The state of the art of protein-protein interaction networks:
The role of the in "silico" approach
Bologna, Italy 8-14 Feb, 2004

Report

The training course organised in Bologna and partially supported by ESF has provided a unique forum for experts and students, mostly young researchers, to meet and fruitfully debate recent and successful achievements in the field of protein-protein interaction networks. The focus was specifically on topics of interactomics, including both an experimental and a theoretical overview of the most recent and promising results. Presently, the biological datasets of physical and functional interactions within proteins in a living cell are rapidly expanding and the amount of data for visualisation and evaluation is daunting. This is specifically due to high throughput proteomics technologies that produce large volumes of protein-protein interaction data organised in networks. Complex functions of living cell are carried out through the concerted activity of many genes and gene products. This activity may be coordinated by the organisation of the genome into regulatory modules or sets of co-regulated genes that share a common function. In the last years high throughput interaction-detection approaches have been introduced that can globally tackle the problem: the yeast two-hybrid system, protein complex purification techniques using mass spectrometry, correlated messenger RNA expression profiles and genetic interaction data. Also computational methods have been developed for interaction predictions derived from gene context analysis.

All these ideas and hypotheses have been analyzed and debated during the forum, in order to highlight the relevance of different experimental and theoretical approaches and that of the critical evaluation of the results out of large scale experiments.

Main topics have been:
· Systems biology and protein-protein interaction networks (Arthur Lesk)
· Experimental methods and their application to different organisms to reveal protein interaction networks (Pierre Legrain, Gianni Cesareni, Anne-Claude Gavin)
· Gene regulation vs protein/DNA binding networks in yeast (Alvis Brazma)
· Fractal networks and their application to biological systems (Guido Caldarelli)
· Function prediction from protein-protein interaction (Manuela Helmer Citterich, Alfonso Valencia, Peer Bork, Christine Brun)
· Prediction of protein-protein interaction (Shoshana Wodak, David Jones, Rita Casadio, Patrick Aloy)
· Regulation of protein complexes and the metabolome (Shoshana Wodak and Janet Thornton)

A major result of the forum was to highlight that several issues are still open. First of all, how reliable are the experimental data and how they can be benchmarked against a trusted reference set. Also how to perform comparative assessment of large data sets of protein-protein interactions and how to interpret it. The Bologna school focused on the problem of data analysis and their representation, as well as on the problem of function prediction from protein-protein interaction networks. Specific questions were addressed: is it feasible to integrate large scale data sets with individual data of protein-protein interaction, including structural data? Do we have interaction networks with enough internal structure to detect/predict "functional modules"? Answers are still elusive, however students had the opportunity to debate the reliability of the methods that are presently available for the prediction of protein-protein interactions, their accuracy and limits with the purpose of stirring new ideas and prompting new developments, both theoretically and practically. This highlighted the role of predictive methods for protein-protein interaction networks, with benefits for the all-scientific community involved in studies of functional genomics and proteomics.

At the completion of the course, participants have been exposed to the presently most debated subjects of protein network prediction including the basic problems related to wide scale prediction of entire genomes, the problem of protein annotation and the most recent results out of the international experiment CAPRI for protein-protein interaction. Furthermore lectures have been generally given in such a way to give first an introduction to the subject and then a more specific overview depending on the speaker's research field. By this approach, also those students who were new to the field could grasp some basic principles before entering the details eventually related to the handling of some specific methods. All the lecturers gave enough information and provided examples to elucidate their own work in a comprehensive and explicative way. We opened a web site where speakers were asked to kindly deposit their presentation so that the students could trace back information, useful schemes and web addresses (www.biocomp.unibo.it/school/html2004).

A tutorial was held in a local computer lab on program application to analyze protein-protein interaction. Three packages were presented by Wolfram Altenhofen, Manuela Helmer Citterich and Ivan Rossi, respectively.

The atmosphere was quite friendly and this promoted discussions between teachers and students. We also organized an open round table discussion (chaired by Janice Glasgow and Arthur Lesk) that focused on the relevance of Bioinformatics in different experimental fields, and some of the students had the opportunity to discuss their own data (selected abstracts are enclosed). Discussion went on during lunch time and coffee breaks. By this, discussion was therefore promoted and encouraged in a very friendly and open way.

List of speakers

List of Participants