|
In
silico
methods for the description of cellular systems by data and
literature mining, predictions and simulations
This
area focuses on the development of new approaches and computational
tools in the area of functional genomics. Functional genomics
opens up new possibilities and raises new requirements. We
see the direct contacts and exchanges with experimental biologists
that will take place within the framework of this programme
as an excellent opportunity for the bioinformatics community
to access information obtained by the state of the art experimental
technologies and to collect requirements and feedback about
its own work and projects. The current scope of in silico
methods is very broad, covering as it does many different
topics that pave the 'virtual path' leading from sequence
to global function. As a starting point for our activities,
we propose to focus on four emerging domains, with the possibility
of incorporating new methodologies as they appear.
1.
Prediction of protein-protein interactions based on the
analysis of multiple sequence alignments. These methods
are related to earlier developments in sequence analysis and
protein structure prediction in the area of bioinformatics.
Recent advances in molecular biology have provided a vast
amount of genetic information for many different organisms.
One of the most challenging current issues is to establish
the possible interactions between different protein components
at different levels, in what has been called 'neighborhood
relationships'. Rather than focusing on direct physical interactions,
a number of computational efforts have recently addressed
the problem of predicting proteins with general functional
relationships. Functional interactions have been predicted
based on comparisons of the species distributions of gene
pairs. These methods assume that genomes encoding one member
of an interaction pair will necessarily also encode its interacting
partner. Marcotte et al. and Enright et al. predicted protein
interactions for those multidomain proteins presenting a variety
of domain arrangements in different organisms. Even though
these approaches all have promising features, they are still
unable to cope with the complexity and extension of protein
interaction networks in real systems. Much remains to be done,
therefore, in the development of new approaches and integration
of existing ones.
2.
Prediction of protein-protein interactions based on the
study of regulatory and other genomic signals with data provided
by genome analysis and genome comparison applications. Dandekar
et al. identified a relationship between genes that are contiguous
in bacterial chromosomes and proteins that formed part of
protein complexes. A different approach was developed by comparing
the frequencies of neighbouring genes in different genomes
and their relationship to the cellular function of the proteins.
Other studies have addressed the problem of predicting protein
function by studying the distribution and conservation of
genomic structures in different systems. Nevertheless, our
knowledge of the evolutionary forces and processes which play
a role in the organization of genomes is far from perfect,
and general approaches able to capture the relationship between
genomic and functional organization have to be developed.
3.
Extraction of information on protein-protein interactions
by systematic analysis of text sources, based on data mining
and text analysis techniques. Very recently, new approaches
have appeared for the extraction of information on protein-protein
interactions. These initial systems are based on previous
experience in the detection of significant, characteristic
keywords in sets of Medline abstracts referring to protein
families, where the use of statistical methods was sufficient
to generate meaningful results without the further need to
implement syntactical analysis. The challenge ahead is to
incorporate more refined statistical methods together with
other new computational techniques in order to improve the
coverage and accuracy of detected interaction networks. Current
approaches would also be extended beyond protein interactions
to related biological issues, such as DNA-protein interactions,
drug-protein binding, tissue distribution and disease-associated
characteristics. Furthermore, problems in molecular biology
will connect with medical informatics, where access to clinical
records and medical information is currently a demanding issue.
4.
Simulation of the behavior of metabolic and signalling
pathways with techniques that include numerical and logical
descriptions of interactions. It is reasonable to think
that in the near future the amount of genomics and functional
information available will be sufficient to define most cellular
functions and interactions. Once all this information has
been integrated, the molecular biology and bioinformatics
communities will be at the point of taking a new step for
the reconstruction of interaction networks and simulation
of their behavior. Even if little practical work has yet been
done in this direction, we would like to stimulate the introduction
of new ideas by increasing the cross-talk of the different
disciplines mentioned in this programme.
Contacts
within the programme
Yaqoub
Ashhab
Francisco Azuaje
Gerhard
Behre
Soren
Brunak
Raffaele
A. Calogero
Rita
Casadio
W
J Coadwell
Werner
Dubitzky
Franca
Fraternali
Robert
Glen
Alessandro
Guffanti
Roderic
Guigo
John
Hancock
Des
Higgins
Turgay
Ibrikci
Juha
Kere
Colm
J Lowery
John
Mitchell
Vaclav
Paces
Syed Asad Rahman
François
Rechenmann
Mischa
Reinhardt
Alfonso
Valencia
Paul
Van der Vet
H.
J. van der Wijk
Rajani
Kanth Vangala
|
