors that assemble within the active site of HIV protease. We showed that the triazole formed in the click chemistry reaction is an effective mimic for the peptide group in traditional inhibitors, forming similar hydrogen-bonding interactions.

Currently, we are moving the FightAIDS@Home system from an outside provider to a new server strategy that will be implemented in the Molecular Graphics Laboratory. FightAIDS@Home enlists the worldwide community in a large computational effort to design effective therapeutic agents to fight AIDS. Personal computers are used in the program when the computers are not in use by their owners, providing an enormous, and largely untapped, computational resource. The current goal is to identify inhibitors that are effective against the wild-type virus and against common mutant forms of the virus. The large computational resources provided by FightAIDS@Home enables the screening of large databases of compounds and use of multiple mutant targets, allowing estimation of the potential of a compound to remain effective when viral mutations occur that cause resistance to drugs currently used to treat HIV disease.

Predicting Protein-Protein Interactions

With the goal of creating a comprehensive tool for predicting protein-protein interactions, we incorporated both SurfDock and AutoDock into the Python programming environment. SurfDock uses a variable-resolution spherical harmonics representation to find candidate orientations, and AutoDock is then used to explore local atomic rearrangements at the interface. We tested the method on a set of 59 protein-protein complexes of known structure and optimized the level of smoothing used in the spherical harmonics approximation of the molecular surfaces. The results of the docking test depended on the force field used to score possible orientations. The best results were obtained with a residue-based pair-wise potential of mean force.

Visual Methods from Atoms to Cells

Understanding structural molecular biology is essential to foster progress and critical decision making among students, policy makers, and the general public. In the past year, we continued our longstanding commitment to science education and outreach with a combination of presentations, popular and professional illustrations and animation, 3-dimensional tangible models, and a presence on the Worldwide Web. In these projects, we use the diverse visualization tools developed in the Molecular Graphics Laboratory to disseminate results that range from atomic structure to cellular function.

We created a 3-dimensional model that demonstrates viral assembly. The model is composed of pentamers from the structure of poliovirus, with embedded magnets on the interacting faces. When 12 or more of these pentamer models are placed in a closed container and gently shaken, they self-assemble in a matter of seconds to form a spherical capsid.

We also continued several regular features that informally present molecular structure and function. The “Molecule of the Month” at the Protein Data Bank (http://www.rcsb.org/pdb) provides an accessible introduction to this central database of biomolecular structure. Each month, a new molecule is presented with a description of its structure, function, and relevance to health and welfare (Fig. 5).

Fig. 5. Three different types of catalase. Catalase was presented as a Molecule of the Month in 2004 after a request from a high school teacher.

Visitors are then given suggestions about how to begin their own exploration of the structures in the data bank. Currently, we are collaborating with T. Herman, Milwaukee School of Engineering, Milwaukee, Wisconsin, to combine material from the “Molecule of the Month” with 3-dimensional models and multimedia tutorials to create educational modules for use at high school and college levels. Other projects include “The Molecular Perspective,” articles in the journal The Oncologist that present structures of interest to clinical oncologists and provide a source of continuing education for physicians, and “Recognition in Action,” a new series in the Journal of Molecular Recognition.

Publications

Berman, H.M., Ten Eyck, L.F., Goodsell, D.S., Haste, N.M., Kornev, A. Taylor, S.S. The cAMP binding domain: an ancient signaling module. Proc. Natl. Acad. Sci. U. S. A. 102:45, 2005.

Brik, A., Alexandros, J., Lin, Y.-C., Elder, J.H., Olson, A.J., Wlodawer, A., Goodsell, D.S., Wong, C.-H. 1,2,3-Triazole as a peptide surrogate in the rapid synthesis of HIV protease inhibitors. Chembiochem 6:1167, 2005.

Gillet, A., Sanner, M., Stoffler, D., Goodsell, D.S., Olson, A.J. Augmented reality with tangible auto-fabricated models for molecular biology applications. In: IEEE Visualization: Proceedings of the Conference on Visualization ’04. IEEE Computer Society, Washington, DC, 2004, p. 235.

Gillet, A., Sanner, M., Stoffler, D., Olson, A. Tangible augmented interfaces for structural molecular biology. IEEE Comput. Graph. Appl. 25:13, 2005.

Gillet, A., Sanner, M., Stoffler, D., Olson, A. Tangible interfaces for structural molecular biology. Structure (Camb.) 13:483, 2005.

Goodsell, D.S. Computational docking of biomolecular complexes with AutoDock. In: Protein-Protein Interactions: A Molecular Cloning Manual, 2nd ed. Golemis, E., Adams, P. (Eds.). Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, in press.

Goodsell, D.S. The molecular perspective: cyclins. Oncologist 9:592, 2004; Stem Cells 22:1121, 2004.

Goodsell, D.S. The molecular perspective: cytochrome c and apoptosis. Oncologist 9:226, 2004; Stem Cells 22:428, 2004.

Goodsell, D.S. The molecular perspective: L-asparaginase. Oncologist 10:238, 2005; Stem Cells 23:710, 2005.

Goodsell, D.S. The molecular perspective: major histocompatibility complex. Oncologist 10:80, 2005; Stem Cells 23:454, 2005.

Goodsell, D.S. The molecular perspective: morphine. Oncologist 9:717, 2004; Stem Cells 23:144, 2005.

Goodsell, D.S. The molecular perspective: nicotine and nitrosamines. Oncologist 9:353, 2004; Stem Cells 22:645, 2004.

Goodsell, D.S. The molecular perspective: polycyclic aromatic hydrocarbons. Oncologist 9:469, 2004; Stem Cells 22:873, 2004.

Goodsell, D.S. Recognition in action: flipping pyrimidine dimers. J. Mol. Recognit. 18:193, 2005.

Goodsell, D.S. Representing structural information. In: Current Protocols in Bioinformatics. Baxeranis, A.D., Davison, D.B. (Eds.). Wiley & Sons, Hoboken, NJ, in press.

Goodsell, D.S. Visual methods from atoms to cells. Structure (Camb.) 13:347, 2005.

Li, C., Xu, L., Wolan, D.W., Wilson, I.A., Olson, A.J. Virtual screening of human 5-aminoimidazole-4-carboxamide ribonucleotide transformylase against the NCI diversity set by use of AutoDock to identify novel nonfolate inhibitors. J. Med. Chem. 47:6681, 2004.

Sanner, M.F. A component-based software environment for visualizing large macromolecular assemblies. Structure (Camb.) 13:447, 2005.

Sanner, M.F. Using the Python programming language for bioinformatics. In: Encyclopedia of Genetics, Genomics, Proteomics and Bioinformatics. Jorde, L.B., Little, P.F.R., Dunn, M.J., et al. (Eds.). Wiley & Sons, Hoboken, NJ, in press.

Zhu, X., Tanaka, F., Hu, Y., Heine, A., Fuller, R., Zhong, G., Olson, A.J., Lerner, R.A., Barbas, C.F. III, Wilson, I.A. The origin of enantioselectivity in aldolase antibodies: crystal structure, site-directed mutagenesis, and computational analysis. J. Mol. Biol. 343:1269, 2004.