Métodos estatísticos aplicados à análise de dados de etiqueta de sequência expressa
Arquivos
Data
2011-02-11
Autores
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Federal de Viçosa
Resumo
Pesquisas de Expressed Sequence Tags (ESTs) são uma ferramenta fundamental para identificação de genes em estudos de seqüenciamento de vários organismos. Dado uma amostra preliminar de EST de uma certa biblioteca de cDNA, vários problemas estatísticos de predição podem surgir. Em particular, é de interesse calcular o número de genes, Δ(t), que podem ser descobertos em uma amostra futura de EST t vezes maior que a amostra original. Esta e outras estatísticas, apresentadas por Susko e Roger (2004), tais como cobertura e o número de leituras necessárias para se descobrir um novo gene são úteis para direcionar protocolos de sequenciamento por meio do cálculo do grau de redundância de uma biblioteca de cDNA. Este cálculo visa maximizar a obtenção de genes durante um sequenciamento de ESTs, porém, este ainda é visto como um procedimento de custo elevado e adequações de técnicas para redução de tal custo é de fundamental importância. O presente trabalho tem como objetivo apresentar os aspectos teóricos da metodologia proposta por Susko e Roger (2004), implementá-la computacionalmente no software livre R e principalmente propor uma abordagem bayesiana para a estimação de Δ(t). Toda a metodologia foi aplicada a dois conjuntos de dados: o primeiro diz respeito a duas bibliotecas de cDNA referentes ao organismo Mastigamoeba Balamuthi e o segundo a duas bibliotecas de cDNA referentes à pele de bovinos F2 (Holandês × Gir) infestados pelo carrapato Riphicephalus (Boophilus) microplus. Para os dois conjuntos de dados as estimativas por intervalo obtidas para Δ(t) foram consideravelmente mais precisas quando se utilizou a inferência bayesiana, indicando que a mesma apresenta-se como uma alternativa viável para estudos relacionados ao cálculo da redundância em análises de ESTs.
Expressed sequence tags (ESTs) surveys are a fundamental tools to identify genes in sequencing studies of various organisms. Given a EST preliminary sample from a certain cDNA library, several prediction statistical problems can arise. Particularly, to calculate the number of genes, Δ (t), which may be discovered in a future EST sample t times larger than the original sample is interesting. This and other ststistics, presented by Susko and Roger (2004), such as coverage and number of necessary readings to discover a new gene are useful for direct sequencing protocols by calculating the degree of redundancy of a cDNA library. This calculation seeks to maximize the obtaining of genes during a EST sequencing, however this is still seen as a costly procedure and adequacy techniques for reducing such costs is of fundamental importance. The present work has as objective to present the theoretical aspects of the methodology proposed by Susko and Roger (2004), to implement computationally the methodology in the free software R and mainly to propose a bayesian approach for estimating Δ (t). All the methodology was applied to two data sets: the first concerns two cDNA libraries from Mastigamoeba balamuthi organism and the second concerns two cDNA libraries from skin of F2 (Holstein × Gyr) bovine infested with the ticks Riphicephalus (Boophilus) microplus. For both data sets the interval estimates obtained for Δ (t) were significantly more accurate when the Bayesian inference was used, indicating that it is an aviable alternative for studies related to the calculation of the redundancy in analysis of ESTs.
Expressed sequence tags (ESTs) surveys are a fundamental tools to identify genes in sequencing studies of various organisms. Given a EST preliminary sample from a certain cDNA library, several prediction statistical problems can arise. Particularly, to calculate the number of genes, Δ (t), which may be discovered in a future EST sample t times larger than the original sample is interesting. This and other ststistics, presented by Susko and Roger (2004), such as coverage and number of necessary readings to discover a new gene are useful for direct sequencing protocols by calculating the degree of redundancy of a cDNA library. This calculation seeks to maximize the obtaining of genes during a EST sequencing, however this is still seen as a costly procedure and adequacy techniques for reducing such costs is of fundamental importance. The present work has as objective to present the theoretical aspects of the methodology proposed by Susko and Roger (2004), to implement computationally the methodology in the free software R and mainly to propose a bayesian approach for estimating Δ (t). All the methodology was applied to two data sets: the first concerns two cDNA libraries from Mastigamoeba balamuthi organism and the second concerns two cDNA libraries from skin of F2 (Holstein × Gyr) bovine infested with the ticks Riphicephalus (Boophilus) microplus. For both data sets the interval estimates obtained for Δ (t) were significantly more accurate when the Bayesian inference was used, indicating that it is an aviable alternative for studies related to the calculation of the redundancy in analysis of ESTs.
Descrição
Palavras-chave
Biblioteca de cDNA, Inferência bayesiana, MCMC, cDNA library, Bayesian inference, MCMC
Citação
PAULA, Fernanda Vital de. Statistical methods applied to expressed sequence tag data analisys. 2011. 68 f. Dissertação (Mestrado em Estatística Aplicada e Biometria) - Universidade Federal de Viçosa, Viçosa, 2011.