Information Technology Reference
In-Depth Information
where p =
2M
N(N
1)
is the connectance, i.e. the average probability that two nodes
are connected. Statistically signicant deviations from these expected values imply
that property is not randomly distributed. The authors introduce dyadicity (D)
and heterophilicity (H) to classify a property's behavior:
D
m
11
m
11
and H
m
10
m
10
(2.4)
a property is called dyadic if D > 1 (antidyadic if H < 1), indicating that the
nodes with property 1 tend to connect more (less) densely among themselves than
expected for random conguration. Similarly the property is heterophilic if H > 1
(heterophobic if H < 1), meaning that the nodes with property 1 have more (fewer)
connections to nodes with property 0 than expected randomly. By studying the
S. cerevisiae PPI network for proteins involved in cellular communications and
signal transduction, Park and Barabasi [92] calculated that this class of proteins
is more dyadic and less heterophilic than expected; this describes the tendency of
proteins belonging to this functional class to have interacting partners of the same
class, and it moreover highlights the modular structure of the class.
2.7. Functional Networks Motifs
Motifs can be dened as a set of nodes in a biological network with specic molecular
functions which are arranged together and perform some 'useful' process. The
behaviors of motifs are generally not separable from the rest of the system and
they constitute only part of a recognizable systems level function. There are several
known motifs in biological networks (e.g. see [93, 94] e.g. in regulatory network
we might identify switches, amplitude lters, oscillators, frequency lters, noise
lters and many others. Shen Orr and collaborators [94] identied three highly
frequent motifs in a regulatory network of the bacterium Escherichia coli. The rst
motif, the feedforward loop, is dened by X, a TF that regulates a second TF Y,
such that both TFs jointly regulate gene or operon Z. Moreover, if X and Y both
positively regulate Z, and X positively (negatively) regulates Y, the feedforward
loop is coherent (incoherent). The second motif, termed single-input module (SIM),
is dened by a set of genes controlled by a single TF. Moreover, the TF controlling a
SIM has the same eect (activation or inhibition) on all the component genes, which
lack any other transcriptional regulation. The third motif, the dense overlapping
regulons, is a layer of overlapping interactions between operons and a group of input
TFs.
By developing an algorithm to evaluate the statistical signicance of all motifs
composed by three and four nodes, Milo et al. [93] analyzed dierent types of
networks. Their datasets were from biology (PPI and nervous system) as well as
from technology (electronic circuits and the world wide web). Dierent motif sets
were found which dier in dierent networks, suggesting that motifs can dene broad
classes of networks, each with specic types of elementary structures. The motifs
