|
Why do you think your paper is
highly cited?
|
“Our study is a method for
inferring gene regulatory networks from
time-course data.”
|
|
Complex biological systems, including gene regulatory
networks, consist of many interacting components. Since most
of the behaviors are continuous and nonlinear, their
theoretical description requires capturing the essence of
the experimentally observed responses.
Inferring gene regulatory networks from time-course data
of molecular concentrations is known to be a difficult
reverse problem. This work showed that it was a practical
solution for characterizing a gene regulatory network to use
S-system equations and a real-coded genetic algorithm.
Does it describe a new discovery, methodology, or synthesis of
knowledge?
S-system notation is a well-studied form of ordinary
differential equations characterized by power-law formalism
(Michael A. Savageau, "Biochemical
Systems Analysis: A Study of Function and Design in
Molecular Biology," Addison-Wesley,
1976). Since a major
bottleneck in S-system modeling is the number of parameters
to be estimated, which is on the order of O(n2)
where n is the number of variables, we introduced Laplacian
regularization theory as the basis for a rationale for
reducing the number of parameters that must be estimated.
Would you summarize the significance of your paper in
layman’s terms?
Our study is a method for inferring gene regulatory
networks from time-course data. Transcriptome data are often
statistically analyzed by differential expression and
co-expression clustering/classification, because in most
cases, the detailed molecular mechanisms of gene regulations
are not yet well understood.
We propose a genetic algorithm which can be used to infer
the dynamics of a small-scale genetic network from only the
concentration profiles. This algorithm will thus help us
understand the mechanism by which a system emerges as a
result of sophisticated gene regulation.
How did you become involved in this research, and were any
particular problems encountered along the way?
Fortunately, the Computational Biology Research Center (CBRC)
at AIST possesses a 1040-CPU cluster of Pentium III 933-MHz
processors, known as Magi, which is one of the world’s
fastest (ranked in the top 500 in 2001). For our research
calculations, we were allocated 200 CPUs, which accelerated
our work, although the algorithm, of course, can be
calculated using an ordinary PC.
Where do you see your research leading in the future?
The Boolean approximation assumes that the variables
saturate in the ON or OFF position. However, gene expression
levels tend to be continuous rather than binary. In the
future, we hope to develop a hybrid method of discrete
inference for large-scale approximation and dynamic modeling
of the detailed modules, which would allow the prediction of
unknown gene regulatory mechanisms.
Are there any social or political implications for your
research?
Our previous optimization algorithm has been packaged and
released by a domestic bioinformatics company. We hope such
an approach will be widely used to reveal a regulatory
picture from transcriptome data.
Shinichi Kikuchi, Ph.D.
Assistant Professor
Institute for Advanced Biosciences
Keio University
Fujisawa, Japan
|
Return to menu of Fast Moving Fronts
