Functional inference from molecular networks in systems biology.

摘要

Biological entities like genes and proteins share complex relationships. Discerning how these cellular components work in cohesion to carry out various functions is a key challenge that many biologists face today. High throughput technologies now measure many aspects of cellular processes from whole genome sequences to dynamic quantities like mRNA quantities using gene chips, protein concentrations using protein chips as well as physical interactions among genes and proteins. Understanding the behavior of these interactions is not only the key to understanding diseases like cancer, but also has valuable applications in areas like metabolic engineering. Elucidation of the behavior of biological networks requires the development and application of various computational tools and techniques.;In this disseration we introduce a variety of techniques and computational tools that can be used to analyze biomolecular datasets, specifically, datasets of biomolecular interactions. The first part of this thesis focuses on the development of a "network-biology" framework to model and analyze static graphs comprised of genetic and protein interaction datasets. By employing machine learning approaches to genetic interaction prediction in Saccharomyces cerevisae, we were able to show that protein interaction networks can be a rich informative source for the prediction of genetic interactions. We also developed a novel graph-theoretic metric that can be used to identify bottlenecks in a network. Using the Saccharomyces cerevisae protein interaction network we were able to show that the new metric is positively correlated with the essentiality of a gene/protein. Our next theme focuses on the importance of the analysis of the dynamic properties of the biological networks to better understand their functionality. To this end we have developed a tool that is capable of locating bifurcation points where interesting qualitative behavior can be observed by optimizing the parameter values. Also identification of functional, physical modules can provide an effective and novel way of looking at biological systems. We therefore employed an insilico evolutionary approach to generate a library of network motifs which can be used to deleniate functional motifs in real world biochemical networks.