Calander Home 2001-2002 2002-2003 Dr. Yiannis Kaznessis Protein fold recognition based on detection of non-local interactions in proteins Date: February 17 Time: Noon to 1 Place: BSBE 4-101 Proteins are the molecular cornerstones of life. The determination of their structure is deemed as a key step toward understanding the behavior of biomolecules and initiating rational approaches for engineering molecular solutions. Experimental efforts, such as x-ray crystallography and NMR techniques are not currently efficient enough to allow for rapid structural determination of the ever-increasing number of newly discovered sequences. Hence, computational methodologies are becoming the sine qua non of protein sequence/structure/function relationship research. A prominent role in computational tools has been claimed by protein fold recognition methods. As protein fold we define the arrangement of secondary structural elements in 3D space and their topological connections. In this work we have focused on the development of an algorithm that incorporates non-local interactions between residues, significantly improving existing protein fold recognition approaches. We have designed pattern recognition techniques that assign scores/probabilities that a target protein sequence belongs to a particular fold family. The feature vectors for pattern recognition are an amalgamate of secondary structural elements and intrachain distance constraints, exploiting the importance of non-local interactions in the foldability, stability, and functionality of protein molecules, as manifested in observed mutation patterns in classes of homologous proteins.