Highlight
A Crystal Ball for Structural Biologists
IGERT Crystal Ball Designers at Iowa State University
Achievement/Results
Protein structure determination is an essential component of modern biological research because the three-dimensional structure of a protein provides key insights into its functions and regulatory mechanisms. BCBLab, an NSF-funded student-run computational biology laboratory, is developing an innovative Crystal Ball Database that will, both figuratively and literally, shed new light on a long-standing problem for structural biologists: protein crystallization. Although the generation of high quality protein crystals is a critical prerequisite for determining the structure of proteins using X-ray crystallography, conditions that promote protein crystal formation are not fully understood. The goal of the Crystal Ball Database project (Crystal Ball) is to transform the “art” of crystal generation into a knowledge-based science.
Crystal Ball was conceived by a group of Ph.D. students enrolled in an NSF Integrative Graduate Education and Research Training (IGERT) Program, in collaboration with the Macromolecular X-ray Crystallography Facility at Iowa State University (ISU). The effort is one of several ongoing interdisciplinary research projects being tackled by BCBLab, which is a student-based initiative designed to provide bioinformatics and computational biology expertise and services to researchers at ISU and New Mexico State University (NMSU).
BCBLab was founded by NSF IGERT students in the Bioinformatics and Computational Biology (BCB) graduate program at ISU. Now in its third year, BCBLab has grown to involve students and faculty at both ISU and NMSU, which are partner institutions in an NSF IGERT program in Computational Molecular Biology, directed by Dr. Drena Dobbs (ISU) and Dr. Desh Ranjan (NMSU). BCBLab helps connect biological scientists on both campuses with BCB graduate students, who can bring bioinformatics tools and practices to bear on problems in the biological sciences. In the process, students obtain hands-on experience in solving real-life bioinformatics and computational biology problems, while developing their bioinformatics skills and analytical techniques. Crystal Ball provides an online database and software to assist X-ray crystallographers in determining protein structures. In X-ray crystallography, three-dimensional structure information is extracted from the diffraction pattern generated when a protein crystal is bombarded with powerful X-ray beams. The pattern of diffraction “spots” generated as a result of the beam’s angle and intensity is used to determine electron densities in a crystal sample, and, ultimately, to infer the positions of individual atoms within a protein crystal.
Technical challenges involved in successful crystallization often constitute the most vexing bottleneck in protein structure determination. Protein crystallization requires the manipulation of a large number of experiment parameters (e.g., buffer components, temperature, pH, protein concentration) to successfully generate a high quality crystal. Crystallographers typically use commercially available screening kits to begin their protein crystallization trials. When initial screens produce favorable results, it is possible to narrow the choice of solution components and identify optimal conditions for crystallization. If favorable results are not obtained in initial screens, crystallographers can spend substantial resources and time (years, in some cases!) performing numerous, additional trials, many times without success.
Crystal Ball is designed to direct researchers to crystallization protocols most likely to be successful, thus eliminating much of the “trial and error” associated with traditional crystallization procedures. In Crystal Ball, results and knowledge from crystallization experiments are compiled and organized in a web-based database and server. Crystal Ball also provides user-friendly software for extracting and analyzing experimental data from crystallization trials, including specific experimental conditions used in both successful and unsuccessful attempts to produce protein crystals.
A unique feature of Crystal Ball is that it provides statistical data regarding both positive (successful) and negative (failed) crystallization experiments. This is important because it permits the application of machine learning algorithms to the protein crystallization problem. Machine learning algorithms can be trained and evaluated on a large dataset of experimental conditions that do or do not produce crystals for known proteins, and then used to predict optimal crystallization conditions for a novel protein sample. Effectively exploiting computational methods to address the protein crystallization problem requires an integrated approach: it demands both a detailed understanding of the strengths and limitations of various machine learning algorithms, and an appreciation of the nuances of protein crystallization variables and procedures. Such understanding is possible only in a truly interdisciplinary training environment.
Results presented by IGERT students at the 2009 NMBIS Conference in Sante Fe suggest that predictions based on Crystal Ball data can, in fact, identify appropriate crystallization conditions for specific proteins of interest. Ultimately, the knowledge and software tools assembled in the Crystal Ball Database – a direct outcome of interdisciplinary collaborations fostered by the IGERT training grant – should greatly accelerate protein crystallization experiments and provide substantial savings in the expense and effort required to experimentally determine protein structures.
Address Goals
The IGERT-supported BCBLab both fosters innovative research that advances the frontiers of knowledge and cultivates a world-class science and engineering workforce by promoting student-motivated research in an interdisciplinary team-based learning environment.
The Crystal Ball Database project illustrates how the uniquely interdisciplinary approach sponsored by the IGERT program has generated an exciting new strategy that promises to advance the frontiers of knowledge in experimental protein structure determination. Because detailed three-dimensional information about the structure of a protein provides important insights into its mechanism and cellular functions, protein structure determination is an essential component of biological research in areas ranging from plant drought tolerance to drug design. Due to the considerable expense and difficulty involved in generating protein crystals, however, the number of experimentally-determined protein structures has lagged far behind the number of protein sequences currently available from genomic sequencing efforts. The powerful integrated computational and biophysical approach taken by the Crystal Ball Database project has the potential to help overcome a major stumbling block that has long hampered the determination of protein structure. Improving the success rate of protein crystallization experiments will could significantly accelerate experimental determination of structures using X-ray crystallography.
The BCBLab has also helped cultivate a world-class science and engineering workforce by promoting student-motivated research in an interdisciplinary team-based learning environment. BCBLab includes scientists with formal training in a wide range of traditional disciplines, including: agronomy; animal science: biophysics; chemistry; chemical engineering; computer science; ecology; evolution; genetics; mathematics; molecular, cellular and developmental biology; physics; plant pathology; plant physiology; statistics; veterinary microbiology; veterinary pathology. The Crystal Ball project and several other research projects undertaken by the BCBLab have led to robust scientific collaborations among graduate and undergraduate students and faculty with backgrounds in these diverse disciplines. Moreover, the success of these efforts has generated institutional commitments to more firmly integrate interdisciplinary efforts into traditionally “wet-lab” focused biological research efforts at both ISU and NMSU. Thus, the IGERT training program is expanding the scientific literacy of many individuals at ISU and NMSU. Specific tangible benefits of BCBLab include:
- providing bioinformatics expertise and service to the university communities at large;
- providing a diverse training ground for both graduate and undergraduate bioinformaticists;
- informing faculty and students regarding the problem-solving capabilities (and limitations) of computational and statistical approaches in bioinformatics and computational biology
- increasing visibility and awareness of the effectiveness of integrated interdisciplinary efforts (as compared with traditional disciplinary or multi-disciplinary efforts)
- serving as a melting pot for biological scientists, statisticians, computer scientists, and engineers, and generating new collaborations that lead to additional future interdisciplinary research projects;
- energizing not just BCBLab participants, but the larger communities that participate and collaborate with IGERT students and faculty at both ISU and NMSU
