Análise funcional de eventos de Splicing alternativo

• Ano de defesa: 2015
• Autor(a) principal: Coelho, Vitor Lima
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Laboratório Nacional de Computação Científica; Serviço de Análise e Apoio a Formação de Recursos Humanos; Brasil; LNCC; Programa de Pós-Graduação em Modelagem Computacional
• Palavras-chave em Português: Bioinformática; Domínios de proteinas; Splicing alternativo; Domain proteins; Alternative Splicing; Bioinformatics; CNPQ::CIENCIAS BIOLOGICAS; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://tede.lncc.br/handle/tede/212

Alternative splicing (AS) is a mechanism that produces more than one gene product at the transcriptional level, by combining different exons of a gene, generating a major part of the proteome diversity in eukaryotes. Over the last decade, as the regulatory role of splicing has become more and more evident, AS turned a key factor in creating different organism complexity given a rather constant repertoire of genes across generations. By inserting, deleting or substituting part of the transcript sequence, AS can obviously also have an impact on functional protein domains. Despite some attempts that have mainly been hampered by technical issues caused by the redundancy in alternative transcript sequences, the large-scale effects of AS on the functional level has been poorly studied so far. This project describes the development of a computational tool called ASTAFUNK - Alternative Splicing Trancriptional Analyses with FUNctional Knowledge - an automated and efficient stand-alone program to study how diversity of a custom transcriptome translates into functional variation, based on standard transcriptome annotations (in GTF, Gene Transfer Format) and domain profiles (in Pfam format). In a nutshell, \afunk{} translates the alternatively spliced parts of open reading frames on the fly into amino acid sequences, which subsequently are aligned with the profile Hidden Markov Models from Pfam employing standard dynamic programming (Viterbi's algorithm) with some technical refinements (i.e., a branch-and-bound approach). In contrast to conventional domain prediction tools (e.g., the HMMER aligner), the ASTAFUNK algorithm has been designed to avoid redundant sequence scans in AS-enriched transcriptomes. In this work, theoretical and practical aspects of the ASTAFUNK approach are evaluated, and the efficient JAVA implementation is made freely available over the internet under the BSD 3-clause open source license.