Personal tools
You are here: Home Research
Document Actions

Research Overview

Our current research focusses on four main areas, for which we develop and apply techniques grounded in computational biophysical chemistry, molecular bioinformatics, and computational pharmaceutical chemistry.

I. Structure, Function, and Molecular Recognition Investigated by Molecular Dynamics Simulations

II. Rigidity Theory for Biomolecules

III. Structure Prediciton by Comparative Modeling and Using (Co-)Evolutionary Information IV. Modulating Biomolecular Interactions

A substantial proportion of the research carried out in our group is based on unbiased and biased molecular dynamics simulations at the atomistic level. To translate the molecular motion obtained from these simulations to meaningful insights into protein structural dynamics and function, we make use of different levels of free energy calculations as post-processing techniques: We perform binding free energy calculations via the Molecular Mechanics Poisson-Boltzmann Surface Area (MM-PBSA) [81] and Molecular Mechanics Generalized Born Surface Area (MM-GBSA) [81,108] approaches, computational alanine scanning of protein-protein interfaces using the DrugScorePPI approach [164], umbrella sampling and potential of mean force (PMF) calculations using the Weighted Histogram Analysis Method (WHAM) [145,152], as well as thermodynamic integration (TI) computations [108]. We closely collaborate with researchers from a variety of experimental disciplines to refine and to corroborate our computational results, and to explain experimental observations on the atomistic level.

Soluble Proteins

In a collaborative approach, we predicted structural intermediates along the reaction cycle of Pyruvate Phosphate Dikinase (PPDK) [145] by computing configurational free energies and confirmed the existence of these intermediates in a series of crystal structures. Employing a similar methodology, we later provided evidence for an alternate binding change mechanism in PPDK, which provided the first explanation, why this enzyme is only active as a dimer [152]. As to further applications on soluble proteins, MD simulations also proved an indispensable tool in the systematic identification and characterization of peptidic Hsp90 dimerization inhibitors [127].

Soluble Proteins

Alternate binding change mechanism in PPDK.

Membrane Proteins

Simulation system containing the ABC transporter MDR3 embedded in a 1,2-dioleoyl-sn-glycero-3-phosphocholine bilayer.

Membrane Proteins

Over the course of the last years, simulations of membrane proteins have become a routine application in our group [162,160,153,151,138]. Our studies on the G protein-coupled receptor (GPCR) TGR5 are another example of how simulations and experiment can fruitfully complement each other: We used MD simulations to narrow down possible multimerization patterns of that receptor [160,138], and to determine that the C-terminus of this GPCR controls membrane trafficking predominantly via its secondary structure [103]. Assisted by FRET measurements, we also provided an explanation on the atomistic level for the prothrombotic phenotype of platelets that carry the HPA-1b polymorphism in αIIbβ3 integrins [162]. We currently perform explicit solvent, explicit membrane MD simulations of the potassium/sodium hyperpolarization-activated cyclic nucleotide-gated ion channel 2 (HCN2) to gain insights into how binding of cyclic nucleotides regulates channel conductivity on a structural level.


The ribosome is one of the major targets for antimicrobial substances, many of which exert their inhibitory effect by binding at the peptidyl transferase center or the ribosomal exit tunnel. We combined MD simulations with free energy calculations [98,119] to study the species selectivity of oxazolidinone antibiotics and to explain how a well-described double mutation in the large ribosomal subunit affects oxazolidinone binding. As to other RNA-containing systems, MD simulations provided insights why ligand binding to the guanine sensing riboswitch aptamer domain (Gsw) is aggravated in the absence of Mg2+ ions, and to detect that the Gsw loop region is dynamically coupled to the ligand binding site [148]. To be able to accurately predict RNA-ligand interactions, we developed a knowledge-based scoring function [30], DrugScoreRNA, which was based on the formalism of the DrugScore approach [4,12].


Atomic mobility of the aptamer domain of the guanine sensing riboswitch.

Binding Free Energy and Entropy Calculations

Schematic of binding free energy calculations.

Binding Free Energy and Entropy Calculations

To facilitate the setup of binding free energy calculations according to the MM-PBSA, MM-GBSA and linear interaction energy (LIE) approaches for sets of ligands typically investigated in lead optimization, we developed the workflow program FEW and extended it to membrane systems [85,114]. By conducting such binding free energy calculations, we aim at identifying procedures and strategies that allow to correctly rank potential drug candidates according to their binding affinity. In this context, correctly and efficiently estimating translational, rotational, and configurational entropy contributions to the binding free energy is still a holy grail [18]. To this end, we developed an efficient tool to approximate configurational entropy changes in terms of the reduction in translational and rotational freedom of the ligand upon protein-ligand binding [141].

