Project 1. Exploiting Structural Genomics Information To Incorporate Protein Flexibility In Drug Design.  


Key project members:  Sander Nabuurs, Simon Folkertsma, David Wood, Gijs Schaftenaar, Markus Wagener. 


Sponsor: Biorange (NBIC)



As an in-silico complement to high-throughput screening (HTS) molecular docking and virtual ligand screening are established computational techniques to predict protein-ligand interactions. For most of these techniques protein flexibility is neglected as it is both extremely difficult to predict in a reliable manner and is also very computer-intensive. Conformational changes within the protein are one of the main factors why most computational or structural models fail to properly explain or predict molecular interactions.

These changes influence not only binding of small molecule ligands to proteins, but also the interaction between proteins themselves. Therefore, building up more knowledge on protein flexibility (or more accurately conformational plasticity) is fundamental to understanding the ways, in which drugs exert biological effects, to modulate protein-protein interactions and to achieve increased affinity between a drug and its target.


Project Goals

In this project we want to improve the prediction of ligand-induced conformational changes, in particular at the protein binding site by developing and combining new and structural bioinformatics technologies. These technologies will be applicable in all stages of drug discovery and will be validated by a/o protein X-ray crystallography and by in vitro/in vivo measurements.


Key goals of the project are:

  • to identify novel ligand binding pockets by predicting protein receptor flexibility.
  • to consider receptor flexibility in small molecule docking (virtual screening).
  • to develop new methods for large scale analysis of binding site architectures.
  • to identify 3D-bioisosteric replacements from a database of small molecule fragments extracted from protein crystal structures.
  • to predict flexibility and rigidity of miniproteins that can be used to mimic the binding surface of large natural proteins.



The project is divided into four sub-projects using different scientific approaches and strategies.


Molecular dynamics (MD) computer simulations based on hydrophobic probes and residue rotamer space search techniques are used to explore the conformational space available to proteins. The search strategy is focused to identify new, potentially transient, binding pockets that are not obvious from the starting structures. The newly identified pockets can be exploited in virtual screening campaigns.


The Fleksy approach for induced-fit docking builds on and integrates existing tools into a flexible pipeline to account for protein plasticity. In this pipeline protein flexibility is described using structure ensembles, which are subsequently used in an ensemble docking experiment. The resulting complexes are optimized using state-of-the-art refinement techniques resulting in high quality complexes, directly suited for further use in drug discovery.


Binding pockets in proteins are described using pharmacophore fingerprints. This allows for a fast similarity search and comparison of binding sites without first having to align the protein structures. Among the many applications of this technique we are first focusing on the identification of bioisostric replacements. Small molecule fragments, extracted from crystal structures, can be described by their local protein environments. Similarity searching techniques can then be used to identify sets of fragment that occur in similar binding sites and are therefore likely to be bioisosteric in nature.


We have explored the use of miniproteins as convenient scaffolds to generate androgen receptor (AR) binders incorporating the FxxLF motif (this motif is necessary for binding of the cofactor to AR). The FxxLF motif was modeled into small, stable proteins from the protein data bank (pdb) with at least one helix so that the predicted structural perturbations were minimal. After synthesis, the binding affinity of the new designed miniproteins to the AR LBD was measured by Fluorescence Polarization (FP). Synthesis and FP measurements were performed in collaboration with the Max Planck Institute of Molecular Physiology in Dortmund, Germany  (Group dr. Luc Brunsveld).


For all approaches it is of utmost importance to establish continuous feedback from experimental techniques (e.g. X-ray crystallography of relevant protein-ligand complexes, binding affinity determinations, or other measurements).


Results to date

  • Hydrophobic probe MD simulation techniques and workflows are set up; preliminary results are encouraging and allow simulation of the partial unfolding of proteins to identify alternative binding site conformations and/or allosteric binding sites.
  • The Rotacal software package is developed for searching the rotameric space of a proteins active site. A prototype of the rotamer approach has been built into the Molden simulation package and tested at several protein families.
  • The Fleksy concept for induced fit docking has been developed into a software tool. The software is actively used within the CMBI CDD group and Schering-Plough. Several binding mode predictions generated by Fleksy have later on been confirmed by both Schering-Plough in-house and public crystal structures. Furthermore, Fleksy has already proven useful in library design, virtual screening and lead optimization.
  • New algorithms for binding site comparison are developed and tested. The methods will be exploited to transfer series of compounds between drug targets and to give access to information on 3D bioisosteric groups.
  • A catalogue of kinase hinge-binding fragments has been created from an analysis of the kinase crystal structure entries of the Protein Data Bank.
  • A dataset of known 3D bioisosteric fragment pairs has been created from an analysis of all protein crystal structures of the PDB. This dataset is being used to validate and optimize the 3D-bioisoster identification tool, and to provide crystal structure examples of the bioisosteric pairs identified by Schering-Plough's 2D-bioisoster database, IBIS.( M. Wagener, J.P.M. Lommerse The quest for bioisosteric replacements. J Chem Inf Model, 2006, 46 , 677-685)
  • Application to protein-protein interaction studies. Naturally occurring miniproteins with a secondary structure stabilized by multiple disulfide bridges can be used as structural templates to mimic the binding surface of large proteins. When predicting the structural changes resulting from targeted mutations prediction and control of protein flexibility is again pivotal for success. Based on modeling and simulation techniques minproteins were designed Four of the eight initially synthesized miniproteins showed high affinity binding better than the binding observed for a reference peptide. Moreover, there is also a clear preference for the FxxLF motif, since the natural unmodified miniproteins showed no binding to AR. The best 2 binders were optimized by a second round of in silico design and mutations, which finally resulted in the synthesis of miniproteins that bind with a higher affinity to AR than the initially designed miniproteins. The results of this project were presented on a poster at various conferences and are currently subject of a communication and article in preparation ("Miniproteins as Cofactor Binding Inhibitors for the Androgen Receptor").



2003 S.J. Teague. Implications of protein flexibility for drug discovery. Nat. Rev. Drug Discov. 2, 527-541.

2006 X. Barril, X. Fradera. Incorporating protein flexibility into docking and structure-based drug design. Expert Opin.Drug.Discov., 1, 335-349.

2007 S.B. Nabuurs, M. Wagener, J. de Vlieg. A Flexible Approach to Induced Fit Docking. J.Med.Chem. 50, 6507-6518.