Computational Intelligence - November 2012 - 20

Guest
Editorial

Clare Bates Congdon
University of Southern Maine, USA
Gary B. Fogel
Natural Selection, Inc., USA

Special Issue on Computational Intelligence
in Bioinformatics

he field of biology is transforming
from a purely wet-lab oriented
science to one with an expanding
computational component. As our ability to sequence DNA, measure the concentration of RNA and proteins in the
cell, and investigate other molecular
characteristics improves, the amount of
available genomic data is increasing at a
rate that exceeds Moore's Law, and the
difficulty associated with data analysis
increases as well. For example, the
sequencing of an entire human genome
was a landmark achievement when first
accomplished in 2000, but is now far
more routine. Over 60 human genomes
have been fully sequenced, and new
sequencing machines are offered for sale
to industry for full-genome sequencing
at a price 5-to-6 orders of magnitude
cheaper than in 2000. Within just one
decade, the landscape of DNA sequencing and our ability to generate tremendous amounts of biological data has
changed dramatically.
However, despite our increasing
ability to collect genetic data in various
forms, our ability to analyze and make
decisions with this data still pales in
comparison. Only a handful of companies offer analytical services to interpret
full genomes, and most physicians still
rely on key biomarkers (or their own
knowledge and intuition) for their
decisions. We remain a long way from
the seemingly elusive goal of being able
to understand data at the same rate it is

Digital Object Identifier 10.1109/MCI.2012.2215120
Date of publication: 16 October 2012

generated. Furthermore, our understanding is ultimately required at multiple levels-from the genome to
proteins, to cells, tissues, organs, and
humans, and even how these interact
with environmental factors-before we
truly understand how the biological
"system" works.
One thing remains clear, however:
biology is inherently a nonlinear process. To model biology appropriately
over all of these levels and especially in
light of so much data, requires a modeling approach that can also handle
nonlinear dynamic processes. This is
where computational intelligence has a
great deal to offer in the field of bioinformatics, the discipline that is expanding the use of biological, medical,
behavioral, or health data, from data
acquisition to organization, analysis,
and visualization [1]. Approaches for
feature selection, pattern recognition,
fuzzification, and optimization, all play
important roles towards this goal.
There are multiple IEEE groups
and venues that address Bioinformatics-related research. While not specific
to computational intelligence, the
IEEE has a new Life Sciences Initiative, http://lifesciences.ieee.org/.
IEEE is also a cosponsor of the IEEE/
ACM Transactions on Computational
Biology and Bioinformatics journal, with
a 2012 impact factor of 1.543. Specific to the IEEE Computational Intelligence Society, the Bioinformatics and
Bioengineering Task Force (BBTC),
http://cis.ieee.org/bioinfor maticsand-bioengineering-tc.html, organizes

IEEE ComputatIonal IntEllIgEnCE magazInE | novEmbEr 2012

the annual IEEE Symposium on
Computational Intelligence in Bioinformatics and Computational Biology
(CIBCB) as well as special sessions at
conferences such as the IEEE World
Congress on Computational Intelligence (WCCI). We hope that this
special issue will pique your interest
in some of these resources and venues.
The four articles that compose this
special issue of IEEE Computational
Intelligence Magazine highlight different aspects of this endeavor, from
gene expression and genetic association with disease to systems biology
modeling. They were selected through
a rigorous peer-review process.
In the opening article, "Data Mining
Over Biological Datasets: An Integrated
Approach Based on Computational
Intelligence," Stegmayer et al. combine
neural networks with evolutionary
computation for data mining. Self-organized maps (SOMs) are used to identify
patterns of coordinated gene expression.
An evolutionary algorithm is used to
infer pathways between the clusters
identified by the SOMs. The approach
is applied to data from Arabidopsis thaliana, a plant species that is a model
organism for genomics research. The
result helps biologists identify meaningful metabolic pathways in large datasets,
saving research time and expense.
In the second article, "An Analysis
Pipeline with Statistical and VisualizationGuided Knowledge Discovery for Michigan-Style Learning Classifier Systems,"
Urbanowicz et al. combine novel data
visualization with statistical evaluation to

http://lifesciences.ieee.org/ http://cis.ieee.org/bioinformatics

Table of Contents for the Digital Edition of Computational Intelligence - November 2012