Modeling biomolecular flexibility and plasticity is important for understanding the function of biological systems. Hence, it is desirable to have a precise knowledge about what can move and how. Given a network representation of a biomolecular structure, its intrinsic flexibility can be efficiently identified by determining the number and spatial distribution of bond-rotational degrees of freedom in the network. The flexibility for a given biomolecule can be analyzed within a few seconds by a graph theory-based approach as implemented in the FIRST program by Thorpe et al. We are using rigidity analysis for investigating large biomolecules such as the ribosome, analyzing the structural determinants of thermostability, approximating the change of vibrational entropy upon binding of biomolecules, understanding allosteric transmission in biomolecules, and sampling of biomolecular conformational spaces. Detailed information about rigidity analysis and its application to biomolecules is provided in [144].

Constraint Network Analysis

The Constraint Network Analysis (CNA) approach [93] aims at linking information from rigidity analysis with biomolecular structure, (thermo-)stability, and function. CNA functions as a front- and back-end to FIRST, for which the C++-based CNA interface module pyFIRST was developed. That way, the computational efficiency of FIRST is preserved in CNA-driven computations [93]. Going beyond the mere identification of flexible and rigid regions in a biomolecule, CNA allows for I) performing constraint dilution simulations that consider a temperature dependence of non-covalent interactions [67], II) computing a comprehensive set of global and local indices for quantifying biomolecular stability [86], and III) performing rigidity analysis on ensembles of network topologies (ENT). For the latter, structural ensembles [19] and ensembles based on the concept of fuzzy non-covalent constraints (ENTFNC) can be used [96]. In order to facilitate the processing of the highly information-rich results obtained from CNA, the VisualCNA plugin [116] for PyMOL and the CNA web server [94] have been developed.

Constraint Network Analysis

Illustration of the Constraint Network Analysis (CNA) approach for characterizing biomacromolecular flexibility [93].

Protein Thermostability

Rigid cluster decomposition of Bacillus subtilis lipase A (PDB ID: 1ISP).

Protein Thermostability

Monitoring the decay of network rigidity along a constraint dilution trajectory helps to improve the understanding of the relationship between biomolecular structure, activity, and thermostability. CNA was successfully applied to a variety of tasks ranging from the comparison of proteins from mesophilic and thermophilic organisms [37,54] to series of orthologs [68,125] to variants with only a few substitution [117,129]. From these studies, we provided direct evidence for the “principle of corresponding states”, according to which mesophilic / thermophilic homologs have similar flexibility and rigidity characteristics at the respective optimal growth temperatures [54]. We also obtained good to very good correlations between predicted and experimental thermostability values [67,117] and emphasized the importance of interface stability that contributes to the thermostability in multimeric structures [125]. For prospective studies, we developed a strategy to predict amino acid substitutions optimal for thermostability improvement [129]. The strategy combines a structural ensemble-based weak spot prediction for the wild type protein by CNA, filtering of weak spots according to sequence conservation, computational site saturation mutagenesis, assessment of variant structures with respect to their structural quality, and screening of the variants for increased structural rigidity by ENTFNC-based CNA.

Allosteric Transmission

Allostery is the process by which biomolecules transmit the effect of binding at one site to another, often distal, functional site. Due to the non-local character of rigidity percolation, rigidity analysis provides insights how altered stability due to binding of an allosteric effector affects sites all across the network. Such a long-range effect was first demonstrated for the protein-protein complex Ras/Raf [19]. Inspired by this observation, a computationally highly efficient approximation of changes in the vibrational entropy (ΔSvib) upon binding to biomolecules has been introduced recently, based on rigidity theory [146]. Compared to state-of-the-art computational methods for computing ΔSvib, this approach yields significant and good to fair correlations for datasets of protein-protein and protein-small molecule complexes as well as in alanine scanning. Recently, an ensemble-based perturbation approach has been introduced for gaining a deeper structure-based understanding of the relationship between changes in static properties and allosteric signal transmission in biomolecules [158]. Applying a free energy perturbation approach to results of rigidity analysis, free energies of cooperativity and pathways of allosteric signaling are computed. The approach was successfully applied on biomolecules showing ligand-based K- and V-type allostery, respectively, and for computing free energies of cooperativity for binding of the allosteric and orthosteric ligands in agreement with the underlying mechanisms of negative and positive cooperativity. As to nucleic acid systems, we proposed an allosteric signal transmission pathway within the large ribosomal subunit [40], which has been confirmed by two independent experimental studies later. In another study by us, FIRST was used to investigate the interplay between the ligand binding site, tertiary loop-loop interactions, and the switching sequence in the aptamer domain of the guanine-sensing riboswitch [156]. Our findings suggest that the distant tertiary interactions and the ligand binding cooperatively stabilize the P1 region, and in this way influence the regulation of genes.


