Homology relationship based on the amino acid properties

VERONICA ALVAREZ; DIAMBRA, L.

Montevideo

Conferencia; ISCB LATIN AMERICA 2010; 2010

International Society for Computational Biology

Several proteins that have diverged beyond significant sequence similarity still retain the 3D foldof their ancestor. These distant homologies were determined only after the tertiary structureshave been solved. There is a need of alternative method based on the primary structures for thedetection of similarity on distantly related proteins.Instead of use the evolutionary information to uncover subtle sequence similarity, as the methodsbased on the substitution matrix [1] and pro¯le hidden Markov model [2], we explore thealternative of evaluate similarity between two proteins by looking at the physical-chemicalproperties of their amino acid sequences. We have applied singular value decomposition (SVD)technique over a set of properties of the amino acid sequence associated to a family of proteins, in order to extract the essential features underlying to the family.We illustrate our approach by constructing one matrix of 1240 aligned homeodomains for eachproperty, and then we apply SVD over all matrix in order to obtain a set of linear independentpatterns, named eigenproperties". The eigenproperties capture the essential features associatedto this set of proteins. When we project the properties associated to a given sequence over thesubspace spanned by the eigenproperties we observe that large projection corresponds to ahomeodomain protein, while other amino acid sequences have small projection. This resultssuggest that amino acid properties could be used as an alternative tool to define similarity of twoproteins.