Functional sites in proteins: A combined proteomic and bioinformatical approach for improvement of a prediction system for linear motifs.

Functional sites in proteins are short stretches of amino acids that serve as binding sites, post-translational modification sites or proteolytic cleavage sites. Most functional features of proteins can be related to functional sites and/or globular domai ns. While there are several efficient methods available for the identification of globular domains, functional sites are notoriously difficult to predict with conventional bioinformatical methods. The primary reason is that such sites are short and often poorly conserved. We and our partners of the ELM consortium, have therefore developed a bioinformatical resource which combines sequence information with molecular and cellular context information to achieve better predictions of functional sites. In this project, we will improve this resource by a proteomic approach to explore the basis for the molecular and cellular context for functional sites using sumoylation as model. We will investigate the permissive sequences for sumoylation by mutagenesis. The e xperimentally derived data will be used as basis for development of new components for the ELM resource. The new instance data, and publicly available data will be used to derive more accurate detection models for the ELM prediction engine. We will aslo e valuate if Hidden Markov Models or position-specific scoring matrices can give better predictions that the currently used regular expressions. While the project will give new knowledge and better prediction of the two types of functional sites studied her e, we anticipate that this knowledge will be applicable to functional sites in general. The impact of an improved ELM resource can not be underestimated: Since each functional site serves a particular role in the overall function of each of the target pro teins, many of them may be considered as potential drug targets. Functional site-interactions will also take a prominent place in proteomic network modelling.