Demichelis, Francesca (2005) On information organization and information extraction for the study of gene expressions by tissue microarray technique. UNSPECIFIED thesis, UNSPECIFIED.

Preview

PDF Download (2902Kb) | Preview

Abstract

Genomic expression studies are the means of depicting molecular profiles characterizing specific disease states. Microarrays allow the tracking and the translation of genome sequences into gene functions, leading to the identification of highly informative genes and pathways with a potential impact on understanding disease development and progression. These technologies concurrently may improve diagnostic and treatment modalities and the detection of novel therapeutic targets. Expression array technology is dramatically expanding the amount of data available on many disease states. These studies typically involve many researchers with different backgrounds, each contributing to some steps of the entire process. In particular, Tissue Microarray technology allows for high-throughput expression profiling of tumor samples by evaluating potentially interesting candidate genes and proteins on a large number of well-characterized tumors, providing information on a population basis. High quality experimental data production is extremely important for the reliability of data analysis. Critical assessment of experimental design and organization and reliability assessment of experimental data together with data preprocessing need to be addressed. A technological approach is also advisable to properly manage data heterogeneity, data quantity and user diversity. The focus of this thesis is to develop a systematic approach to processing and better understanding data generated from Tissue Microarray technology, overcoming the limitations of other current approaches. This thesis addresses Tissue Microarray data collection and organization, enhancing data sharing, usability, and process automation. We faced preprocessing issues, identifying critical points and some solutions. We also focused on a specific issue in data classification, proposing a novel classification model based on a Bayesian hierarchical approach, able to handle data uncertainty. Three Tissue Microarray experiments are presented as case studies with the purpose of providing real world examples to illustrate some of the critical points made in this thesis.

Actions (login required)