PhD-defence by Glennie Helles


Predicting the tertiary structure of proteins from their primary sequence - known as ab initio protein structure prediction - is considered one of today's most challenging problems in molecular biology. It is also the main focus of this PhD-thesis where we approach the problem from two different but complementary sides.

First, we look at parallel metaheuristics and show experimentally how a specific design can improve search efficiency, such that we are able to locate lower energy structures much faster than by using any of the other preferred parallel metaheuristics. Secondly, we investigate two different ways of decreasing the solution space, namely by local and by global constraints.

Theoretically, atoms can rotate freely around the covalent bonds along the backbone of a protein, but from Ramachandran plots of proteins we know that they do not. Looking at local constraints we thus construct a neural network and use it to predict a probability distribution of dihedral angles for coil residues. This is useful because it can be used to guide search algorithms towards the most probable dihedral angle area for coil residues.

Turning our attention to global constraints, we investigate beta-topologies. Knowing the correct beta-topology would decrease the solution space tremendously, but it is unfortunately not yet possible to reliably predict the correct beta-topology of proteins.  However, we devise a way of ranking beta-topologies such that we can put a limit on the number of beta-topologies we need to explore in order to be sure that we search close to the native structure.

Assessment Committee:

Chairman: Associate Professor Kim Steenstrup Pedersen (DIKU, University of Copenhagen)
Professor Salam Al-Karadaghi (Department of Biochemistry and Structural Biology, University of Lund)

Senior Lecturer Marcus Brazil (Department of Electrical and Electronic Engineering, the University of Melbourne)

Academic supervisor:

Associate Professor, Head of Department Martin Zachariasen (DIKU, University of Copenhagen)

For an electronic copy of the thesis, please contact Jette Møller,, + 45 35 32 14 57.