Ensemble- and rigidity theory-based perturbation approach to analyze dynamic allostery [158].

Multi-Scale Modeling of Macromlecular Conformational Transitions

Multi-scale modeling of macromolecular conformational changes using NMSim[57].

Multi-Scale Modeling of Macromlecular Conformational Transitions

Modeling conformational transitions of macromolecules is computationally challenging. Recently, coarse-grained normal mode approaches based on elastic network theory have emerged as efficient alternatives for investigating large-scale conformational changes. It has been shown that functionally important conformational changes of proteins can be described by low frequency normal modes, which are robust and insensitive to higher coarse-graining [52]. Accordingly, we have introduced a three-step approach for multi-scale modeling of macromolecular conformational changes. The first two steps are based on recent developments in rigidity and elastic network theory (termed Rigid Cluster Normal Mode Analysis) [24,26]. In the final step, the recently introduced idea of constrained geometric simulations of diffusive motions in proteins is extended. New macromolecule conformers are generated by deforming the structure along low-energy normal mode directions predicted by RCNMA plus random direction components [57]. Recently, NMSim has been used to sample large-scale domain motions during phosphate group transfer in the pyruvate phosphate dikinase (PPDK). From this, an unknown intermediate state of PPDK has been identified, which was confirmed by X-ray crystallography [145,152]. In connection with quantitative FRET studies and integrative structure modeling, NMSim has been used for unbiased and FRET-guided generation of structural ensembles [139]. NMSim is accessible via a web server [75].

Knowledge of a protein structure is essential to understand its function, evolution, dynamics, stability, interactions and for knowledge-based protein- or drug-design. Experimental structure determination rates, however, are far exceeded by that of next-generation sequencing, resulting in less than 1/1000th of the known proteins having an experimentally resolved structure. Computational structure prediction seeks to alleviate this problem.


We are building a suite of meta-programs called TopSuite for computational structure prediction in order to benefit from the strengths of different methods and counteract their weaknesses. The suite includes programs for template selection (TopThreader), sequence and structure alignment (TopAligner), template-based model building (TopBuilder), model quality assessment (TopScore), model combination and refinement (TopRefiner), ab initio residue contact prediction (TopContact), protein-protein contact prediction (TopInterface) and protein-protein docking (TopDock). We developed the pipeline TopModel that combines TopThreader, TopAligner, TopScore, TopBuilder and TopRefiner to produce protein structure predictions based on detected templates. TopModel combines and improves predictions from a wide range of primary predictors using large diverse datasets, deep neural networks, and highly accurate model quality assessment to optimize template selection, template-target alignment, model selection, and model combination and refinement. TopModel has been used in several collaborative projects [100,124,135,163] producing high quality models for a wide range of systems.


TopModel workflow.


Hybrid modeling with high precision FRET information.

Hybrid FRET Modeling

Due to the sparse experimental information provided by low resolution techniques such as FRET or EPR, determining atomic resolution protein structures using only these experiments is impossible. Fortunately, computational methods can provide complementary information, such as detailed structural features. The combination of structure prediction with experimental input in a hybrid approach can lead to generation and verification of detailed multi-domain protein structure models [139] or quarternary structures [78], because experimental data such as FRET distance information can guide the model building. The key to hybrid modelling lies in the fine interplay between the computer simulations and the experiments to accomplish the most effective synergies between the strengths of both sides.

Comparative Modeling and Binding Mode Prediction of GPCRs

G-protein coupled receptors (GPCRs) are currently among the most important drug targets, yet only a few GPCRs have been crystallized. Thus, when developing drugs for not yet crystallized GPCRs, computational structure prediction is required for knowledge-based drug design. While the number of unique X-ray crystal structures of GPCRs is steadily increasing, they must still often be modeled at low sequence identities. This necessitates the use of as much evolutionary and mutational information as possible as well as a multi-template approach. As such, the sequences of the family of a specific GPCR can be used to guide the alignment of the templates to the target sequence as is done in the TopModel pipeline. For binding mode predictions in models of GPCRs, we perform molecular docking to a variety of models to include the flexibility of the binding pocket. When complemented by mutational analysis and molecular dynamics simulations in an iterative, integrated modeling approach, as done with TGR5 [120], this can result in a very detailed, predictive binding mode model.

