Visualization in Biology and Medicine - Scientific Visualization: Uncertainty, Multifield, Biomedical, and Scalable Visualization - page 236

comparative genomics research findings, as well as supported the validation of data
and computational algorithms that determine paired features.More information about
MizBee, as well as source code, executables, and example data, can be found at http://
mizbee.org .
22.3 Functional Genomics
The field of functional genomics is, at a high level, about answering the question: how
do genes work together in a cell to perform different functions? Cellular functions,
such as metabolism or reproduction, are controlled by many interrelated chemical
reactions that are catalyzed by genes, or more precisely the products of genes called
proteins. These chemical reactions are complex and highly connected, forming com-
plicated networks that biologist need to discover, unravel, and understand. Finding
differences and similarities in networks from different experimental conditions, in
different cell types, and in different species is an important component of functional
genomics.
22.3.1 Data in Functional Genomics
Scientists working in functional genomics predominately work with two kinds of
data: gene expression and molecular networks. Gene expression is a continuous
measurement of how much a gene is on or off in a cell. It is primarily derived using
microarray technology where the expression levels of many genes are measured at
once. The resulting data is stored as a table of values where most often the rows
are genes and the columns are different samples such as time points, experimental
conditions, tissue types, or species.
Molecular networks are large graphs representing chemical reactions that occur in
a cell. Some of the very well-characterized networks, such as that for metabolism, are
generated from years of careful experimentation. These networks are shared through
curated libraries such as theBioCyc or KEGGdatabases [ 4 , 17 ]. These large networks
are often broken into smaller, more manageable subsets called molecular pathways ,
usually consisting of a dozen reactions or fewer. Other networks, such as protein-
protein interaction networks, are generated using machine-learning algorithms that
look for correlations in large sets of gene expression measurements.
22.3.2 Challenges for Visualization
Whether visualizing gene expression or molecular networks, scale is a major chal-
lenge. For gene expression, the table of measurements can contain thousands of data
points, with current trends moving towards tables with more than two dimensions.
For molecular networks, the graphs can contain many nodes with high connectivity.