Access Type

Open Access Embargo

Date of Award

January 2016

Degree Type

Dissertation

Degree Name

Ph.D.

Department

Computer Science

First Advisor

Sorin Draghici

Abstract

The identification of biological processes involved with a certain phenotype, such as a disease or drug treatment, is the goal of the majority of life sciences experiments.

Pathway analysis methods are used to interpret high-throughput biological data to identify such processes by incorporating information on biological systems to translate data into biological knowledge.

Although widely used, current methods share a number of limitations.

First, they do not take into account the individual contribution of each gene to the phenotype in analysis.

Second, most of the methods include parameters of difficult interpretation, often arbitrarily set.

Third, the results of all methods are affected by the fact that pathways are not independent entities, but communicate with each other by a phenomenon referred to as crosstalk.

Crosstalk effects heavily influence the results of pathway analysis methods, adding a number of false positives and false negatives, making them difficult to interpret.

We developed methods that address these limitations by i) allowing for the incorporation of individual gene contributions, ii) developing objective methods for the estimation of parameters of pathway analysis methods, and iii) developing an approach able to detect, quantify, and correct for crosstalk effects.

We show on a number of real and simulated data that our approaches increase specificity and sensitivity of pathway analysis, allowing for a more effective identification of the processes and mechanisms underlying biological phenomena.

Share

COinS