engleski

Bioinformatics classification of proteins functional families

The important problem in protein classification are functionally connected protein families without a common ancestor sequence (like coregulators of nuclear receptors) and can not be classified by multiple sequence comparison methods. The goal of this project is to construct optimal representation of sequences based on adjacency histogram, physical chemical properties of amino acids and then analyze the representations using eigenanalysis methods. The working hypothesis is that the information from sequence can be preserved in adjacency histogram with arbitrary precision. Because the histogram representation is size invariant with respect to the length of sequence and the level of information in histogram can be controlled this approach fills the space between dwo significantly different methods of protein studies: the comparison of sequences and study of protein motifs. The general result will be a stand alone methodology of protein classification with operator controlled resolution. The specific results will be classification and expansion of family of nuclear receptor coregulators. Specific goal of this project is classification of subset of nuclear receptor coregulators family (216 proteins, 1252 with all orthologs) represents significant insight into cell's regulatory process. The general goal of protein classification method with arbitrary precision resolution will result in studies of other protein families besides the one of nuclear receptor coregulators. Methodology is applicable to large scale problems and with use of approximate computational methods and parallel computesr can be applied on the whole proteome.

proteins; classification; eigenanalysis

nije evidentirano