Research

Our research experience is in the field of computational methodology for molecular systems, with an emphasis on algorithmic developments for studying the structure and dynamics of proteins using computational methods in conjunction with data from experiments, e.g. by NMR spectroscopy. Today most NMR protein structures are solved using algorithms and software that have been developed by the group and its present and former coworkers.

Algorithmic developments for computer-aided chemistry/structural biology

• Algorithms for the calculation of three-dimensional (3D) protein/DNA/RNA structures on the basis of geometric restraints, e.g. from NMR data. These programs, DIANA, DYANA and CYANA, have been or are being used to calculate thousands of protein structures in academic and industrial research.

• An efficient algorithm for molecular dynamics simulation in torsion angle space (Güntert et al. 1997, Güntert et al. 2001) that enables efficient protein structure calculation by torsion angle dynamics driven simulated annealing.

• Automatic adaption of MD time steps based on the accuracy of energy conservation (Güntert et al. 1997).

• An optimized torsion angle dynamics algorithm, implemented in CYANA, that runs a typical structure calculation about an order of magnitude faster than with other programs for the same purpose.

• Algorithms for the structure calculation of proteins that occur as symmetric multimers or fibrils.

• Two programs OPAL and OPALp for energy refinement and classical molecular dynamics simulation that made/make efficient use of vector and parallel high-performance computers, respectively.

• An algorithm for torsion angle space regularization of structures with distorted geometry, which serves also for obtaining average molecular structures in torsion angle space.

• Automated determination of optimal residue ranges for the superposition of protein structures.

Computational methods development for the analysis of experimental data

• A computational method for fully automated structure determination of proteins in solution (FLYA) that yields, without human intervention, 3D protein structures starting from a set of multidimensional NMR spectra. The FLYA method substitutes all manual spectra analysis, and can thus overcome a main efficiency limitation of the NMR method for protein structure determination (Scott et al. 2006, López-Méndez et al. 2006, Takeda et al. 2007, Güntert 2009).

• A fully automated method for NMR chemical shift assignment and protein structure determination using exclusively NOESY spectra (Ikeya et al. 2011)

• Automation of the assignment of cross peaks in NOESY-type NMR spectra, the principle source of conformational data for NMR protein structure determination (Herrmann et al. 2002, Herrmann et al. 2002, Jee et al. 2003, Güntert 2003, Güntert 2004, Güntert 2009). Automated NOESY assignment replaces the traditionally most time-consuming analysis step in an NMR structure determination by an efficient and objective computational method and plays a crucial role in NMR-based structural genomics/proteomics.

• The software package PROSA for NMR data processing (Güntert et al. 1992), and methods for automatic baseline correction (FLATT, IFLAT) and the measurements of scalar couplings (INFIT).

• The first generally applicable algorithm for automated NMR chemical shift assignment (Bartels et al. 1996, Bartels et al. 1997), a crucial and difficult step of NMR spectrum interpretation. The algorithm, GARANT, can use any combination of spectra as input data and determines assignments by searching for an optimal matching between expected and observed peaks using a genetic algorithm combined with local optimization.

• Widely used computational tools for interactive and semi-automated NMR spectrum analysis (EASY, XEASY, KUJIRA).

• An analytic implementation of the spin-1/2 product operator formalism for the simulation of NMR pulse sequences using the symbolic computing package Mathematica (POMA).

• Computational methods for determining stereospecific assignments in NMR (HABAS, GLOMSA, FOUND).

• Methods for the interpretation of experimental data in dynamic macromolecular systems as ensemble-averaged restraints.

Programming language development

For the algorithm development it is crucial to rely on a basic software package (e.g. the program CYANA) that was developed completely “in-house”. Only in this case can one fully understand its design and develop new algorithms efficiently and with minimal overhead for scientifically uninteresting issues such as the user interface, input and output, and data storage. To facilitate this, we have defined and implemented and interpreted programming language, INCLAN, that integrates seamlessly into Fortran/C/C++-based programs (Güntert et al. 1992, Güntert et al. 1997). INCLAN provides control structures, parallelization (by MPI or shared-memory), the combination of arithmetic expressions with shell-like variables, centralized syntax checking of commands, graphics, etc. The CYANA, DYANA, PROSA, GARANT, OPALp and OPAL software packages employ INCLAN for their user interface and as a scripting language. Extensive functionality of these programs is implemented in INCLAN rather than in the underlying Fortran/C/C++ source code, which renders these programs highly adaptable also for sophisticated schedules, and makes them easily usable for other researchers.

