Year: 2012
Type: dissertation
Abstract: An important problem in the early phases of drug-discovery and disease prognosis is to distinguish between genes that are ubiquitously expressed and genes that are preferentially expressed in one or a few tissues.Although several data sources and methods have been published explicitly for this purpose, it is still not evident how to retrieve these genes and how much confidence we can put in the results.We therefore investigate current data with the aim of answering queries, such as: give me all genes that are preferentially expressed in e.g.liver.In this project we perform a meta-analysis on normal human tissues using a computational approach.Four large-scale microarray datasets across a broad range of normal human tissues were selected for the investigation of gene expression patterns.Methods from publications were implemented and tested on genes with known tissue-specificity.The parameters of the methods were optimized in an iterative procedure using these as training genes.Eventually we ended up using the Decision Function and ROKU-SPMa method we modified and implemented.The results from separate datasets were combined using a rule-based score that represents the degree of specificity.Also, the coverage (indicating the presence of genes in the datasets) was used for assessing the confidence of the results.In total we produced a list of 27523 genes annotated with score and coverage.Among these, 1955 were predicted specific with the highest possible score.The results were evaluated using three external databases: TiSGeD (microarray data), TiGER (EST data) and HPA (protein expression data).The 117 top candidates of specific genes, with highest score and highest coverage, were compared to these databases.96.5% of our predictions were supported by at least one of the databases and agreed better with the consensus than the others.This project is carried out within Research and Development Information at AstraZeneca Mölndal.The project was aimed to serve for an in-house integrated knowledge platform called PharmaConnect, which extracts, integrates and analyses knowledge to support systematic, evidence based decision making, regardless of discipline or content source.I would like to give my sincere appreciation to those who supported and helped me through this master thesis project.My gratitude is beyond words, to my supervisor Daniel Dalevi and Marcus Bjäreland, for giving me this great opportunity to carry out a thesis work at AstraZeneca and for their endless support and encouragement throughout the completion of the project.I owe my deepest gratitude to Daniel Dalevi for his patience, motivation, enthusiasm and immense knowledge to drive me into the right direction, inspire me during the dark times and help me with his brilliant ideas.I would like to thank Marcus Bjäreland for providing breakthrough suggestions and the guide me throughout the project.I want to thank Olle Nerman
Author: Jing Guo
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Keywords: tissue-specific gene expression, gene expression profiling, microarray data, meta-analysis, decision function, normal human tissues, tiger, Human Protein Atlas, knowledge integration, drug discovery, disease prognosis, query answering, rule-based scoring, coverage analysis, knowledge platform, AstraZeneca
Open Access status: green
OpenAlex ID: W164181315
