Computational methods to analyze molecular determinants behind phenotypes
Year: 2013
Type: dissertation
Abstract: Phenotype is a collection of an organism's observable features that can be characterized both on individual level and on single cell level. Phenotypes are largely determined by their molecular processes which also explains their inheritance and plasticity. Some of the molecular background of phenotypes can be characterized by inherited genetic variations and alterations in gene expression. The high-throughput measurement technologies enable the measurement of molecular determinants in cells. However, measurement technologies produce remarkable large data sets and the research questions have become increasingly complex. Thus computational methods are needed to discover molecular mechanisms behind the phenotypes. In many cases, analysis of molecular determinants that contribute to the phenotype proceeds by first identifying putative candidates by using a priori information and high-throughput measurements. Then further analysis can focus on most promising molecules. In many cases, the aim is to identify relevant markers or targets from a set of candidate molecules. Often biomedical studies result in a long list of candidate genes, and to interpret these candidates, information on their context in cell functions is needed. This context information can give insight to synergistic effects of molecular machinery in cells when functions of individual molecules do not explain the observed phenotype. In addition, the context information can be used to generate candidates. One of the methods in this thesis provides a computational data integration method that provides a link in between candidate genes from molecular pathways and genetic variants. It uses publicly available biological knowledge bases to systematically create functional context of candidate genes. This approach is especially important when studying cancer, that is dependent of complex molecular signaling. Genotypes associated with inherited disease predispositions have been studied successfully in the past, however, traditional methods are not applicable in wide variety of analysis conditions. Thus, this thesis introduces a method that uses haplotype sharing to identify genetic loci inherited by multiple distantly related individuals. It is flexible and can be used in various settings, also with very limited number of samples. Increasing the number of biological replicates in gene expression analysis increases the reliability of the results. In many cases, however, the number of samples is limited. Therefore, pooling gene expression data from multiple published studies can increase the understanding of the molecular background behind cell types. This is shown in this thesis by an analysis that identifies gene expression differences in two cell types using publicly available gene expression samples from previous studies. Finally, when candidate molecules are available to characterize phenotypes, they can be compiled into biomarkers. In many cases, a combination of multiple molecules serves as a better biomarker than a single molecule. This thesis also includes a machine learning approach that is used to discover a classifier that predicts the phenotype.
Source: Työväentutkimus Vuosikirja
Author: Sirkku Karinen
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W2103026372