The Systems Immunology Laboratory uses computer modeling to address biological questions that cannot easily be understood by experiment.

We collaborate closely with a number of research groups at IFReC and also develop software tools for general use.
Computer Servers
Our Research
Docking with experimental restraints
One of the most important general features of proteins is that they interact with other macromolecules. New protein-protein, protein-DNA, and protein-RNA interactions are continuously being revealed by novel experimental techniques. However, since theses interactions are often short-lived and/or dynamic in nature, we often cannot visualize them at atomic resolution using traditional structural biology techniques, such as x-ray crystallography or NMR spectroscopy. On the other hand, we can often obtain important clues about the location of binding sites through site-directed mutagenesis. The aim of this project is to support such wet experiments with in silico modeling in order to determine the structure and dynamics of protein-protein and protein-nucleotide complexes.

Dynamic models of immune responses
One of the long-term goals of our lab is to establish a dynamical model of the immune system that reproduces or even predicts the behavior of immune cells under various stimuli. Towards this goal, we have developed a coarse-grained formulation for modeling the dynamic behavior of cells, based on stochasticity and heterogeneity, rather than on biochemical reactions. Our formulation allows us to build a model of a cell population without requiring very precise biochemical parameters, but still provides continuous time course predictions of each molecular state as biochemical reaction equations do. As a proof of concept, we have simulated the TNF-NFkB system using our formulation1.

Regulation of gene expression
We also have a number of ongoing projects aimed at understanding regulation of gene expression. In one study, we developed a new measure for predicting combinatorial regulation by pairs of transcription factors from promoter sequence motifs. A second study focuses on the prediction of sequence motifs at specific distances from transcription start sites of genes with similar expression profiles. Also, in a collaborative study with the Laboratory of Host Defense, we are analyzing the changes of various epigenetic markers and their effects on gene expression under TLR stimulation.

Structure-based protein function prediction
In addition to such systems-level studies, our group uses structure to infer the functions of proteins of interest. Structure-based protein function prediction is a complex problem that requires a collection of software tools, which must be integrated into a serviceable pipeline. For this purpose we have developed programs for predicting multiple sequence alignments (MAFFT), pairwise structural alignments (ASH), 3D structural modeling (Spanner), side-chain and loop structure prediction (OSCAR), and functional matching (SeSAW). We have successfully used these tools to predict the function of Regnase-12, and to predict the effects of point mutations in the kinase ROP16 from Toxiplasma gondii3 and in LMP7, a component of the human immunoproteasome4. In our latest pipeline, we have included methods for the prediction of local intrinsic disorder with which we carry out composition-based function prediction using our tool IDD Navigator5. Since structure prediction is essential our function prediction work, we have developed an empirical atomic-level energy force field (OSCAR). OSCAR energies were derived from series expansions and the parameters were trained with all known protein structures. The energy functions have been shown to be more accurate than other types of energy functions for protein side chain modeling and loop selection6. We also developed a fast side-chain modeling program (OSCAR-star) based on a rigid side chain rotamer model7. We have similarly derived protein backbone potentials as Fourier series. These were then used for prediction of loop conformations (OSCAR-loop) and the accuracy was found to be better than that of other loop modeling algorithms for short loops (<10 residues). For longer loops, the prediction accuracy was improved by concurrently sampling with Spanner8.

1Teraguchi S, et al., Phys Rev E 2011, 84:062903  2Matsushita, et al., Nature 2009, 458:1185-1190  3Yamamoto, et al., J Exp Med 2009, 206:2747-2760  4Kitamura, et al., J Clin Invest 2011, 121:4150-4160  5Patil A, Teraguchi S, et al., Pac Symp Biocomput 2011, Teraguchi S, Patil A, et al.BMC Bioinformatics 2010, 11:S7  6Liang S, et al., 2011, 32:1680-1686; Liang S, et al., Proteins 2011, 79:2260-2267  7Liang S, et al., Bioinformatics 2011, 27:2913-2914  8Liang S, et al., J Chem Theory Comput 2012, in press 
Future Endeavors
In the future we will continue to pursue the two complementary missions of collaboration and software development. Of particular interest are collaborations that fuse the systems biology and structural approaches. For example, we, along with Prof. Inagakis group in Hokkaido University and the laboratory of Host Defense at IFReC, are actively engaged in constructing three-dimensional models of signaling complexes formed downstream from activated TLR receptors. In addition, the same three labs are investigating the RNA structural specificity of Regnase-1 though a combination of modeling, NMR, and site-directed mutagenesis. Finally, we are pursuing the mechanism of store operated calcium sensors in immune cells by modeling the structural dynamics of the protein STIM1, in collaboration with Assoc. Prof. Yoshihiro Baba of the Lymphocyte Differentiation Lab.