Software development

The abovementioned algorithmic developments have been implemented in a number of softwares. Many of these are widely used in academic and industrial research.

Experimental techniques

Computational methods have also been important in the development of new experimental techniques that allow the study of biomolecular systems that have so far been impossible or difficult to analyze:

• Optimal stereo- and regiospecific isotope labeling of amino acids (stereo-array isotope labeling; SAIL), which yielded the NMR solution structure of the 41 kDa maltodextrin-binding protein MBP. This is the first solution structure determination of a protein larger than 30 kDa using detailed experimental information on the side-chains (Kainosho et al. 2006).

• Structure determination of proteins in living cells (Sakakibara et al. 2009, Ikeya et al. 2010).

• Transmembrane segment enhanced labeling for the backbone assignment of helical membrane proteins (Reckel et al. 2008)

Applications

The aforementioned computational methods and software packages have been applied to a variety of biomolecular systems. Some examples include:

• NMR solution structures of more than 40 proteins, including those listed below, and others.

• Participation in the “Protein 3000 Project”, the Japanese National Project on Protein Structural and Functional Analyses that has determined 3000 new protein structures. More than half of them were solved using the CYANA software (e.g., Pantoja-Uceda et al. 2004, López-Méndez et al. 2004, Nameki et al. 2004, Scott et al. 2004, Pantoja-Uceda et al. 2005, Nameki et al. 2005, Scott et al. 2005, Li et al. 2005, Hamada et al. 2006, Kuwasako et al. 2006, Ohnishi et al. 2007, Kobayashi et al. 2007, Kuwasako et al. 2008, Kuwasako et al. 2008, Nagata et al. 2008, Lin et al. 2008, Ohnishi et al. 2009, He et al. 2009, He et al. 2009, Tsuda et al. 2009, He et al. 2009, He et al. 2010, Handa et al. 2010, Yamashita et al. 2011, Tsuda et al. 2011).

• Structure determination of integral membrane proteins: OmpX (Fernández et al. 2004), the C-terminal catalytic fragment of presenilin 1 (Sobhanifar et al. 2010), and bacteriorhodopsin (unpublished).

• Structures of various mammalian prion proteins that are involved in transmissible spongiform encephalopathies such as BSE or the human Creutzfeldt-Jakob disease (Calzolai et al. 2000, Zahn et al. 2003, Calzolai et al. 2005, Lysek et al. 2005).

• Atomic resolution structure of the fungal prion protein FgHET-s(218-289) in amyloid fibrils determined by solid state NMR (unpublished).

• Structural basis for the selectivity of the external thioesterase of the surfactin non-ribosomal peptide synthetase (Koglin et al. 2008).

• Structural basis for specific protein-RNA recognition (Kuwasako et al. 2008, Kuwasako et al. 2008, Nagata et al. 2008, Ohnishi et al. 2009, He et al. 2009, Tsuda et al. 2009, Yamashita et al. 2011, Tsuda et al. 2011)

• Conformational stability and activity of p73 require a second helix in the tetramerization domain (Coutandin et al. 2009)

• Elucidation of an intramolecular regulation mechanism for pheromone binding and release by the Bombyx mori pheromone-binding protein (Horst et al. 2001, Lee et al. 2002).

• Structures of the Na,K-ATPase nucleotide-binding domain free and in complex with ATP (Hilge et al. 2003)

• NMR structures of the periplasmic chaperones CcmE (Enggist et al. 2002) and FimC (Pellecchia et al. 1998, Pellecchia et al. 1998).

• NMR solution structures of the calreticulin P-domain (Ellgaard et al. 2001, Ellgaard et al. 2002).

• Analysis of the role of water molecules in specific protein-DNA recognition in one of the earliest nanosecond molecular dynamics simulations of a protein-DNA complex (Billeter et al. 1996).

• NMR studies of two Alzheimer peptides with widely different plaque-competence, Aβ(1–40)ox and Aβ(1–40)ox showed that there are no significant conformational differences between their monomeric forms in water (Riek et al. 2001)

• Structural studies of metallothioneins, proteins containing Zn or Cd clusters (Riek et al. 1999, Peroza et al. 2009)

• Structure-guided fragment-based in silico drug design of dengue protease inhibitors (Knehans et al. 2011)