Information Technology Reference
In-Depth Information
biological processes at a systems level. A protein-protein interaction network (PIN)
consists of all reported protein-protein interactions in an organism detected us-
ing specic assays, such as yeast two-hybrid (now implemented in bacterial systems
too [40]), immuno-precipitation and tandem-anity purication [41{43]. In general,
the data are only of a qualitative nature; that is, interactions are either present or
not but their strength is not quantied. The collection, verication and validation
of such data pose considerable statistical challenges, and together form an active
eld of bioinformatics research.
2.4.1. Experimental approach
The advent of the yeast two-hybrid (Y2H) system in 1989 marked a milestone in
proteomics. Exploiting the modular nature of TFs, Y2H allows measurements of
the activation of reporter genes based on interactions between two chimeric pro-
teins of interest. The Y2H system is today increasingly used in high-throughput
applications intended to map genome-scale PPIs from viruses to humans. Although
some signicant technical limitations apply, Y2H has made a great contribution to
our general understanding of the topology of interaction networks [44]. In addition
to Y2H, especially in the past decade a variety of methods have been developed
to detect and quantify PPIs, including surface plasmon resonance spectroscopy,
NMR, peptide tagging combined with mass spectrometry and uorescence-based
technologies (see Table 2.1).
2.4.2. Computational approach
The prediction of protein-protein interactions is certainly amongst the most am-
bitious targets of modern bioinformatics. The increase of available genomic se-
quences has boosted the progress for detecting functions of genes and proteins and
also allowed the development of complementary approaches to the high throughput
experimental methods. Several approaches attempt to reconstruct the PIN and
the most recent methods can be dened as based on the genomic-context [63], with
four major families: methods based on phylogenetic proles, methods based on gene
clustering, gene fusion methods, and gene neighbor methods. Each of them mainly
takes advantage of one specic genomic feature but in recent times there have been
proposals of methods exploiting an integration of dierent genomic features two of
which are the joint observation method and the algorithm implemented in the web
server STRING. The joint observation method selects the PPIs that are predicted
or identied by more than one method. Its rationale is based on the understanding
that the condence of PPI prediction relies on the amount of supporting evidence,
and that the condence increases with more evidence (i.e., methods). STRING cal-
culates a combined score for each pair of proteins assuming that the features from
various sources are independent [64].
