The 4th meeting on the Critical Assessment of Predicted Interaction (CAPRI) held at the Mare Nostrum, Barcelona

The Critical Assessment of PRedicted Interaction (CAPRI) is a community-wide effort, aimed at testing the performance of protein docking methods in blind prediction. CAPRI was initiated in 2001 and modeled after Critical Assessment of methods for Structure Prediction (CASP). Both experiments rely on the good will of structural biologists who provide target structures before publication. As structural determination of protein complexes comes at a slower and an unpredictable pace, CAPRI opens prediction rounds whenever appropriate targets become available. Groups developing docking or scoring methods are invited to predict the structure of target protein complexes by docking and scoring the corresponding component structures. The predicted complexes are assessed for each round, and the results are made publicly available online at the CAPRI web site, managed initially by Kim Henrick and more recently by Sameer Velankar at the European Bioinformatics Institute (Hinxton, UK) (http://www.ebi.ac.uk/msd-srv/capri/). Every 2–3 years, whenever a sufficient number of targets have been assessed, the results are reviewed and the prediction performance of different groups is evaluated and discussed. Three such periodic evaluations have taken place previously. Rounds 1–2 of CAPRI, with a total of seven targets were evaluated in 2002 in a meeting at La Londe des Maures on the Mediterranean coast of France. Rounds 3–5 with a total of 10 targets were evaluated in 2005, in a meeting in Gaeta, Italy, a sea port south of Rome. Rounds 6–12 with a total of nine targets were evaluated in 2007, in a meeting in the city of Toronto on the shores of Lake Ontario, Canada. Each of these evaluations was the subject of a special issue of this Journal (Proteins 2003:52, Proteins 2005:60, and Proteins 2007:69). The present issue of Proteins is dedicated to the fourth CAPRI evaluation of rounds 13–19 that took place over the period of 2007–2009 and comprised a total of 14 targets. This evaluation was completed a few months ago and presented at a meeting in the Mediterranean city of Barcelona, Spain. This multicultural and cosmopolitan city is a metaphor of the diversity of targets and high level of participation that this CAPRI edition achieved. Perhaps not by chance, Mare Nostrum (Latin for Mediterranean Sea) is also the name of the supercomputer hosted by the local organizer Barcelona Supercomputing Center (BSC), and a proof of the increasingly commitment of computational resources to structure prediction. A total of 64 groups and 12 automatic servers submitted 4420 docking predictions to the CAPRI web site. In addition, a total of 41 groups and four servers also participated in the scoring experiment, submitting 1296 predictions. The article of J. Janin (Université Paris-Sud, Orsay, France) describes the targets of this edition, more challenging than ever. Some of the targets were a première in CAPRI, like the protein/nucleic acids interactions or the domain–domain assembly. In addition, the number of cases that required homology-based modeling of (at least) one of their subunits has significantly increased with respect to past editions. Altogether, this collection of targets is giving a more realistic picture of the new challenges in the field. The article by S. Wodak (Hospital for sick children and University of Toronto, Canada) and M. Lensink (Université Libre de Bruxelles, Belgium) explains the assessment of the results by comparison with the crystallographic structures. Some of the targets appear to be clearly easy, with correct predictions produced by many of the docking methods. For most of the targets, the main difficulties arise from the conformational flexibility and/or the need for biological information. The other contributions to this issue are by predictor groups. This is the CAPRI edition with the largest number of targets and participant groups so far. However, in spite of the higher number of participants, the