Professor Burkowski is primarily interested in the mathematical models and computational algorithms that are related to structural bioinformatics. In biology, modeling the structure of a large molecule is very important if one is to understand its behaviour. Structure models have been utilized in the analysis of DNA, RNA, and proteins. In particular, proteins have a very complicated structure that allows them to interact with each other and with smaller molecules. The magic that is life takes place in the cavities of macromolecular surfaces. The primary area of research, for Dr. Burkowski, involves the geometric modeling of protein binding sites. A protein is a long chain-like molecule, typically involving thousands of atoms, and folded into a compact 'low energy' structure. Living cells contain tens of thousands of different proteins that provide both structure and control of life processes. For example, enzymes help regulate chemical processes dealing with metabolism and energy utilization. Although most proteins have a very complicated structure, a biologist will often simplify analysis of protein function by considering the surface of the protein molecule since that is the part of the protein that chemically interacts with other molecules such as DNA, drugs, or other proteins. The analysis of the interaction of proteins with a small molecule, such as a drug, is especially important, not only because of the therapeutic benefits, but also because it gives researchers a start in understanding how proteins function in more complex interactions with larger molecules.
Small molecules such as drugs will bind to a protein surface over a small area called the receptor site. By binding to the receptor, drugs either enhance or inhibit the activity of the protein. To design a drug with a high medical value, it is important to know how the drug binds to the protein. It is also important to establish the selectivity of the drug. It must bind to the receptor of a particular protein and not react with any other type of protein since such interactions may produce unfortunate side effects for the cell.
For the last four years Professor Burkowski and some of his graduate students have been developing computational tools such as SVMs (Support Vector Machines) to classify drugs in their ability to bind to such receptor sites. SVMs depend on the formulation of mathematical function called a kernel. Most people using SVMs employ standard, well studied kernels that have provided fairly reliable results over most application areas. The main thrust of our research is the customization of a kernel, primarily through so-called kernel alignment techniques, with the objective of increasing the predictive accuracy of these classifiers. The overall goal is the development of rapid filtering techniques which can be applied to databases of compounds to select molecules that can act as candidates for further testing. While such mathematical models do not replace clinical trials, they can be used to aid in the discovery and design of drug candidates while helping to eliminate drugs that have poor specificity. Dr. Burkowski has just completed the writing of an undergraduate text for students studying structural bioinformatics. It was first published in the fall of 2008.
Degrees and awards
BSc, MMath, PhD (Waterloo)
Burkowski, F.J. and Wong, W.W.L, Predicting multiple binding modes using a kernel method based on a Vector Space Model Molecular Descriptor, Int. J. Computational Biology and Drug Design (IJCBDD), Vol. 2, No. 1, 2009. pp. 58-80.
Wong, W.W.L. and Burkowski, F.J., A constructive approach for discovering new drug leads: Using a kernel methodology for the inverse-QSAR problem, J. Cheminformatics, 2009, 1:4, 1-27
F.J. Burkowski. Algorithms for Structural Bioinformatics. Chapman & Hall/CRC, Boca Raton, 2008.
F.J. Burkowski. Optimization via Gene Expression Algorithms. Chapter 8 in Handbook of Bioinspired Algorithms and Applications, Eds.: S. Olariu and A.Y. Zomaya. Chapman & Hall/CRC, Boca Raton, pp. 121-134, 2006.