-----------------------------------------------------------------------
                                             BIOINFORMATICS COLLOQUIUM
						College of Science
                                              George Mason University
                       -----------------------------------------------------------------------

					    Multi-Domain Data Analysis
						 
                                                   Jason Kinser

   				            George Mason University

Abstract:
Real world problems produce a plethora of data within multiple domains.  A biological problem, 
for example may produce DNA sequences, proteins, cell images, functions, and gene expressions.  
These different types (domains) of data can be linked through multiple channels.  For example, 
proteins can be compared through alignment, functional similarity, metabolic pathway links, or 
gene ontologies.  Data that has multiple domains and multiple types of connections requires a different 
approach to data analysis.  Traditional approaches create a database for each domain and queries in 
one domain are unaffected by the presence of data in the other domains.  The philosophy of this talk 
is that this is an incorrect approach, and a new method is presented that combines the data from multiple 
domains into a single search space.  Within this space it is possible to establish multi-domain clusters, 
multi-domain associative memories, and multi-domain predictions.