Research Interests

I. Outline: Molecular modeling package "GALAXY"

The ultimate goal of our research is to understand and to predict the functions of biomolecules
in atomic details with theoretical and computational methods. Such research can also provide
basic tools for the design of small molecules or proteins that can contribute to promoting human health.

For such purposes, we have been developing a biomolecular modeling program package
GALAXY and a web server GalaxyWeb implementing the GALAXY methods [1]
(click here to connect to the web server,

GALAXY is a software package for biomolecular modeling which consists of core programs
that perform basic modeling tasks such as protein structure prediction, protein-ligand docking,
and protein-protein docking and applied programs aimed at more specific goals such as ligand
binding site prediction or protein homo-oligomeric state prediction. The GALAXY program package
has been designed to be modular, enabling convenient addition of new programs in the future.

The philosophy underlying the GALAXY development is to provide more "practically useful"
information to experimentalists, such as more "reliable" model structures, by fusion of available
tools and knowledge in diverse disciplines such as chemistry, physics, computer science, or bioinformatics.
GALAXY currently utilizes various scoring functions (e.g. molecular mechanics,
implicit solvation models, statistical scores) and sampling techniques (e.g. molecular dynamics,
genetic algorithm, Monte Carlo simulation, conformational space annealing) to achieve the best
performance in each modeling task.

II. Protein Structure Prediction

    Template-based modeling in CASP experiments

    The CASP (Critical Assessment of the techniques for protein Structure Prediction) experiment
    is a community-wide protein structure prediction competition in which more than 100 groups from
    various countries test their prediction methods. In our first participation in CASP (CASP9, 2010),
    our automated prediction server "Seok-server" (based on the "GalaxyTBM" method [2])
    was evaluated to be "among the top six servers" in the template-based modeling (TBM)
    category (the largest and the most important category in CASP)
    according to the official assessment by the CASP organizers.

    GalaxyTBM: Automated protein structure prediction

    Our stand-alone tool for template-based modeling is called "GalaxyTBM" [2]
    (web service available at GalaxyWeb, click here).
    GalaxyTBM combines a number of modules in the GALAXY package such as GalaxyCassiopeia and
    GalaxyRefine and external programs for template detection and sequence alignment.

    One of the merits of GalaxyTBM is that it automatically refines unreliable model regions in
    relatively short computation time.
    Unreliable loops, termini, or side-chains are automatically
    detected and reconstructed by Galaxy modules. The generated model structures are not only
    globally accurate but also locally realistic. We keep improving the components of GalaxyTBM,
    both in the performance and speed, to provide better services to users.

    Protein Loop Modeling

    Protein loop modeling is one of the areas of our specialty. Loop modeling is applicable to the
    studies related to protein flexibility. All our loop modeling methods have grounds on the loop
    closure algorithm called "triaxial loop closure" [3,4]. For example, an ensemble of loop conformations
    can be generated by "FALC" [5,6] that combines the loop closure and the local structure sampling
    technique of fragment assembly (web service available at FALC web server,
    click here).
    In the more advanced loop modeling method "GalaxyPS" [7,8,9]
    (web service available at GalaxyWeb, (click here),
    loop closure is used as an elementary move set during global energy optimization.

    In contrast to traditional loop modeling studies that focus primarily on reconstructing loop
    structures in the experimentally resolved structures, we are extending "GalaxyPS" loop modeling to a
    broader range of problems.
    For example, even in perturbed crystal structures or structures in different
    functional state, GalaxyPS produces accurate predictions in the atomic level [9,10]. Therefore, this
    approach is expected to be more useful for solving more practical problems.

    Loop modeling method can be further applied to improving template-based models.
    A template-based model refinement tool called "GalaxyRefine" was developed based on GalaxyPS.
    GalaxyRefine has been applied to CASP experiments since CASP9, and the model quality improvement by
    GalaxyRefine has been demonstrated in CASP9 TBM and refinement categories [8].

III. Protein-Ligand Docking

    Another essential computational tool necessary for understanding protein function is protein-ligand
    docking. A protein-ligand docking method predicts the bound pose of the ligand and the binding affinity.
    Protein-ligand docking is one of the essential tools for computer-aided drug discovery.

    Our recent research in this area focuses on incorporating receptor (protein) flexibility to our
    docking program "GalaxyDock"
    [11,12]. A large number of proteins undergo conformational transitions to
    perform their biochemical functions. Nevertheless, many ligand-docking tools that treat proteins as
    rigid are used widely because of their efficiency and simplicity. A more practical docking program
    (with less computational cost and more accurate results) that considers receptor flexibility will bring
    a large impact to the field.

