Ferramenta computacional para identificação de micro-organismos com base em assinaturas genômicas

• Ano de defesa: 2015
• Autor(a) principal: Andrighetti, Tahila [UNESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Nucleotídeos; Bioinformática; Entropia; Genoma humano; Micro-organismos; Seqüenciamento de nucleotídeo; Nucleotide sequence
• Link de acesso: http://hdl.handle.net/11449/132017; http://www.athena.biblioteca.unesp.br/exlibris/bd/cathedra/11-11-2015/000851881.pdf

Microbial communities play a crucial role in all ecosystems on Earth since they metabolize essential compounds. Given this relevant role they are investigated in Medicine, Biotechnology, Ecology, Food Sciences among other fields. However, only 1% of all known micro-organisms species can be cultivated in vitro. The unravelling of their functions and taxonomic classification demands the development of new approaches. With the advent of new sequencing strategies, the entire genome of microrganisms on a given habitat can be experimentally extracted, but the fragments obtained are small (<1500 bps), and the data processing remains a huge challenge. The most used metagenomic analysis tools classify the sequences by homology. However, the computational time grows exponentially as the read length decreases. There is an evident need for alternative methods that can analyze metagenomic data quickly and accurately. This study proposes a new bacteria sequences identification method to be used in metagenomic data. The genomes of 2164 bacterial strains were obtained from the GenBank and distributed into test and control sets. Each group was randomly fragmented into sequences of 64, 128, 256, 512, 1024, 2048, and 4096 base pair. The sequences organization measures applied in the reads were: GC content, dinucleotide abundance and diplets, triplets and tetraplets entropy. The average and standard deviation of the control sequences values of each species, genus and families of bacteria were calculated. Combinations of genomic signatures and entropy were performed allowing classifying bacteria sequences into family, genus and species. The performance of the proposed methodology was determined by measuring sensitivity, specificity, accuracy and harmonic mean for the test set. The results indicated that the GC content presented the best performance among the signatures investigated. We also considered combinations of features, the combination considering GC ...