Dissecting and predicting biomolecular complexes

A structural bioinformatics and modelling project from the Computational Structural Biology Research group

Bonvin Master Project

Contact person: Dr. A.M.J.J. Bonvin
(Bloembergen NMR building, room 1.22)


The fact that an entire genome can nowadays be sequenced for less than $10’000 led to a boom in genetic information, which in turn attracted particular attention to biomolecular interactions. It is estimated that a human cell is regulated by over 300’000 protein interactions, but only a small fraction of these have been structurally characterized by experimental methods such as x-ray crystallography or NMR. Other biochemical and biophysical methods can, however, obtain partial structural information on these interactions, while bioinformatics analysis of the can also contribute important evolutionary data. Combining these predictions and/or partial experimental information with methods for structure prediction of interactions – docking – allows the generation of atomic structural models that complement the experimental techniques.

All docking methods share three common elements: first, three-dimensional (3D) structural models of the individual components must be available; second, they must explore the conformational landscape of the interaction and generate candidate structural models of the complex, what is called sampling; finally, they must assess the generated models and select those that are more likely to be representatives of the native complex, what is called scoring.  We have developed for this purpose an information-driven docking approach called HADDOCK. It currently the best docking method in the world as assessed in a blind international competition. HADDOCK is unique because it can use external information to bias the calculations towards the ‘right’ answer. Nevertheless, there are still many challenges open related to describing larger and more complex systems, improving our scoring functions and assessing the impact of a variety of data and energy functions on the prediction performance.

Possible Projects

We are looking for motivated students to help us further develop HADDOCK. Some examples of possible projects are:

  1. Test several scoring functions and decide which improve the performance of HADDOCK.
  2. Measure the impact of the electrostatics treatment on the docking performance.
  3. Assess how to best make use of mass-spectrometry information, e.g. chemical cross-links, to guide the modelling process.
  4. Work on interface predictors for protein-RNA complexes
  5. Build coarse-grained (simplified) representations of nucleic acid molecules. This dramatically decreases the computational cost of modelling large assemblies like the ribosome and the nucleosome.

You will work independently - although with guidance - on your own project, which will involve basic computer skills (can be learned on the job), and biochemistry/biophysics knowledge of protein structures. And you will make a real-world impact in structural biology research considering the large and worldwide user base of HADDOCK!