    A related application method developed in our lab is the ligand binding site prediction method
    "GalaxySite" [13]. A key idea is to combine molecular docking and information from similar proteins.
    Molecular docking allows more precise prediction compared to other similarity-based methods.
    This method is accessible at GalaxyWeb (click here).

IV. Protein-Protein Docking

    We are also developing prediction methods for protein-protein interactions. Prediction of protein-
    protein complex structure and binding affinity is crucial for understanding and regulating protein
    functions. Our protein-protein docking tool "GalaxyPPDock" is being tested in CAPRI [14], a community-
    wide protein-protein interaction prediction experiment (Critical Assessment of PRedicted Interactions).

    Understanding interaction between symmetric protein chains is a topic related to protein-protein
    docking. We developed a prediction method for homo-oligomer structure called "GalaxyGemini" [15].
    GalaxyGemini makes use of known quaternary structures of template oligomers,
    selected based on sequence similarity and binding interface similarity to the target protein.
    Web service for GalaxyGemini is available at GalaxyWeb (click here).

V. Applications

    The next stage in the GALAXY development is to develop more tools for practical applications.
    The first type of application methods combine incomplete experimental data obtained by structural
    biologists (which may not be very useful as themselves) and structure prediction methods
    (which may not be accurate enough as themselves) to produce high-accuracy protein model structures.

    For example, structure factors collected by X-ray crystallographers can be used to guide structure
    prediction even when phase information is not available and homology models are not accurate enough for
    molecular replacement. In addition, chemical shift data collected from NMR spectroscopy can assist
    refinement of template-based models. We are studying how to interpret and combine various experimental
    data with GalaxyTBM.

    The second type of application methods aim at obtaining insights into conformational changes and
    molecular interactions involved in proteins of pharmaceutical interests such as antibody proteins and
    G protein-coupled receptors. Development of methods specialized for those targets by applying the current
    loop modeling and ligand docking techniques are undergoing, for example, modeling of intra- and extracellular
    loops of G protein-coupled receptors and modeling of antibody loop structures in the complementarity
    determining region.


[1] J. Ko, H. Park, L. Heo, and C. Seok*, "GalaxyWEB server for protein structure prediction and refinement", Nucleic Acids Res. 40 (W1), W294-W297 (2012).
[2] J. Ko, H. Park, and C. Seok*, "GalaxyTBM", BMC Bioinformatics, 13, 198 (2012).
[3] E. A. Coutsias, C. Seok*, M. P. Jacobson, and K. A. Dill, "A Kinematic View of Loop Closure", J. Comput. Chem. 25, 510 (2004).
[4] E. A. Coutsias, C. Seok, M. J. Wester, and K. A. Dill, "Resultants and Loop Closure", Int. J. Quantum Chem. 106, 176 (2006).
[5] J. Lee*, D. Lee, H. Park, E. A. Coutsias, and C. Seok*, "Protein loop modeling by using fragment assembly and analytical loop closure", Proteins, 78, 3428-3436 (2010).
[6] J. Ko, D. Lee, H. Park, E. A. Coutsias, J. Lee*, and C. Seok*, "The FALC-Loop web server for protein loop modeling", Nucleic Acids Res. 39, W210-W214 (2011).
[7] H. Park, J. Ko, K. Joo, J. Lee, C. Seok*, and J. Lee*, "Refinement of protein termini in template-based modeling using conformational space annealing", Proteins, 79, 2725-2734 (2011).
[8] H. Park and C. Seok*, "Refinement of unreliable local regions in template-based protein models", Proteins, 80, 1974-1986 (2012).
[9] H. Park, G. R. Lee, and C. Seok*, in preparation.
[10] G. R. Lee, W. Shin, H. Park, S. Shin, and C. Seok*, "Conformational sampling of flexible ligand-binding protein loops", Bull. Korean Chem. Soc. 33 (3), 770-774 (2012).
[11] W. Shin, L. Heo, J. Lee, J. Ko, C. Seok*, and J. Lee*, "LigDockCSA: protein-ligand docking using conformational space annealing", J. Comput. Chem. 32 (15), 3226-3232, (2011).
[12] W. Shin and C. Seok*, "GalaxyDock: Protein-ligand docking with flexible protein side-chains", submitted.
[13] H. Lim, W. Shin, and C. Seok*, "GalaxySite: Ligand binding site prediction using molecular docking", submitted.
[14] Sarel J Fleishman et al. (H. Park, J. Ko, H. Lee, C. Seok), "Community-wide assessment of protein-interface modeling suggests improvements to design methodology", J. Mol. Biol. 414 (2), 289-302 (2011).
[15] H. Lee, H. Park, J. Ko, and C. Seok*, "GalaxyGemini: a web server for protein oligomer structure prediction based on similarity", submitted.


Previous research pages

2007 2009