Modeling protein assemblies: Critical Assessment of Predicted Interactions (CAPRI) 15 years hence.

Protein–protein interactions and protein assemblies, which often also include other macromolecular components, play a crucial role in all cellular processes. Their dysregulation or disruption often leads to disease. Characterizing these interactions and understanding the principles that governs them is therefore more than ever at the center stage of today’s molecular biology. Much of what we know about protein complexes is derived from data on the three-dimensional (3D) structures of these complexes determined by experimental methods. But the number of protein assemblies for which detailed structural information is available represents only a small fraction of the protein assemblies in the cell that can be detected by proteomics and other methods. On the other hand, structural biology has been very successful in mapping out the structural repertoire of individual proteins, many of which are the building blocks of larger complexes. Therefore, computational methods for modeling protein complexes by using available information on the structure of protein components and the fast-growing data on protein sequences have a very important role to play in populating the uncharted landscape of protein assemblies. The Critical Assessment of Predicted Interactions (CAPRI) (http://www.ebi.ac.uk/msd-srv/capri/), a community-wide initiative established in 2001, exactly 15 years ago, has been a key factor in fueling progress in this endeavor. It has done so by offering computational biologists the opportunity of testing their algorithms in blind predictions of experimentally determined 3D structures of protein complexes, the “targets,” provided to CAPRI prior to publication. Due to the paucity of available targets, a CAPRI prediction Round is initiated each time a target (or a few targets) become available, and completed 3 to 6 weeks later. Registered participants are invited to predict the 3D structure of the target protein assembly starting from sequence information alone, or from the unbound structures when those are available. A Round also includes a scoring challenge in which the correct assembly mode(s) must be identified out of a pool of decoys (incorrect models). Groups can participate in either or both challenges (as predictors and scorers). Upon completion of a Round, models submitted by all groups are evaluated against the target structure and ranked using a set of criteria established in close collaboration with the CAPRI community. The evaluation is likewise performed blindly, by concealing the identity of the participants submitting the models from the independent team that performs the assessment. To maintain strict confidentiality, only the assessment team has access to the target coordinates until their release by the authors. Since its inception, the focus of CAPRI has expanded significantly, to include complexes of proteins with other large molecules, including peptide, nucleic acids, and sugars. Several CAPRI Rounds have also focused on the problems of predicting protein binding affinities and modeling the positions of water molecules in protein–protein interfaces. To this day, 35 CAPRI prediction Rounds were completed with a total of 109 targets. Results for Rounds 1–27 were presented at 5 Assessment Meetings, held in 2002, 2004, 2007, and 2009, 2013 across Europe. This issue describes the prediction results for the 20 targets of Rounds 28–35 that took place during the years 2013–2016. These results were presented at the sixth CAPRI evaluation meeting held April 2016 in Tel-Aviv, Israel (http://www.cs.tau.ac.il/conferences/CAPRI2016/) organized by Dr. Haim Wolfson (Tel-Aviv University) and Dr. Ora Schueler-Furman (Hebrew University of Jerusalem). One of these Rounds (Round 30) was a prediction experiment run jointly by CAPRI and CASP (Critical Assessment of Structure Prediction) during the summer of 2014. The prediction results for the 25 targets of that Round, reported in the issue of Proteins dedicated to