Comparative Modeling and Binding Mode Prediction of GPCRs

Binding mode model of taurolithocholic acid in TGR5 with mutationally verified interacting residues.

Protein-ligand and protein-protein interactions play an important role for the function of living organisms. Thus, one of the research areas of our group is focused on analyzing regions of binding interactions. In particular, we apply our computational tools and expertise to investigate ligand binding sites and protein-protein interfaces to understand the interaction of binding partners and to identify possibilities to modulate such interactions.

Analysis of Binding Epitopes

A long-standing development in our groups is the characterization of binding pockets using knowledge-based potentials [4,23]. The family of DrugScore pair-potentials has been successfully used to describe binding sites between small molecules and proteins [14,31], small molecules and RNA [30] and protein-protein interactions [104,107]. Currently, we are seeking to update and improve these knowledge-based functions by exploiting the ever-growing Protein Data Bank. Furthermore, we develop and utilize state-of-the-art modelling and simulations techniques, including enhanced sampling methods and free energy calculations, to understand the conformational interplay between binding partners as well as the associated energetics [18,26,73,81,85,157]. Our investigation of the per-residue decomposition of binding energetics represents a methodological hallmark [16,98]. We actively extend these approaches towards an application to membrane-bound systems [114]. The integration of such simulations with experimental data allows the elucidation of interactions with atomistic accuracy that agree with in vitro time scales [139,162]. Next to analysis of known binding epitopes, we investigated possibilities to identify transient binding pockets [73]; a topic, which is currently revisited.


Structure of the aptamer domain of a guanine-sensing riboswitch (left) and computed two-dimensional potential of mean force (2D-PMF, right) [157].


Artistic representation of a teroxazole-based peptidomimetic PPII [77,101].

Key Applications

We have successfully studied a number of highly relevant targets, in close collaboration with experimentalists. RUNX1/ETO is an oncogene, required for the onset and maintenance of acute myeloid leukemia (AML). We were able to identify small molecule inhibitors that supress the RUNX1/ETO tetramerization via the NHR2 domain by mimicking α-helical structures [77] and targeting hotspot residues [97]. Our lead compound significantly reduced dissemination of leukemic cells in mice and increased the survival rate [143]. Another validated target for cancer treatment is HSP90. We modelled the dimeric structure, studied binding site hotspots, and investigated the behavior of inhibitory peptides by homology modelling and MD simulations [104,127]. ETR1 is a prototype of plant ethylene receptors, which promotes fruit ripening. The control of fruit ripening has an enormous economic potential. By combining homology modelling, free ligand diffusion MD simulations, and rigidity theory with experimental site-directed mutagenesis, we deduced the site and potential mode of action of ripening inhibitory peptides on ETR1. NsrR is a response regulator involved in resistance to the lantibiotic nisin. By using structural modelling, we shed light on the DNA-binding interface of NsrR [130]. PEPC, phosphoenolpyruvate carboxylase, plays a central role in carbon fixation in C4 plants, which use a more effective photosynthetic pathway than C3 plants. Selective inhibitors of PEPC in C4 plants are promising starting points of herbicide development. Using docking and shape-based virtual screening, we were able to identify new classes of potential C4 plant herbicides [134,154].

Dimerization Epitopes of GPCRs

An emerging target for the modulation of protein-protein interactions are dimerization interfaces of GPCRs. The homo- and heterodimerization of GPCRs has been shown to influence the membrane trafficking, signaling, and degradation of these proteins. Furthermore, agonistic and antagonistic signals can be transferred from one receptor in a heterodimer to the other. This opens up new possibilities to influence the behavior and signaling of GPCRs by targeting their dimerization epitope. In order to employ knowledge-driven approaches, however, the dimerization epitopes of the GPCR of interest have to be discerned. We unraveled the dimerization interfaces of TGR5 in live cells in an integrated modeling approach combining modeling, molecular dynamics simulations and FRET. Here, we modeled TGR5 on GPCR dimers known from X-ray crystal structures and combined them with an exhaustive conformational ensemble of an attached fluorophore and its linker. We used this to compare the expected FRET distance distributions to the ones measured in live cells and could, thus, show that TGR5 adopts a dimerization epitope involving transmembrane helix 1 and helix 8 [138].

Dimerization Epitopes of GPCRs

TGR5 dimer with fluorophore positioning weighted by a Boltzmann probability.

Powered by Plone CMS, the Open Source Content Management System