Monday, May 2, 2005
12 Noon - 1:00 p.m.
3105 Engineering Bldg.
Computational Protocols for Functional Annotations of Proteins
Division of Biomedical Informatics, Children's Hospital Research Foundation,
3333 Burnet Avenue, Cincinnati, OH 45229, http://folding.chmcc.org
Computational protocols for protein structure and function prediction play an essential role in post-genomic efforts to transform the available genomic data into a rich source of medically relevant hypotheses. Currently used methods, while very successful in certain aspects of functional annotations, require further improvements to provide reliable predictions pertaining to such critical aspects of protein function as interactions with other proteins and membranes, post-translational modifications and effects of mutations and polymorphisms. In this talk we discuss several novel representations, algorithms and computational protocols for structural and functional annotations of proteins, including improved prediction of protein-protein interaction sites, phosphorylation sites and membrane domains in membrane channels and pores.
The underlying common denominator for these studies is the observation that accurately predicted (from the amino acid sequence) structural profiles, consisting of several important attributes of amino acid residues in a protein, such as the level of solvent exposure, secondary structures or propensities to different environments, can provide a compact and informative representation of an amino acid and its context. Consequently, such profiles can be used to yield improved predictions of other functional and structural features of proteins, such as protein-protein interaction sites, membrane domains and post-translational modifications. We are using the above principle and advanced, specifically tailored machine learning techniques, in order to develop reliable structural and functional annotation of proteins. This is joint work with Rafal Adamczak, Aleksey Porollo, Michael Wagner, Karthikeyan Swaminathan and Baoqiang Cao.
B. Cao, R. Adamczak, A. Porollo, M. Jarell and J. Meller; Prediction-based Structural Profiles Enhance Recognition of Membrane Proteins, submitted
M. Wagner, R. Adamczak, A. Porollo and J. Meller; Linear Regression Models for Solvent Accessibility Prediction in Proteins, Journal of Computational Biology, Vol. 12 (3): 355-369 (2005)
R. Adamczak, A. Porollo and J. Meller; Combining Prediction of Secondary Structures and Solvent Accessibility in Proteins, Proteins: Structure, Function and Bioinformatics, to appear (2005), Epub ahead of print
A. Porollo, R. Adamczak and J. Meller; Polyview: A Flexible Visualization Tool for Structural and Functional Annotations of Proteins, Bioinformatics, vol. 20 (15): 2460-62 (2004)
R. Adamczak, A. Porollo and J. Meller; Accurate Prediction of Solvent Accessibility Using Neural Networks Based Regression, Proteins: Structure, Function and Bioinformatics, 56(4):753-67 (2004)