BIOMED Home >> | Who We Are | Faculty | Research | Undergraduate Program | Graduate Programs | Students | Alumni  | Contact Us

Print friendly version of this event. Mail this event to a friend.


Seminar - Academic Conduct and Research Ethics

Seminar - Introduction to Neuroengineering at Drexel

Master's Thesis Defense - A System for Mechanically Testing in-situ Clavicular Fracture Fixation Devices

Ph.D. Research Proposal - Computational Support for Medical Device Usability Analysis: A Formal Methods Approach

Master's Thesis Defense - New Continuous-Flow Total Artificial Heart for Use in Smaller Sized Adults and Pediatric Patients

Ph.D. Thesis Defense - Application of Machine Learning and Functional Data Analysis in Classification and Clustering of Functional Near Infrared Spectroscopy Signal in Response to Noxious Stimuli

EVENTS Archive
Ph.D. Thesis Defense - Identification of Activation of Transcription Factors from Microarray Data
Date: March 5, 2007
Time: 2:00 PM
Location: Bossone Research Enterprise Center, Room: 702

Andrew Kossenkov
Advisors: Aydin Tozeren, Ph.D. and Michael Ochs, Ph.D.

Signaling pathways play a critical role in cell survival and development by regulation of transcription factor activity causing necessary gene products to be produced in response to different stimuli. Although the task of detecting activities of signaling pathways is extremely difficult, recent advances in microarray technology promise progress in the field. There are many clustering and pattern recognition algorithms that have been applied to analysis of microarray data. However, these methods lack an ability to address the biological nature of the data and force assignment of one gene to a single co-expression group, while ignoring the fact that many individual genes are regulated by different signaling pathways in response to different stimuli, and therefore the genes should be assigned to multiple groups of co-expression. Another issue in microarray analysis is a low signal-to-noise ratio provided by the technology, yet most of the clustering methods do not even take errors of the measurements into consideration.

Bayesian Decomposition is an algorithm that decomposes microarray data into a set of biologically meaningful expression patterns that could be linked to certain signaling pathways and groups of genes that contain these patterns, allowing assignment of one gene to multiple patterns of expression. To address the problem of low signal-to-noise we modified the Bayesian Decomposition algorithm to allow inclusion of prior gene co-regulation information to improve statistical power. We also created the Automated Sequence Annotation Pipeline to provide microarray data mining processes with annotation information at all steps and particularly to deduce the co-regulation information for a given set of genes from transcription factor database TRANSFAC.

We validated enhancements done to Bayesian Decomposition on simulated and real biological data and showed that using co-regulation information can improve ability of the method to recover correct results. The designed data mining process that uses the Automated Sequence Annotation Pipeline and the modified Bayesian Decomposition was applied to determine transcription factor activities linked to patient outcome in gastrointestinal stromal tumor (GIST) patients undergoing treatment with imatinib mesylate (IM, Gleevec). The study demonstrates genes that can be potentially used as biomarkers to predict GIST patient response to Gleevec treatment and activity of transcription factors that can contribute to difference in the response.


The Bossone Research Enterprise Center is located at the corner of 32nd and Market Streets.

Phone 215.895.2215 | Fax 215.895.4983 | Email
Copyright 2015, Drexel University, All Rights Reserved.