Bayesian Pathway Analysis
  • FENS
  • Bayesian Pathway Analysis

You are here

Title  : Bayesian Pathway Analysis
Speaker: Hasan H. Otu (Harvard Medical School/Bilgi University)
Date   : Oct 31, 2012
Time   : 13:40
Place  : FENS L035

Please find below the abstract of the talk and a short bio of our speaker.

Kind regards,

Husnu

Abstract:
High Throughput Biological Data (HTBD) production has been increasing at an unprecedented pace with the advancements of microarrays and next-gen sequencing technologies. From a life science perspective HTBD data analysis results make most sense when interpreted within the context of biological networks and pathways. Bayesian Networks (BN) represent dependency structure for a set of random variables using directed acyclic graphs and have been used with increasing popularity in mathematics and computational sciences over the past 20 years. BNs model both linear and non-linear interactions, handle stochastic events in a probabilistic framework accounting for noise, and emphasize only strong relations in noisy data. These properties make BNs excellent candidates for HTBD analysis. In applications of BNs to HTBD analysis, generally, nodes represent genes and edges represent interaction relations.

In this talk I will describe a method we have devised, Bayesian Pathway Analysis (BPA), with applications to synthetic and real data. In the BPA approach, known biological pathways are modeled as BNs and pathways that best explain given HTBD are found. "Gene Set Enrichment" (GSE) or "Gen Ontology" (GO) based approaches that analyze microarray data within the context of pathways or functional groups consider the genes in a pathway or group as a "list", calculate some sort of a score for each list representing the pathway's or group's significance without involving in their model the topology via which genes in a given pathway or group interact with each other. Proposed method, for the first time, integrates pathway topology (graph representing gene interactions) when analyzing HTBD within the context of pathways. BPA tests fitness of the HTBD to the pathways (which are modeled as BNs) through the Bayesian Dirichlet Equivalent (BDe) scoring scheme. Significance of the scores are assessed using randomization via bootstrapping and False Discovery Rate (FDR) corrected p-values are calculated for each pathway accounting for multiple hypothesis testing.

Bio:
Hasan H. Otu obtained his BS degree in 1996 and MS degree in 1997, both from Bogazici University, Department of Electrical and Electronics Engineering. In 2002, he graduated from the University of Nebraska-Lincoln with a PhD in Electrical Engineering. He is a faculty member at Harvard Medical School (2003 - ) where he was a research fellow between 2002-2003. Dr. Otu is the founding director of Bioinformatics Core at Beth Israel Deaconess Medical Center, Harvard Medical School and Associate Director of Proteomics Core at Dana Farber Harvard Cancer Center. Since 2010, Dr. Otu has been acting as the founding chair of Department of Bioengineering at Istanbul Bilgi University. Dr. Otu's research interests are in the area of Bioinformatics focusing on macromolecular sequence analysis, microarrays, biomarker discovery and systems biology, analyzing high throughput biological data within the context of networks.