Scordis, Philip; (2001) New approaches to facilitate genome analysis. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text New_approaches_to_facilitate_g.pdf Download (13MB) | Preview

Abstract

In this era of concerted genome sequencing efforts, biological sequence information is abundant. With many prokaryotic and simple eukaryotic genomes completed, and with the genomes of more complex organisms nearing completion, the bioinformatics community, those charged with the interpretation of these data, are becoming concerned with the efficacy of current analysis tools. One step towards a more complete understanding of biology at the molecular level is the unambiguous functional assignment of every newly sequenced protein. The sheer scale of this problem precludes the conventional process of biochemically determining function for every example. Rather we must rely on demonstrating similarity to previously characterised proteins via computational methods, which can then be used to infer homology and hence structural and functional relationships. Our ability to do this with any measure of reliability unfortunately diminishes as the pools of experimentally determined sequence data become muddied with sequences that are themselves characterised with "in silico" annotation.Part of the problem stems from the complexity of modelling biology in general, and of evolution in particular. For example, once similarity has been identified between sequences, in order to assign a common function it is important to identify whether the inferred homologous relationship has an orthologous or paralogous origin, which currently cannot be done computationally. The modularity of proteins also poses problems for automatic annotation, as similar domains may occur in proteins with very different functions. Once accepted into the sequence databases, incorrect functional assignments become available for mass propagation and the consequences of incorporating those errors in further "in silico" experiments are potentially catastrophic. One solution to this problem is to collate families of proteins with demonstrable homologous relationships, derive a pattern that represents the essence of those relationships, and use this as a signature to trawl for similarity in the sequence databases. This approach not only provides a more sensitive model of evolution, but also allows annotation from all members of the family to contribute to any assignments made. This thesis describes the development of a new search method (FingerPRINTScan) that exploits the familial models in the PRINTS database to provide more powerful diagnosis of evolutionary relationships. FingerPRINTScan is both selective and sensitive, allowing both precise identification of super-family, family and sub-family relationships, and the detection of more distant ones. Illustrations of the diagnostic performance of the method are given with respect to the haemoglobin and transfer RNA synthetase families, and whole genome data.FingerPRINTScan has become widely used in the biological community, e.g. as the primary search interface to PRINTS via a dedicated web site at the university of Manchester, and as one of the search components of InterPro at the European Bioinformatics Institute (EBI). Furthermore, it is currently responsible for facilitating the use of PRINTS in a number of significant annotation roles, such as the automatic annotation of TrEMBL at the EBI, and as part of the computational suite used to annotate the Drosophila melanogaster genome at Celera Genomics.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as XMLJSONCSV

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item