From single genes to co-expression networks: extracting knowledge
from barley functional genomics
P. Faccioli
1,
*
,
, P. Provero
2,
, C. Herrmann
3
, A.M. Stanca
1
, C. Morcia
1
and V. Terzi
1
1
Istituto Sperimentale per la Cerealicoltura, C.R.A., Via S. Protaso 302, I-29017, Fiorenzuola d’Arda (PC),
Italy (*author for correspondence; e-mail p.faccioli@iol.it);
2
Dipartimento di Genetica, Biologia e
Biochimica, Universita
`
di Torino, Via Santena 5bis, I-10100, Torino, Italy;
3
Laboratoire de Ge
´
ne
´
tique et
Physiologie du De
´
veloppement, IBDM, CNRS/INSERM/Universite
´
de la Me
´
diterrane
´
e, Campus Luminy,
13288, Marseille Cedex 9, France,
These two authors contributed equally to this work
Received 3 February 2005; accepted in revised form 30 May 2005
Key words: barley, co-expression networks, functional genomics, Systems Biology
Abstract
The paper reports an ‘in silico’ approach to gene expression analysis based on a barley gene co-expression
network resulting from the study of several publicly available cDNA libraries. The work is an application
of Systems Biology to plant science: at the end of the computational step we identified groups of potentially
related genes. The communities of co-expressed genes constructed from the network are remarkably
characterized from the functional point of view, as shown by the statistical analysis of the Gene Ontology
annotations of their members. Experimental, lab-based testing has been carried out to check the rela-
tionship between network and biological properties and to identify and suggest effective strategies of
information extraction from the network-derived data.
Introduction
Systems Biology is an emerging discipline focusing
the attention on the structure and dynamics of
cellular functions observed as a whole rather than
isolated parts (Kitano, 2002a). The idea behind
this approach is that of a ‘whole-istic biology’ thus
following the Ludwig von Bertalanffy’s General
System Theory (von Bertalanffy, 1969).
The recent advent of high-throughput technol-
ogy and the exponential increase in computer
power have thus moved biology into a revolution-
ary mode, shifting the focus of molecular biolo-
gists from single genes to whole genomes (Kitano,
2002b).
With a reference to functional genomics, meth-
ods examining mRNA and protein population are
now available that can offer insights into gene
expression characterization (Faccioli et al., 2001,
2002). Moreover, the increasing number of cDNA
libraries from which ESTs are obtained is opening
the way for ‘in silico’ analysis of tissue-specific
transcription patterns (Rudd, 2003).
Particularly, the possibility of exploring gene
function is extremely attractive in a context of
high-throughput data generation and computa-
tional inference based on similarities in gene
expression has been proved to be a valuable tool
for function characterization (Altman and Ray-
chaudhuri, 2001), being a powerful and, obviously,
more feasible alternative to a ‘brute-force’ analysis
of all gene and protein sequences (Wu et al., 2002).
The assumption on which clustering-based data
mining relies is that genes sharing a common
Plant Molecular Biology (2005) 58:739–750 Ó Springer 2005
DOI 10.1007/s11103-005-8159-7