ABSTRACT
O splicing alternativo é uma das maiores fontes de diversidade genética e a caracterização desta variabilidade é fundamental para que se possa decifrar o transcriptoma. Foi demonstrado que certos genes são alternativamente processados em tecidos tumorais e suas variantes são associadas a progressão tumoral e invasão em diferentes tumores. Portanto, o entendimento da associação do splicing alternativo ao câncer possui um grande valor diagnóstico e terapêutico. Combinamos uma análise computacional de dados de expressão gênica com validações experimentais para a geração de um sistema que seleciona exons associados a tumores em tecidos específicos. Foram definidos critérios para a busca por exons super expressos em tumores específicos a tecidos e foi desenvolvida uma análise estatística para calcular a probabilidade de um exon ser associado a um tumor. Assim, foram selecionados 1295 genes contendo 2878 exons a serem associados a tumor... O grupo final de candidatos inclui 1386 exons pertencendo a 638 genes. Em linhagens celulares tumorais foram confirmados 4 de 10 exons como super-expressos em tumores, cujos protótipos dos mesmos genes (que não possuem os exons) não foram super expressos em tecido tumoral. Em amostras tumorais de pacientes, 5 de 6 exons validados experimentalmente foram confirmados como sendo associados a tumor. Classificações funcionais dos genes candidatos demonstraram que nossa lista final é enriquecida com genes relacionados funcionalmente com câncer. Os candidatos foram validados outra vez através de uma comparação com trabalhos publicados. Este trabalho demonstra a importância da combinação de sistemas de seleção computacionais com validações experimentais. Análises experimentais em larga escala validarão até que nível nossos exons candidatos diferencialmente expressos têm um potencial diagnóstico e/ou terapêutico...(AU)
Altemative splicing is one of the maJOr sources of the transcriptional diversity found in human cells and its characterization is fundamental to decipher the human transcriptome. Certain genes have been shown to be altematively spliced in tumor tissues and their isoforms have been shown to be associated with spreading and progression in severa! human tumors. Therefore, understanding the association between altemative splicing and cancer is of great diagnostic and therapeutic value and will generate a broader understanding of the involvement and regulation of altemative splicing in tumorigênesis. We combined the use of a transcriptome database for a computational analysis of gene expression data with experimental validations in order to develop a system capable of selecting exons that are predominantly represented in tumor samples of different tissues as compared to their respective normal counterparts. A statistical analysis was developed to calculate the probability that an exon was indeed tumor associated and various criteria were defined to decrease the probability of selecting false-positive candidates. This way we selected 1295 genes containing 2878 exons having a an elevated expression levei in tumor samples, which we called tumor-associated exons. The validation of a few candidates was performed by RT -PCR and showed that our list of candidates included cases of exons belonging to genes that are overexpressed in tumors in general, independently of their splicing variants. The selection of such cases was not the object of our study as we were looking for tumor associated variants that could represent cases of differentially regulated altemative splicing in cancer. To increase the probability of finding bona fide regulated splicing variants that were really tumor-associated, we perforrned a Serial Analysis of Gene Expression (SAGE) analysis, excluding those genes that are upregulated in specifíc tumors. Our final group of candidates included 1386 exons belonging to 638 genes. In tumor cell lines, 4 of 10 validated exons were confirrned as over expressed in tumor while their prototype variants (that don't include the candidate exons) were not over expressed in tumor. In patient tumor samples, 5 of 6 experimentally validated exons were confirmed to be tumor associated. Functional classification of our candidate genes showed that our final list is slightly inflated with cancer-related genes. We validated our candidates once more by comparing them with published studies. Our work shows the importance of the combination of computational selection systems with experimental validations. Large scale experimental analyses will validate to what extent our candidate exons might be of therapeutic or diagnostic use (AU)