Graphics Reference
In-Depth Information
Introduction
1.1
his chapter reviews data visualization techniques that are used to reconstruct ge-
netic networks from genomics data. Reconstructed genetic networks are predicted
interactions among genes of interest and these interactions are inferred from ge-
nomics data, e.g., microarray data or/and DNA sequence data. Genomics data are
generally contaminated and high-dimensional. he dimensionality of a microarray
is the number of genes in it, and it usually numbers in the thousands at least. It is
important to examine and clean data carefully to attain meaningful inferences. hus
visualization tools that are used in the preprocessing of data associated with genetic
network reconstruction are also reviewed.
Withthe advent ofhigh-throughputgenomics andproteomicsdata, simultaneous
interrogation of the status of each component in a cell is now possible. One chal-
lenging issue in the post-genomics era is to understand the biological interaction
networks that are relevant to cell function (Barabási and Oltvai, ). Biological
networks include protein-protein interaction pathways and genetic networks. he
latter include direct interactions or those in which the biological principle is known,
such as transcriptional networks and metabolic networks, as well as subtle ones like
synthetic geneticinteraction networks(Tongetal., ;Tongetal., ),transcrip-
tional compensation interactions (Wong and Roth, ), among others. Although
protein chips and semiautomatic yeast two-hybrid screens are available, they are not
yet as widely used as microarray gene expression data. Hence, in this article we focus
on visualization for genetic network reconstruction from microarray data.
With the abundant information produced by microarray technology, various ap-
proaches to inferring genetic networks have been proposed. Most of them can be
groupedinto three classes:discretevariable models,continuous variable models,and
graph models. he discrete variable models discretize gene expression into a few
states. he dynamics of gene expression may be perceived as transitions of finite
states. Typical discrete variable models are Boolean networks (Liang et al., ;
Akutsua et al., ) and discrete Bayesian networks (for example, Friedman et al.,
). In general, continuous variable models characterize the expression of a gene
or changes in it as a linear or nonlinear function of other genes (for instance, Beal
et al., ).Graph models,forexample Schäfer andStrimmer ( ),depictgenetic
interactions through directedgraphs (“digraphs") instead of characterizing the inter-
actions quantitatively. For exhaustive literature reviews of both static and dynamic
models used to reconstruct genetic networks, we refer the reader to De Jong ( ),
van Someren et al. ( ), and Shieh et al. ( ).
Visualization for Data Preprocessing
1.2
Outlier Detection
1.2.1
Although many microarray data sets in yeast have been made available, microar-
ray experiments conducted under similar treatments are still sparse compared to the
Search WWH ::




Custom Search