Characterization of protein families, sequence patterns, and functional annotations in large data sets

Sammanfattning: Bioinformatics involves storing, analyzing and making predictions on massive amounts of protein and nucleotide sequence data. The thesis consists of six papers and is focused on proteins. It describes the utilization of bioinformatics techniques to characterize protein families and to detect patterns in gene expression and in polypeptide occurrences. Two protein families were bioinformatically characterized - the membrane associated proteins in eicosanoid and glutathione metabolism (MAPEG) and the Tripartite motif (TRIM) protein families.In the study of the MAPEG super-family, application of di?erent bioinformatic methods made it possible to characterize many new members leading to a doubling of the family size. Furthermore, the MAPEG members were subdivided into families. Remarkably, in six families with previously predominantly mammalian members, ?sh representatives were also now detected, which dated the origin of these families back to the Cambrium ”species explosion”, thus earlier than previously anticipated. Sequence comparisons made it possible to de?ne diagnostic sequence patterns that can be used in genome annotations. Upon publication of several MAPEG structures, these patterns were con?rmed to be part of the active sites.In the TRIM study, the bioinformatic analyses made it possible to subdivide the proteins into three subtypes and to characterize a large number of members. In addition, the analyses showed crucial structural dependencies between the RING and the B-box domains of the TRIM memberRo52. The linker region between the two domains, denoted RBL, is knownto be disease associated. Now, an amphipathic helix was found to be acharacteristic feature of the RBL region, which also was used to divide the family into three subtypes.The ontology annotation treebrowser (OAT) tool was developed to detect functional similarities or common concepts in long lists of proteins or genes, typically generated from proteomics or microarray experiments. OAT was the ?rst annotation browser to include both Gene Ontology (GO) and Medical Subject Headings (MeSH) into the same framework. The complementarity of these two ontologies was demonstrated. OAT was used in the TRIM study to detect di?erences in functional annotations between the subtypes.In the oligopeptide study, we investigated pentapeptide patterns that were over- or under-represented in the current de facto standard database of protein knowledge and a set of completed genomes, compared to what could be expected from amino acid compositions. We found three predominant categories of patterns: (i) patterns originating from frequently occurring families, e.g. respiratory chain-associated proteins and translation machinery proteins; (ii) proteins with structurally and/or functionally favored patterns; (iii) multicopy species-speci?c retrotransposons, only found in the genome set. Such patterns may in?uence amino acid residue based prediction algorithms. These ?ndings in the oligopeptide study were utilized for development of a new method that detects translated introns in unveri?ed protein predictions, which are available in great numbers due to the many completed and ongoing genome projects.A new comprehensive database of protein sequences from completed genomes was developed, denoted genomeLKPG. This database was of central importance in the MAPEG, TRIM and oligopeptide studies. The new sequence database has also been proven useful in several other studies.