PROTEIN STRUCTURE DETERMINATION.
State of the art experimental and computational methods in protein structure determination

Organisers:

Ugur Sezerman, Biological Sciences and Bioengineering programme, Sabanci Univeristy, Turkey
Rita Casadio, Department of Biology, Univerisyt of Bologna, Italy
Burak Erman, Material Sciences Programme, Sabanci Univeristy, Turkey

For further information, visit the workshop website or e-mail: Ugur Sezerman (ugur@sabanciuniv.edu)

Background:

The main purpose of structural genomics is to find the three-dimensional structures of proteins and through the structure understand their function and how they interact with each other. The function of a protein is directly related to its structure. It has been shown that misfolded proteins disrupt the function of that protein. At the cellular level, protein's function depends on its interaction with other molecules. If we can obtain the structures of protein complexes, we can understand the binding mechanism of these proteins.

There are two ways to determine protein structure in structural genomics, the first one is the experimental determination of protein structures using NMR or X-ray crystallography, and the second one is the computational structure prediction.

Recent improvements in structure determination using NMR improved the resolution of the structure, but NMR can only tackle with proteins of certain size, which remains as the major drawback of this technique. On the other hand the X-ray crystallography can obtain high-resolution (less than 1 Angstrom) structures of the proteins of any size, but the major bottleneck in this method is the expression, purification and the crystallisation of the protein. As the number of sequences coming in from the several genome projects is increasing, the need for high throughput accurate structural prediction methods is imminent.

Computational methods involve ab initio methods, threading and comparative modelling. Ab initio methods can handle proteins of small size because of the extensive computational demands. Threading tries to fit the sequence of a protein into a known fold. The success of this method depends on the sensitivity of the target function and the fitting algorithm. This method fails for proteins with a novel fold. The most successful computational approach for protein structure determination is comparative modelling. Comparative modelling involves identifying the best template with a known structure from the PDB. The identification is done by PSI- BLAST. Models for the protein can be built by modelling tools available on the web such as Modeller, SwissModel and 3D-pssm. New trends in comparative modelling include using multiple templates. This improves the accuracy of the prediction. Still, if the sequence identity between the template and the target is low (less than 30%) the method does not yield accurate models.

There are several other approaches that can provide additional information to improve the accuracy of the predictions. Secondary structure prediction methods involve statistical rules derived from structures. Fold recognition algorithms use several computational methods such as neural networks, HMMs or SVMs. They can also provide information on which fold the protein is closely related to. Predictions can be improved by searching databases of complete protein domains (CATH, ProDom, SCOP), collections of structural or functional sequence motifs (BLOCKS, PRINTS), or libraries of conserved sequence patterns associated with specific functions (PROSITE). The new approaches for constructing the contact maps of proteins aid the prediction by finding out contacting residues from the sequence information. These additional constraints limit the conformational search space and thus simplify the problem.

A number of computational methods are developed to study protein-protein interactions in recent years. Different docking algorithms try to determine the bound structure of two or more proteins. A priori knowledge of the interacting proteins 3-Dimensional structures is necessary for the current docking procedures. Many docking methods are based on the rigid body approximation, other procedures use a geometric criterion on a simplified representation of the protein surface, and some others combine a shape complementarity search with subsequent energy refinement. One major drawback of the rigid docking algorithms is the conformational change of proteins upon binding. Recently developed flexible docking algorithms can dock a flexible ligand into the active site of a protein. Some recent methods use data other than protein structure to infer possible interaction networks, namely protein interaction mapping based on the analysis of gene fusion events or on comparison of genome sequences. All the methods discussed so far are the means that are used in structural genomics to obtain the structure of the proteins or protein complexes.

Venue:

The location of the workshop will be in Kemer, Antalya which is a town lying at the foot of Taurus mountains on the turquoise coast. The hotel where the meeting will be held is surrounded by pine forests, orange grooves and historical sites. The conference centre of the hotel has a large conference hall equipped with electronic facilities and Internet connections. Also a networking of 20 PC, connected to the Internet is available for demonstrations.

Dates:

November 1st - 3rd, 2002

Provisional Programme:

Attendance:

The course is limited to a maximum of 60 participants.

Registration:

Registration will be taken through the workshop website